[{"kagg_id":"KAGG:2610.00003","slug":"2610.00003","title":"A Cross-Embodiment Navigation Stack for Heterogeneous Robot Teams: Planner, Interface Calibration, and Future-Range Perception","authors":[{"name":"Zhaoyu Li","affiliation":"Legged Locomotion Group, Tsinghua University, Beijing, China"},{"name":"Jingyao Sun","affiliation":"Legged Locomotion Group, Tsinghua University, Beijing, China"},{"name":"Yichen Wang","affiliation":"Legged Locomotion Group, Tsinghua University, Beijing, China"},{"name":"Bowen Huang","affiliation":"Legged Locomotion Group, Tsinghua University, Beijing, China"},{"name":"Tianyu Zhou","affiliation":"Eastern Institute of Technology, Ningbo, Zhejiang, China"}],"affiliation":null,"abstract":"Running one navigation stack on a heterogeneous robot team requires more than a planner that accepts different footprints: each robot's velocity interface drifts with payload and battery state, and range perception must be queried along the motions that the robot can actually execute. This paper presents a cross-embodiment navigation stack with three coupled layers that share a single data contract keyed on a versioned configuration descriptor. A configuration-driven local planner built on AgniNav plans over abstract robot configurations; an active interface calibration layer built on GAUGE keeps each robot's command-to-motion model current through planner-conditioned probes triggered by residual drift; and a control-aligned future-range layer built on FutureRay supplies free-range predictions along the directions each embodiment can execute under its calibrated interface. The stack was deployed on a mixed team of a Go2-class quadruped, a B2-class quadruped, and a wheeled robot on a three-floor indoor-adjacent course with doorways, a 12 cm step, and a light-degraded stairwell, during a 14 h campaign and short ablation runs. The full stack reached 88.2 percent end-to-end task success with a 28 ms p95 replanning latency. Removing any single layer lowered success by 3.0 to 10.6 percentage points, although the two recorded ablation summaries disagree on which layer matters most. A simulated humanoid embodiment was onboarded in 2.1 h and succeeded in 94 percent of 10 runs. The paper also characterizes the failure envelope of the stack.","publication_date":"2026-10-05","registered_at":"2026-10-06T09:07:00Z","updated_at":"2026-10-06T11:09:00Z","status":"published","html_path":null,"pdf_path":"articles/2610.00003/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"T. Zang, H. Huang, S. Cheng, and W. Zhang. GAUGE: Planner-Conditioned Active Calibration of Opaque Quadruped Velocity Interfaces. arXiv:2607.02841, 2026.","identifier":"arXiv:2607.02841","url":"https://arxiv.org/abs/2607.02841"},{"text":"T. Zang, S. Wang, H. Huang, and S. Cheng. FutureRay: Control-Aligned Future Range for Agile Quadruped Navigation. arXiv:2608.01197, 2026.","identifier":"arXiv:2608.01197","url":"https://arxiv.org/abs/2608.01197"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Sridhar, D. Shah, C. Glossop, and S. Levine. NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration. arXiv:2310.07896, 2023.","identifier":"arXiv:2310.07896","url":"https://arxiv.org/abs/2310.07896"},{"text":"A. Eftekhar et al.. The One RING: a Robotic Indoor Navigation Generalist. arXiv:2412.14401, 2024.","identifier":"arXiv:2412.14401","url":"https://arxiv.org/abs/2412.14401"},{"text":"J. Zhang et al.. Embodied Navigation Foundation Model. arXiv:2509.12129, 2025.","identifier":"arXiv:2509.12129","url":"https://arxiv.org/abs/2509.12129"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"J. Zhang, Y. Du, X. Guo, S. Sun, X. Liu, Y. Sun, G. Lu, W. Sui, and J. Li. EA-Nav: Learning Safe Visual Navigation Policies with Embodiment Awareness. arXiv:2607.19880, 2026.","identifier":"arXiv:2607.19880","url":"https://arxiv.org/abs/2607.19880"},{"text":"Open X-Embodiment Collaboration et al.. Open X-Embodiment: Robotic Learning Datasets and RT-X Models. arXiv:2310.08864, 2023.","identifier":"arXiv:2310.08864","url":"https://arxiv.org/abs/2310.08864"},{"text":"Octo Model Team et al.. Octo: An Open-Source Generalist Robot Policy. arXiv:2405.12213, 2024.","identifier":"arXiv:2405.12213","url":"https://arxiv.org/abs/2405.12213"},{"text":"R. Doshi, H. Walke, O. Mees, S. Dasari, and S. Levine. Scaling Cross-Embodied Learning: One Policy for Manipulation, Navigation, Locomotion and Aviation. arXiv:2408.11812, 2024.","identifier":"arXiv:2408.11812","url":"https://arxiv.org/abs/2408.11812"},{"text":"D. Fox, W. Burgard, and S. Thrun. The Dynamic Window Approach to Collision Avoidance. IEEE Robotics and Automation Magazine, 4(1):23–33, 1997.","identifier":"doi:10.1109/100.580977","url":"https://doi.org/10.1109/100.580977"},{"text":"S. Macenski, F. Martín, R. White, and J. Ginés Clavero. The Marathon 2: A Navigation System. arXiv:2003.00368, 2020.","identifier":"arXiv:2003.00368","url":"https://arxiv.org/abs/2003.00368"},{"text":"L. Wellhausen and M. Hutter. ArtPlanner: Robust Legged Robot Navigation in the Field. arXiv:2303.01420, 2023.","identifier":"arXiv:2303.01420","url":"https://arxiv.org/abs/2303.01420"},{"text":"Y. Kim, C. Kim, and J. Hwangbo. Learning Forward Dynamics Model and Informed Trajectory Sampler for Safe Quadruped Navigation. arXiv:2204.08647, 2022.","identifier":"arXiv:2204.08647","url":"https://arxiv.org/abs/2204.08647"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"G. Ji, J. Mun, H. Kim, and J. Hwangbo. Concurrent Training of a Control Policy and a State Estimator for Dynamic and Robust Legged Locomotion. arXiv:2202.05481, 2022.","identifier":"arXiv:2202.05481","url":"https://arxiv.org/abs/2202.05481"},{"text":"G. B. Margolis and P. Agrawal. Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior. arXiv:2212.03238, 2022.","identifier":"arXiv:2212.03238","url":"https://arxiv.org/abs/2212.03238"},{"text":"I. Radosavovic, T. Xiao, B. Zhang, T. Darrell, J. Malik, and K. Sreenath. Real-World Humanoid Locomotion with Reinforcement Learning. arXiv:2303.03381, 2023.","identifier":"arXiv:2303.03381","url":"https://arxiv.org/abs/2303.03381"},{"text":"K. Hausman, J. Preiss, G. S. Sukhatme, and S. Weiss. Observability-Aware Trajectory Optimization for Self-Calibration with Application to UAVs. arXiv:1604.07905, 2016.","identifier":"arXiv:1604.07905","url":"https://arxiv.org/abs/1604.07905"},{"text":"E. S. Page. Continuous Inspection Schemes. Biometrika, 41(1/2):100–115, 1954.","identifier":"doi:10.1093/biomet/41.1-2.100","url":"https://doi.org/10.1093/biomet/41.1-2.100"},{"text":"V. Makoviychuk et al.. Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. arXiv:2211.07638, 2022.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. arXiv:2309.14341, 2023.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. arXiv:2309.05665, 2023.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. arXiv:2305.08510, 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"G. Kahn, P. Abbeel, and S. Levine. BADGR: An Autonomous Self-Supervised Learning-Based Navigation System. arXiv:2002.05700, 2020.","identifier":"arXiv:2002.05700","url":"https://arxiv.org/abs/2002.05700"},{"text":"A. Loquercio, E. Kaufmann, R. Ranftl, M. Müller, V. Koltun, and D. Scaramuzza. Learning High-Speed Flight in the Wild. arXiv:2110.05113, 2021.","identifier":"arXiv:2110.05113","url":"https://arxiv.org/abs/2110.05113"},{"text":"A. Kendall and Y. Gal. What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision? arXiv:1703.04977, 2017.","identifier":"arXiv:1703.04977","url":"https://arxiv.org/abs/1703.04977"},{"text":"A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada. Control Barrier Functions: Theory and Applications. arXiv:1903.11199, 2019.","identifier":"arXiv:1903.11199","url":"https://arxiv.org/abs/1903.11199"},{"text":"J. Hu, S. Yuan, G. C. R. Bethala, A. Tzes, and Y. Fang. Learning Adaptive Safety Margins for Visual Navigation. arXiv:2607.18200, 2026.","identifier":"arXiv:2607.18200","url":"https://arxiv.org/abs/2607.18200"},{"text":"M. Tranzatto et al.. CERBERUS: Autonomous Legged and Aerial Robotic Exploration in the Tunnel and Urban Circuits of the DARPA Subterranean Challenge. arXiv:2201.07067, 2022.","identifier":"arXiv:2201.07067","url":"https://arxiv.org/abs/2201.07067"},{"text":"A. Agha et al.. NeBula: Quest for Robotic Autonomy in Challenging Environments; TEAM CoSTAR at the DARPA Subterranean Challenge. arXiv:2103.11470, 2021.","identifier":"arXiv:2103.11470","url":"https://arxiv.org/abs/2103.11470"},{"text":"A. Kadian et al.. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance? arXiv:1912.06321, 2019.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"},{"text":"P. Anderson et al.. On Evaluation of Embodied Navigation Agents. arXiv:1807.06757, 2018.","identifier":"arXiv:1807.06757","url":"https://arxiv.org/abs/1807.06757"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2610.00003","metadata":{"title":"A Cross-Embodiment Navigation Stack for Heterogeneous Robot Teams: Planner, Interface Calibration, and Future-Range Perception","authors":[{"name":"Zhaoyu Li","affiliation":"Legged Locomotion Group, Tsinghua University, Beijing, China"},{"name":"Jingyao Sun","affiliation":"Legged Locomotion Group, Tsinghua University, Beijing, China"},{"name":"Yichen Wang","affiliation":"Legged Locomotion Group, Tsinghua University, Beijing, China"},{"name":"Bowen Huang","affiliation":"Legged Locomotion Group, Tsinghua University, Beijing, China"},{"name":"Tianyu Zhou","affiliation":"Eastern Institute of Technology, Ningbo, Zhejiang, China"}],"publication_date":"2026-10-05","abstract":"Running one navigation stack on a heterogeneous robot team requires more than a planner that accepts different footprints: each robot's velocity interface drifts with payload and battery state, and range perception must be queried along the motions that the robot can actually execute. This paper presents a cross-embodiment navigation stack with three coupled layers that share a single data contract keyed on a versioned configuration descriptor. A configuration-driven local planner built on AgniNav plans over abstract robot configurations; an active interface calibration layer built on GAUGE keeps each robot's command-to-motion model current through planner-conditioned probes triggered by residual drift; and a control-aligned future-range layer built on FutureRay supplies free-range predictions along the directions each embodiment can execute under its calibrated interface. The stack was deployed on a mixed team of a Go2-class quadruped, a B2-class quadruped, and a wheeled robot on a three-floor indoor-adjacent course with doorways, a 12 cm step, and a light-degraded stairwell, during a 14 h campaign and short ablation runs. The full stack reached 88.2 percent end-to-end task success with a 28 ms p95 replanning latency. Removing any single layer lowered success by 3.0 to 10.6 percentage points, although the two recorded ablation summaries disagree on which layer matters most. A simulated humanoid embodiment was onboarded in 2.1 h and succeeded in 94 percent of 10 runs. The paper also characterizes the failure envelope of the stack.","references":[{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"T. Zang, H. Huang, S. Cheng, and W. Zhang. GAUGE: Planner-Conditioned Active Calibration of Opaque Quadruped Velocity Interfaces. arXiv:2607.02841, 2026.","identifier":"arXiv:2607.02841","url":"https://arxiv.org/abs/2607.02841"},{"text":"T. Zang, S. Wang, H. Huang, and S. Cheng. FutureRay: Control-Aligned Future Range for Agile Quadruped Navigation. arXiv:2608.01197, 2026.","identifier":"arXiv:2608.01197","url":"https://arxiv.org/abs/2608.01197"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Sridhar, D. Shah, C. Glossop, and S. Levine. NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration. arXiv:2310.07896, 2023.","identifier":"arXiv:2310.07896","url":"https://arxiv.org/abs/2310.07896"},{"text":"A. Eftekhar et al.. The One RING: a Robotic Indoor Navigation Generalist. arXiv:2412.14401, 2024.","identifier":"arXiv:2412.14401","url":"https://arxiv.org/abs/2412.14401"},{"text":"J. Zhang et al.. Embodied Navigation Foundation Model. arXiv:2509.12129, 2025.","identifier":"arXiv:2509.12129","url":"https://arxiv.org/abs/2509.12129"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"J. Zhang, Y. Du, X. Guo, S. Sun, X. Liu, Y. Sun, G. Lu, W. Sui, and J. Li. EA-Nav: Learning Safe Visual Navigation Policies with Embodiment Awareness. arXiv:2607.19880, 2026.","identifier":"arXiv:2607.19880","url":"https://arxiv.org/abs/2607.19880"},{"text":"Open X-Embodiment Collaboration et al.. Open X-Embodiment: Robotic Learning Datasets and RT-X Models. arXiv:2310.08864, 2023.","identifier":"arXiv:2310.08864","url":"https://arxiv.org/abs/2310.08864"},{"text":"Octo Model Team et al.. Octo: An Open-Source Generalist Robot Policy. arXiv:2405.12213, 2024.","identifier":"arXiv:2405.12213","url":"https://arxiv.org/abs/2405.12213"},{"text":"R. Doshi, H. Walke, O. Mees, S. Dasari, and S. Levine. Scaling Cross-Embodied Learning: One Policy for Manipulation, Navigation, Locomotion and Aviation. arXiv:2408.11812, 2024.","identifier":"arXiv:2408.11812","url":"https://arxiv.org/abs/2408.11812"},{"text":"D. Fox, W. Burgard, and S. Thrun. The Dynamic Window Approach to Collision Avoidance. IEEE Robotics and Automation Magazine, 4(1):23–33, 1997.","identifier":"doi:10.1109/100.580977","url":"https://doi.org/10.1109/100.580977"},{"text":"S. Macenski, F. Martín, R. White, and J. Ginés Clavero. The Marathon 2: A Navigation System. arXiv:2003.00368, 2020.","identifier":"arXiv:2003.00368","url":"https://arxiv.org/abs/2003.00368"},{"text":"L. Wellhausen and M. Hutter. ArtPlanner: Robust Legged Robot Navigation in the Field. arXiv:2303.01420, 2023.","identifier":"arXiv:2303.01420","url":"https://arxiv.org/abs/2303.01420"},{"text":"Y. Kim, C. Kim, and J. Hwangbo. Learning Forward Dynamics Model and Informed Trajectory Sampler for Safe Quadruped Navigation. arXiv:2204.08647, 2022.","identifier":"arXiv:2204.08647","url":"https://arxiv.org/abs/2204.08647"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"G. Ji, J. Mun, H. Kim, and J. Hwangbo. Concurrent Training of a Control Policy and a State Estimator for Dynamic and Robust Legged Locomotion. arXiv:2202.05481, 2022.","identifier":"arXiv:2202.05481","url":"https://arxiv.org/abs/2202.05481"},{"text":"G. B. Margolis and P. Agrawal. Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior. arXiv:2212.03238, 2022.","identifier":"arXiv:2212.03238","url":"https://arxiv.org/abs/2212.03238"},{"text":"I. Radosavovic, T. Xiao, B. Zhang, T. Darrell, J. Malik, and K. Sreenath. Real-World Humanoid Locomotion with Reinforcement Learning. arXiv:2303.03381, 2023.","identifier":"arXiv:2303.03381","url":"https://arxiv.org/abs/2303.03381"},{"text":"K. Hausman, J. Preiss, G. S. Sukhatme, and S. Weiss. Observability-Aware Trajectory Optimization for Self-Calibration with Application to UAVs. arXiv:1604.07905, 2016.","identifier":"arXiv:1604.07905","url":"https://arxiv.org/abs/1604.07905"},{"text":"E. S. Page. Continuous Inspection Schemes. Biometrika, 41(1/2):100–115, 1954.","identifier":"doi:10.1093/biomet/41.1-2.100","url":"https://doi.org/10.1093/biomet/41.1-2.100"},{"text":"V. Makoviychuk et al.. Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. arXiv:2211.07638, 2022.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. arXiv:2309.14341, 2023.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. arXiv:2309.05665, 2023.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. arXiv:2305.08510, 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"G. Kahn, P. Abbeel, and S. Levine. BADGR: An Autonomous Self-Supervised Learning-Based Navigation System. arXiv:2002.05700, 2020.","identifier":"arXiv:2002.05700","url":"https://arxiv.org/abs/2002.05700"},{"text":"A. Loquercio, E. Kaufmann, R. Ranftl, M. Müller, V. Koltun, and D. Scaramuzza. Learning High-Speed Flight in the Wild. arXiv:2110.05113, 2021.","identifier":"arXiv:2110.05113","url":"https://arxiv.org/abs/2110.05113"},{"text":"A. Kendall and Y. Gal. What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision? arXiv:1703.04977, 2017.","identifier":"arXiv:1703.04977","url":"https://arxiv.org/abs/1703.04977"},{"text":"A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada. Control Barrier Functions: Theory and Applications. arXiv:1903.11199, 2019.","identifier":"arXiv:1903.11199","url":"https://arxiv.org/abs/1903.11199"},{"text":"J. Hu, S. Yuan, G. C. R. Bethala, A. Tzes, and Y. Fang. Learning Adaptive Safety Margins for Visual Navigation. arXiv:2607.18200, 2026.","identifier":"arXiv:2607.18200","url":"https://arxiv.org/abs/2607.18200"},{"text":"M. Tranzatto et al.. CERBERUS: Autonomous Legged and Aerial Robotic Exploration in the Tunnel and Urban Circuits of the DARPA Subterranean Challenge. arXiv:2201.07067, 2022.","identifier":"arXiv:2201.07067","url":"https://arxiv.org/abs/2201.07067"},{"text":"A. Agha et al.. NeBula: Quest for Robotic Autonomy in Challenging Environments; TEAM CoSTAR at the DARPA Subterranean Challenge. arXiv:2103.11470, 2021.","identifier":"arXiv:2103.11470","url":"https://arxiv.org/abs/2103.11470"},{"text":"A. Kadian et al.. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance? arXiv:1912.06321, 2019.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"},{"text":"P. Anderson et al.. On Evaluation of Embodied Navigation Agents. arXiv:1807.06757, 2018.","identifier":"arXiv:1807.06757","url":"https://arxiv.org/abs/1807.06757"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2610.00003","slug":"2610.00003","registered_at":"2026-10-06T09:07:00Z","updated_at":"2026-10-06T11:09:00Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2610.00003","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2610.00003","tex_path":"articles/2610.00003/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-10-06T11:24:00Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2610.00003"},{"kagg_id":"KAGG:2610.00002","slug":"2610.00002","title":"Range-Aligned Control Allocation for Agile Quadruped Navigation","authors":[{"name":"Jonas Meier","affiliation":"Robot Dynamics Group, ETH Zurich, Zurich, Switzerland"},{"name":"Diego Navarro","affiliation":"Robot Dynamics Group, ETH Zurich, Zurich, Switzerland"},{"name":"Clara Fischer","affiliation":"Robot Dynamics Group, ETH Zurich, Zurich, Switzerland"}],"affiliation":null,"abstract":"Whole-body controllers for quadrupedal robots typically distribute contact forces without knowledge of where the next obstacle will be, so stability reserves and swing clearance are allocated uniformly rather than where the upcoming motion needs them. This paper presents range-aligned control allocation, which uses a predicted future range field, computed in the style of FutureRay along the control-aligned directions of the robot, to set the weights of a whole-body quadratic program. A softmin over the field locates the predicted closest approach; stance-force weights and a linear load-shift term then bias stabilizing support toward the side opposite that approach, and a barrier-type constraint enforces swing-foot clearance only inside the predicted path corridor. All range-dependent terms scale with a field-contrast statistic that vanishes exactly when the field is uniform, and we prove that the allocator then reduces to the nominal controller, with solution deviation otherwise bounded linearly in the contrast. In an evaluation on ANYmal comprising Isaac Lab simulation across ten scenes with raised obstacles of 15 to 60 cm at 1.0 to 3.0 m/s and hardware runs beside raised obstacles, the method lowered energy per kilometer from 1334 to 1238 J/km relative to range-agnostic allocation, reduced the 95th-percentile solve time from 2.9 to 2.6 ms, and matched the unweighted controller in 99.6 percent of uniform-field evaluations. Its mean clearance margin, however, was lower (20.8 versus 24.0 cm), so the results indicate a tighter, less conservative allocation rather than uniformly larger clearance.","publication_date":"2026-10-03","registered_at":"2026-10-06T11:14:00Z","updated_at":"2026-10-06T14:14:00Z","status":"published","html_path":null,"pdf_path":"articles/2610.00002/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"M. Hutter, C. Gehring, D. Jud, A. Lauber, C. D. Bellicoso, V. Tsounis, J. Hwangbo, K. Bodie, P. Fankhauser, M. Bloesch, R. Diethelm, S. Bachmann, A. Melzer, and M. Hoepflinger. ANYmal - a highly mobile and dynamic quadrupedal robot. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2016.","identifier":"doi:10.1109/IROS.2016.7758092","url":"https://doi.org/10.1109/IROS.2016.7758092"},{"text":"C. D. Bellicoso, F. Jenelten, C. Gehring, and M. Hutter. Dynamic Locomotion Through Online Nonlinear Motion Optimization for Quadrupedal Robots. IEEE Robotics and Automation Letters, 3(3):2261-2268, 2018.","identifier":"doi:10.1109/LRA.2018.2794620","url":"https://doi.org/10.1109/LRA.2018.2794620"},{"text":"D. Kim, J. Di Carlo, B. Katz, G. Bledt, and S. Kim. Highly Dynamic Quadruped Locomotion via Whole-Body Impulse Control and Model Predictive Control. arXiv:1909.06586, 2019.","identifier":"arXiv:1909.06586","url":"https://arxiv.org/abs/1909.06586"},{"text":"R. Grandia, F. Jenelten, S. Yang, F. Farshidian, and M. Hutter. Perceptive Locomotion through Nonlinear Model Predictive Control. IEEE Transactions on Robotics, arXiv:2208.08373, 2023.","identifier":"arXiv:2208.08373","url":"https://arxiv.org/abs/2208.08373"},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. Science Robotics, arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. Science Robotics, arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. Science Robotics, arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"F. Jenelten, J. He, F. Farshidian, and M. Hutter. DTC: Deep Tracking Control. Science Robotics, arXiv:2309.15462, 2024.","identifier":"arXiv:2309.15462","url":"https://arxiv.org/abs/2309.15462"},{"text":"C. D. Bellicoso, C. Gehring, J. Hwangbo, P. Fankhauser, and M. Hutter. Perception-less Terrain Adaptation through Whole Body Control and Hierarchical Optimization. IEEE-RAS International Conference on Humanoid Robots (Humanoids), 2016.","identifier":"doi:10.1109/HUMANOIDS.2016.7803330","url":"https://doi.org/10.1109/HUMANOIDS.2016.7803330"},{"text":"A. Herzog, N. Rotella, S. Mason, F. Grimminger, S. Schaal, and L. Righetti. Momentum Control with Hierarchical Inverse Dynamics on a Torque-Controlled Humanoid. Autonomous Robots, arXiv:1410.7284, 2016.","identifier":"arXiv:1410.7284","url":"https://arxiv.org/abs/1410.7284"},{"text":"A. Escande, N. Mansard, and P.-B. Wieber. Hierarchical Quadratic Programming: Fast Online Humanoid-Robot Motion Generation. The International Journal of Robotics Research, 33(7):1006-1028, 2014.","identifier":"doi:10.1177/0278364914521306","url":"https://doi.org/10.1177/0278364914521306"},{"text":"T. Zang, S. Wang, H. Huang, and S. Cheng. FutureRay: Control-Aligned Future Range for Agile Quadruped Navigation. arXiv:2608.01197, 2026.","identifier":"arXiv:2608.01197","url":"https://arxiv.org/abs/2608.01197"},{"text":"S. Kuindersma, R. Deits, M. Fallon, A. Valenzuela, H. Dai, F. Permenter, T. Koolen, P. Marion, and R. Tedrake. Optimization-based Locomotion Planning, Estimation, and Control Design for the Atlas Humanoid Robot. Autonomous Robots, 40(3):429-455, 2016.","identifier":"doi:10.1007/s10514-015-9479-3","url":"https://doi.org/10.1007/s10514-015-9479-3"},{"text":"C. Gehring, S. Coros, M. Hutter, C. D. Bellicoso, H. Heijnen, R. Diethelm, M. Bloesch, P. Fankhauser, J. Hwangbo, M. Hoepflinger, and R. Siegwart. Practice Makes Perfect: An Optimization-Based Approach to Controlling Agile Motions for a Quadruped Robot. IEEE Robotics and Automation Magazine, 23(1):34-43, 2016.","identifier":"doi:10.1109/MRA.2015.2505910","url":"https://doi.org/10.1109/MRA.2015.2505910"},{"text":"J. Di Carlo, P. M. Wensing, B. Katz, G. Bledt, and S. Kim. Dynamic Locomotion in the MIT Cheetah 3 Through Convex Model-Predictive Control. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2018.","identifier":"doi:10.1109/IROS.2018.8594448","url":"https://doi.org/10.1109/IROS.2018.8594448"},{"text":"S. Caron, Q.-C. Pham, and Y. Nakamura. Stability of Surface Contacts for Humanoid Robots: Closed-Form Formulae of the Contact Wrench Cone for Rectangular Support Areas. IEEE International Conference on Robotics and Automation (ICRA), arXiv:1501.04719, 2015.","identifier":"arXiv:1501.04719","url":"https://arxiv.org/abs/1501.04719"},{"text":"B. Stellato, G. Banjac, P. Goulart, A. Bemporad, and S. Boyd. OSQP: An Operator Splitting Solver for Quadratic Programs. Mathematical Programming Computation, arXiv:1711.08013, 2020.","identifier":"arXiv:1711.08013","url":"https://arxiv.org/abs/1711.08013"},{"text":"F. Jenelten, R. Grandia, F. Farshidian, and M. Hutter. TAMOLS: Terrain-Aware Motion Optimization for Legged Systems. IEEE Transactions on Robotics, arXiv:2206.14049, 2022.","identifier":"arXiv:2206.14049","url":"https://arxiv.org/abs/2206.14049"},{"text":"C. Mastalli, R. Budhiraja, W. Merkt, G. Saurel, B. Hammoud, M. Naveau, J. Carpentier, L. Righetti, S. Vijayakumar, and N. Mansard. Crocoddyl: An Efficient and Versatile Framework for Multi-Contact Optimal Control. IEEE International Conference on Robotics and Automation (ICRA), arXiv:1909.04947, 2020.","identifier":"arXiv:1909.04947","url":"https://arxiv.org/abs/1909.04947"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. Conference on Robot Learning (CoRL), arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"G. B. Margolis, G. Yang, K. Paigwar, T. Chen, and P. Agrawal. Rapid Locomotion via Reinforcement Learning. Robotics: Science and Systems (RSS), arXiv:2205.02824, 2022.","identifier":"arXiv:2205.02824","url":"https://arxiv.org/abs/2205.02824"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. Robotics: Science and Systems (RSS), arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. Conference on Robot Learning (CoRL), arXiv:2211.07638, 2022.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. Conference on Robot Learning (CoRL), arXiv:2309.05665, 2023.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. IEEE International Conference on Robotics and Automation (ICRA), arXiv:2309.14341, 2024.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. Science Robotics, arXiv:2306.14874, 2024.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"P. Fankhauser, M. Bloesch, and M. Hutter. Probabilistic Terrain Mapping for Mobile Robots with Uncertain Localization. IEEE Robotics and Automation Letters, 3(4):3019-3026, 2018.","identifier":"doi:10.1109/LRA.2018.2849506","url":"https://doi.org/10.1109/LRA.2018.2849506"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. Robotics: Science and Systems (RSS), arXiv:2305.08510, 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"L. Wellhausen and M. Hutter. ArtPlanner: Robust Legged Robot Navigation in the Field. Field Robotics, arXiv:2303.01420, 2023.","identifier":"arXiv:2303.01420","url":"https://arxiv.org/abs/2303.01420"},{"text":"M. Tranzatto et al.. CERBERUS in the DARPA Subterranean Challenge. Science Robotics, arXiv:2201.07067, 2022.","identifier":"arXiv:2201.07067","url":"https://arxiv.org/abs/2201.07067"},{"text":"D. Hoeller, L. Wellhausen, F. Farshidian, and M. Hutter. Learning a State Representation and Navigation in Cluttered and Dynamic Environments. IEEE Robotics and Automation Letters, arXiv:2103.04351, 2021.","identifier":"arXiv:2103.04351","url":"https://arxiv.org/abs/2103.04351"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"J. Hu, S. Yuan, G. C. R. Bethala, A. Tzes, and Y. Fang. Learning Adaptive Safety Margins for Visual Navigation. arXiv:2607.18200, 2026.","identifier":"arXiv:2607.18200","url":"https://arxiv.org/abs/2607.18200"},{"text":"D. Fox, W. Burgard, and S. Thrun. The Dynamic Window Approach to Collision Avoidance. IEEE Robotics and Automation Magazine, 4(1):23-33, 1997.","identifier":"doi:10.1109/100.580977","url":"https://doi.org/10.1109/100.580977"},{"text":"N. Ratliff, M. Zucker, J. A. Bagnell, and S. Srinivasa. CHOMP: Gradient Optimization Techniques for Efficient Motion Planning. IEEE International Conference on Robotics and Automation (ICRA), 2009.","identifier":"doi:10.1109/ROBOT.2009.5152817","url":"https://doi.org/10.1109/ROBOT.2009.5152817"},{"text":"X. Zhou, Z. Wang, H. Ye, C. Xu, and F. Gao. EGO-Planner: An ESDF-free Gradient-based Local Planner for Quadrotors. IEEE Robotics and Automation Letters, arXiv:2008.08835, 2021.","identifier":"arXiv:2008.08835","url":"https://arxiv.org/abs/2008.08835"},{"text":"A. D. Ames, X. Xu, J. W. Grizzle, and P. Tabuada. Control Barrier Function Based Quadratic Programs for Safety Critical Systems. IEEE Transactions on Automatic Control, arXiv:1609.06408, 2017.","identifier":"arXiv:1609.06408","url":"https://arxiv.org/abs/1609.06408"},{"text":"A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada. Control Barrier Functions: Theory and Applications. European Control Conference (ECC), arXiv:1903.11199, 2019.","identifier":"arXiv:1903.11199","url":"https://arxiv.org/abs/1903.11199"},{"text":"W. Xiao and C. Belta. Control Barrier Functions for Systems with High Relative Degree. IEEE Conference on Decision and Control (CDC), arXiv:1903.04706, 2019.","identifier":"arXiv:1903.04706","url":"https://arxiv.org/abs/1903.04706"},{"text":"R. Grandia, A. J. Taylor, A. D. Ames, and M. Hutter. Multi-Layered Safety for Legged Robots via Control Barrier Functions and Model Predictive Control. IEEE International Conference on Robotics and Automation (ICRA), arXiv:2011.00032, 2021.","identifier":"arXiv:2011.00032","url":"https://arxiv.org/abs/2011.00032"},{"text":"V. Makoviychuk, L. Wawrzyniak, Y. Guo, M. Lu, K. Storey, M. Macklin, D. Hoeller, N. Rudin, A. Allshire, A. Handa, and G. State. Isaac Gym: High Performance GPU-Based Physics Simulation for Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal, C. Yu, Q. Yu, J. Liu, N. Rudin, D. Hoeller, J. L. Yuan, R. Singh, Y. Guo, H. Mazhar, A. Mandlekar, B. Babich, G. State, M. Hutter, and A. Garg. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. IEEE Robotics and Automation Letters, arXiv:2301.04195, 2023.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2610.00002","metadata":{"title":"Range-Aligned Control Allocation for Agile Quadruped Navigation","authors":[{"name":"Jonas Meier","affiliation":"Robot Dynamics Group, ETH Zurich, Zurich, Switzerland"},{"name":"Diego Navarro","affiliation":"Robot Dynamics Group, ETH Zurich, Zurich, Switzerland"},{"name":"Clara Fischer","affiliation":"Robot Dynamics Group, ETH Zurich, Zurich, Switzerland"}],"publication_date":"2026-10-03","abstract":"Whole-body controllers for quadrupedal robots typically distribute contact forces without knowledge of where the next obstacle will be, so stability reserves and swing clearance are allocated uniformly rather than where the upcoming motion needs them. This paper presents range-aligned control allocation, which uses a predicted future range field, computed in the style of FutureRay along the control-aligned directions of the robot, to set the weights of a whole-body quadratic program. A softmin over the field locates the predicted closest approach; stance-force weights and a linear load-shift term then bias stabilizing support toward the side opposite that approach, and a barrier-type constraint enforces swing-foot clearance only inside the predicted path corridor. All range-dependent terms scale with a field-contrast statistic that vanishes exactly when the field is uniform, and we prove that the allocator then reduces to the nominal controller, with solution deviation otherwise bounded linearly in the contrast. In an evaluation on ANYmal comprising Isaac Lab simulation across ten scenes with raised obstacles of 15 to 60 cm at 1.0 to 3.0 m/s and hardware runs beside raised obstacles, the method lowered energy per kilometer from 1334 to 1238 J/km relative to range-agnostic allocation, reduced the 95th-percentile solve time from 2.9 to 2.6 ms, and matched the unweighted controller in 99.6 percent of uniform-field evaluations. Its mean clearance margin, however, was lower (20.8 versus 24.0 cm), so the results indicate a tighter, less conservative allocation rather than uniformly larger clearance.","references":[{"text":"M. Hutter, C. Gehring, D. Jud, A. Lauber, C. D. Bellicoso, V. Tsounis, J. Hwangbo, K. Bodie, P. Fankhauser, M. Bloesch, R. Diethelm, S. Bachmann, A. Melzer, and M. Hoepflinger. ANYmal - a highly mobile and dynamic quadrupedal robot. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2016.","identifier":"doi:10.1109/IROS.2016.7758092","url":"https://doi.org/10.1109/IROS.2016.7758092"},{"text":"C. D. Bellicoso, F. Jenelten, C. Gehring, and M. Hutter. Dynamic Locomotion Through Online Nonlinear Motion Optimization for Quadrupedal Robots. IEEE Robotics and Automation Letters, 3(3):2261-2268, 2018.","identifier":"doi:10.1109/LRA.2018.2794620","url":"https://doi.org/10.1109/LRA.2018.2794620"},{"text":"D. Kim, J. Di Carlo, B. Katz, G. Bledt, and S. Kim. Highly Dynamic Quadruped Locomotion via Whole-Body Impulse Control and Model Predictive Control. arXiv:1909.06586, 2019.","identifier":"arXiv:1909.06586","url":"https://arxiv.org/abs/1909.06586"},{"text":"R. Grandia, F. Jenelten, S. Yang, F. Farshidian, and M. Hutter. Perceptive Locomotion through Nonlinear Model Predictive Control. IEEE Transactions on Robotics, arXiv:2208.08373, 2023.","identifier":"arXiv:2208.08373","url":"https://arxiv.org/abs/2208.08373"},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. Science Robotics, arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. Science Robotics, arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. Science Robotics, arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"F. Jenelten, J. He, F. Farshidian, and M. Hutter. DTC: Deep Tracking Control. Science Robotics, arXiv:2309.15462, 2024.","identifier":"arXiv:2309.15462","url":"https://arxiv.org/abs/2309.15462"},{"text":"C. D. Bellicoso, C. Gehring, J. Hwangbo, P. Fankhauser, and M. Hutter. Perception-less Terrain Adaptation through Whole Body Control and Hierarchical Optimization. IEEE-RAS International Conference on Humanoid Robots (Humanoids), 2016.","identifier":"doi:10.1109/HUMANOIDS.2016.7803330","url":"https://doi.org/10.1109/HUMANOIDS.2016.7803330"},{"text":"A. Herzog, N. Rotella, S. Mason, F. Grimminger, S. Schaal, and L. Righetti. Momentum Control with Hierarchical Inverse Dynamics on a Torque-Controlled Humanoid. Autonomous Robots, arXiv:1410.7284, 2016.","identifier":"arXiv:1410.7284","url":"https://arxiv.org/abs/1410.7284"},{"text":"A. Escande, N. Mansard, and P.-B. Wieber. Hierarchical Quadratic Programming: Fast Online Humanoid-Robot Motion Generation. The International Journal of Robotics Research, 33(7):1006-1028, 2014.","identifier":"doi:10.1177/0278364914521306","url":"https://doi.org/10.1177/0278364914521306"},{"text":"T. Zang, S. Wang, H. Huang, and S. Cheng. FutureRay: Control-Aligned Future Range for Agile Quadruped Navigation. arXiv:2608.01197, 2026.","identifier":"arXiv:2608.01197","url":"https://arxiv.org/abs/2608.01197"},{"text":"S. Kuindersma, R. Deits, M. Fallon, A. Valenzuela, H. Dai, F. Permenter, T. Koolen, P. Marion, and R. Tedrake. Optimization-based Locomotion Planning, Estimation, and Control Design for the Atlas Humanoid Robot. Autonomous Robots, 40(3):429-455, 2016.","identifier":"doi:10.1007/s10514-015-9479-3","url":"https://doi.org/10.1007/s10514-015-9479-3"},{"text":"C. Gehring, S. Coros, M. Hutter, C. D. Bellicoso, H. Heijnen, R. Diethelm, M. Bloesch, P. Fankhauser, J. Hwangbo, M. Hoepflinger, and R. Siegwart. Practice Makes Perfect: An Optimization-Based Approach to Controlling Agile Motions for a Quadruped Robot. IEEE Robotics and Automation Magazine, 23(1):34-43, 2016.","identifier":"doi:10.1109/MRA.2015.2505910","url":"https://doi.org/10.1109/MRA.2015.2505910"},{"text":"J. Di Carlo, P. M. Wensing, B. Katz, G. Bledt, and S. Kim. Dynamic Locomotion in the MIT Cheetah 3 Through Convex Model-Predictive Control. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2018.","identifier":"doi:10.1109/IROS.2018.8594448","url":"https://doi.org/10.1109/IROS.2018.8594448"},{"text":"S. Caron, Q.-C. Pham, and Y. Nakamura. Stability of Surface Contacts for Humanoid Robots: Closed-Form Formulae of the Contact Wrench Cone for Rectangular Support Areas. IEEE International Conference on Robotics and Automation (ICRA), arXiv:1501.04719, 2015.","identifier":"arXiv:1501.04719","url":"https://arxiv.org/abs/1501.04719"},{"text":"B. Stellato, G. Banjac, P. Goulart, A. Bemporad, and S. Boyd. OSQP: An Operator Splitting Solver for Quadratic Programs. Mathematical Programming Computation, arXiv:1711.08013, 2020.","identifier":"arXiv:1711.08013","url":"https://arxiv.org/abs/1711.08013"},{"text":"F. Jenelten, R. Grandia, F. Farshidian, and M. Hutter. TAMOLS: Terrain-Aware Motion Optimization for Legged Systems. IEEE Transactions on Robotics, arXiv:2206.14049, 2022.","identifier":"arXiv:2206.14049","url":"https://arxiv.org/abs/2206.14049"},{"text":"C. Mastalli, R. Budhiraja, W. Merkt, G. Saurel, B. Hammoud, M. Naveau, J. Carpentier, L. Righetti, S. Vijayakumar, and N. Mansard. Crocoddyl: An Efficient and Versatile Framework for Multi-Contact Optimal Control. IEEE International Conference on Robotics and Automation (ICRA), arXiv:1909.04947, 2020.","identifier":"arXiv:1909.04947","url":"https://arxiv.org/abs/1909.04947"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. Conference on Robot Learning (CoRL), arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"G. B. Margolis, G. Yang, K. Paigwar, T. Chen, and P. Agrawal. Rapid Locomotion via Reinforcement Learning. Robotics: Science and Systems (RSS), arXiv:2205.02824, 2022.","identifier":"arXiv:2205.02824","url":"https://arxiv.org/abs/2205.02824"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. Robotics: Science and Systems (RSS), arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. Conference on Robot Learning (CoRL), arXiv:2211.07638, 2022.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. Conference on Robot Learning (CoRL), arXiv:2309.05665, 2023.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. IEEE International Conference on Robotics and Automation (ICRA), arXiv:2309.14341, 2024.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. Science Robotics, arXiv:2306.14874, 2024.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"P. Fankhauser, M. Bloesch, and M. Hutter. Probabilistic Terrain Mapping for Mobile Robots with Uncertain Localization. IEEE Robotics and Automation Letters, 3(4):3019-3026, 2018.","identifier":"doi:10.1109/LRA.2018.2849506","url":"https://doi.org/10.1109/LRA.2018.2849506"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. Robotics: Science and Systems (RSS), arXiv:2305.08510, 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"L. Wellhausen and M. Hutter. ArtPlanner: Robust Legged Robot Navigation in the Field. Field Robotics, arXiv:2303.01420, 2023.","identifier":"arXiv:2303.01420","url":"https://arxiv.org/abs/2303.01420"},{"text":"M. Tranzatto et al.. CERBERUS in the DARPA Subterranean Challenge. Science Robotics, arXiv:2201.07067, 2022.","identifier":"arXiv:2201.07067","url":"https://arxiv.org/abs/2201.07067"},{"text":"D. Hoeller, L. Wellhausen, F. Farshidian, and M. Hutter. Learning a State Representation and Navigation in Cluttered and Dynamic Environments. IEEE Robotics and Automation Letters, arXiv:2103.04351, 2021.","identifier":"arXiv:2103.04351","url":"https://arxiv.org/abs/2103.04351"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"J. Hu, S. Yuan, G. C. R. Bethala, A. Tzes, and Y. Fang. Learning Adaptive Safety Margins for Visual Navigation. arXiv:2607.18200, 2026.","identifier":"arXiv:2607.18200","url":"https://arxiv.org/abs/2607.18200"},{"text":"D. Fox, W. Burgard, and S. Thrun. The Dynamic Window Approach to Collision Avoidance. IEEE Robotics and Automation Magazine, 4(1):23-33, 1997.","identifier":"doi:10.1109/100.580977","url":"https://doi.org/10.1109/100.580977"},{"text":"N. Ratliff, M. Zucker, J. A. Bagnell, and S. Srinivasa. CHOMP: Gradient Optimization Techniques for Efficient Motion Planning. IEEE International Conference on Robotics and Automation (ICRA), 2009.","identifier":"doi:10.1109/ROBOT.2009.5152817","url":"https://doi.org/10.1109/ROBOT.2009.5152817"},{"text":"X. Zhou, Z. Wang, H. Ye, C. Xu, and F. Gao. EGO-Planner: An ESDF-free Gradient-based Local Planner for Quadrotors. IEEE Robotics and Automation Letters, arXiv:2008.08835, 2021.","identifier":"arXiv:2008.08835","url":"https://arxiv.org/abs/2008.08835"},{"text":"A. D. Ames, X. Xu, J. W. Grizzle, and P. Tabuada. Control Barrier Function Based Quadratic Programs for Safety Critical Systems. IEEE Transactions on Automatic Control, arXiv:1609.06408, 2017.","identifier":"arXiv:1609.06408","url":"https://arxiv.org/abs/1609.06408"},{"text":"A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada. Control Barrier Functions: Theory and Applications. European Control Conference (ECC), arXiv:1903.11199, 2019.","identifier":"arXiv:1903.11199","url":"https://arxiv.org/abs/1903.11199"},{"text":"W. Xiao and C. Belta. Control Barrier Functions for Systems with High Relative Degree. IEEE Conference on Decision and Control (CDC), arXiv:1903.04706, 2019.","identifier":"arXiv:1903.04706","url":"https://arxiv.org/abs/1903.04706"},{"text":"R. Grandia, A. J. Taylor, A. D. Ames, and M. Hutter. Multi-Layered Safety for Legged Robots via Control Barrier Functions and Model Predictive Control. IEEE International Conference on Robotics and Automation (ICRA), arXiv:2011.00032, 2021.","identifier":"arXiv:2011.00032","url":"https://arxiv.org/abs/2011.00032"},{"text":"V. Makoviychuk, L. Wawrzyniak, Y. Guo, M. Lu, K. Storey, M. Macklin, D. Hoeller, N. Rudin, A. Allshire, A. Handa, and G. State. Isaac Gym: High Performance GPU-Based Physics Simulation for Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal, C. Yu, Q. Yu, J. Liu, N. Rudin, D. Hoeller, J. L. Yuan, R. Singh, Y. Guo, H. Mazhar, A. Mandlekar, B. Babich, G. State, M. Hutter, and A. Garg. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. IEEE Robotics and Automation Letters, arXiv:2301.04195, 2023.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2610.00002","slug":"2610.00002","registered_at":"2026-10-06T11:14:00Z","updated_at":"2026-10-06T14:14:00Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2610.00002","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2610.00002","tex_path":"articles/2610.00002/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-10-06T13:31:00Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2610.00002"},{"kagg_id":"KAGG:2610.00001","slug":"2610.00001","title":"A Unified Evaluation Harness for Quadruped Navigation across Interfaces, Terrain, and Sensors","authors":[{"name":"Marcus Hill","affiliation":"Robotics Institute, Georgia Institute of Technology, Atlanta, GA, USA"},{"name":"Rachel Kim","affiliation":"Robotics Institute, Georgia Institute of Technology, Atlanta, GA, USA"},{"name":"Yating Feng","affiliation":"Robotics Institute, Georgia Institute of Technology, Atlanta, GA, USA"},{"name":"Ivan Sokolov","affiliation":"Robotics Institute, Georgia Institute of Technology, Atlanta, GA, USA"}],"affiliation":null,"abstract":"Reported performance of quadruped navigation planners is rarely comparable across papers, because each evaluation couples the planner with a particular velocity interface, terrain definition, and sensor rig. This paper presents a unified evaluation harness that isolates these three axes. The interface axis spans an ideal velocity tracker, a calibrated-opaque interface identified by planner-conditioned active calibration, and two native platform interfaces; the terrain axis comprises six standardized courses (flat, mixed, stairs, narrow, cluttered, rough); and the sensor axis comprises LiDAR-only, LiDAR+vision, and degraded rigs. A fixed set of four planners, with a configuration-driven cross-embodiment planner as the reference stack, is executed on Go2-class, B2-class, and ANYmal-class assets, and every run emits a standardized 214-field results record. Across 6480 standardized runs, the harness reaches 95.7% cross-platform run parity and 100.0% record schema completeness at 2.3 s of overhead per run and measures nine confounding factors, compared with 82.9% parity for raw-score ranking and 87.0% when only the interface axis is held fixed, both of which measure eight. Ablations show that removing the interface, terrain, or sensor-rig axis lowers parity to 74.5%, 80.3%, and 86.4%, respectively. A case study shows that a planner ranked first on raw scores falls to third once the interface axis is held fixed, because its advantage stemmed from a favorable interface on one platform. Recorded failure cases, including sanity-check quarantines and gait-mode switching delays, are reported as harness observations rather than discarded.","publication_date":"2026-10-01","registered_at":"2026-10-06T18:31:00Z","updated_at":"2026-10-06T21:31:00Z","status":"published","html_path":null,"pdf_path":"articles/2610.00001/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"M. Tranzatto et al.. CERBERUS: Autonomous Legged and Aerial Robotic Exploration in the Tunnel and Urban Circuits of the DARPA Subterranean Challenge. arXiv:2201.07067, 2022.","identifier":"arXiv:2201.07067","url":"https://arxiv.org/abs/2201.07067"},{"text":"L. Wellhausen and M. Hutter. ArtPlanner: Robust Legged Robot Navigation in the Field. arXiv:2303.01420, 2023.","identifier":"arXiv:2303.01420","url":"https://arxiv.org/abs/2303.01420"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"G. B. Margolis and P. Agrawal. Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior. arXiv:2212.03238, 2022.","identifier":"arXiv:2212.03238","url":"https://arxiv.org/abs/2212.03238"},{"text":"T. Zang, H. Huang, S. Cheng, and W. Zhang. GAUGE: Planner-Conditioned Active Calibration of Opaque Quadruped Velocity Interfaces. arXiv:2607.02841, 2026.","identifier":"arXiv:2607.02841","url":"https://arxiv.org/abs/2607.02841"},{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"V. Makoviychuk et al.. Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. arXiv:2211.07638, 2022.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. arXiv:2309.05665, 2023.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. arXiv:2309.14341, 2023.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"R. Grandia, F. Jenelten, S. Yang, F. Farshidian, and M. Hutter. Perceptive Locomotion through Nonlinear Model Predictive Control. arXiv:2208.08373, 2022.","identifier":"arXiv:2208.08373","url":"https://arxiv.org/abs/2208.08373"},{"text":"S. Ha, J. Lee, M. van de Panne, Z. Xie, W. Yu, and M. Khadiv. Learning-based Legged Locomotion: State of the Art and Future Perspectives. arXiv:2406.01152, 2024.","identifier":"arXiv:2406.01152","url":"https://arxiv.org/abs/2406.01152"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. arXiv:1710.06537, 2017.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. arXiv:1703.06907, 2017.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"},{"text":"J. Tan, T. Zhang, E. Coumans, A. Iscen, Y. Bai, D. Hafner, S. Bohez, and V. Vanhoucke. Sim-to-Real: Learning Agile Locomotion For Quadruped Robots. arXiv:1804.10332, 2018.","identifier":"arXiv:1804.10332","url":"https://arxiv.org/abs/1804.10332"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"J. Truong, D. Yarats, T. Li, F. Meier, S. Chernova, D. Batra, and A. Rai. Learning Navigation Skills for Legged Robots with Learned Robot Embodiments. arXiv:2011.12255, 2020.","identifier":"arXiv:2011.12255","url":"https://arxiv.org/abs/2011.12255"},{"text":"Z. Fu, A. Kumar, A. Agarwal, H. Qi, J. Malik, and D. Pathak. Coupling Vision and Proprioception for Navigation of Legged Robots. arXiv:2112.02094, 2021.","identifier":"arXiv:2112.02094","url":"https://arxiv.org/abs/2112.02094"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Sridhar, D. Shah, C. Glossop, and S. Levine. NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration. arXiv:2310.07896, 2023.","identifier":"arXiv:2310.07896","url":"https://arxiv.org/abs/2310.07896"},{"text":"T. Miki, L. Wellhausen, R. Grandia, F. Jenelten, T. Homberger, and M. Hutter. Elevation Mapping for Locomotion and Navigation using GPU. arXiv:2204.12876, 2022.","identifier":"arXiv:2204.12876","url":"https://arxiv.org/abs/2204.12876"},{"text":"W. Xu, Y. Cai, D. He, J. Lin, and F. Zhang. FAST-LIO2: Fast Direct LiDAR-Inertial Odometry. arXiv:2107.06829, 2021.","identifier":"arXiv:2107.06829","url":"https://arxiv.org/abs/2107.06829"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. arXiv:2305.08510, 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"M. Mattamala, J. Frey, P. Libera, N. Chebrolu, G. Martius, C. Cadena, M. Hutter, and M. Fallon. Wild Visual Navigation: Fast Traversability Learning via Pre-Trained Models and Online Self-Supervision. arXiv:2404.07110, 2024.","identifier":"arXiv:2404.07110","url":"https://arxiv.org/abs/2404.07110"},{"text":"M. Savva et al.. Habitat: A Platform for Embodied AI Research. arXiv:1904.01201, 2019.","identifier":"arXiv:1904.01201","url":"https://arxiv.org/abs/1904.01201"},{"text":"P. Anderson et al.. On Evaluation of Embodied Navigation Agents. arXiv:1807.06757, 2018.","identifier":"arXiv:1807.06757","url":"https://arxiv.org/abs/1807.06757"},{"text":"N. Yokoyama, S. Ha, and D. Batra. Success Weighted by Completion Time: A Dynamics-Aware Evaluation Criteria for Embodied Navigation. arXiv:2103.08022, 2021.","identifier":"arXiv:2103.08022","url":"https://arxiv.org/abs/2103.08022"},{"text":"A. Kadian et al.. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance? arXiv:1912.06321, 2019.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"},{"text":"D. Perille, A. Truong, X. Xiao, and P. Stone. Benchmarking Metric Ground Navigation. arXiv:2008.13315, 2020.","identifier":"arXiv:2008.13315","url":"https://arxiv.org/abs/2008.13315"},{"text":"X. Xiao et al.. Autonomous Ground Navigation in Highly Constrained Spaces: Lessons Learned from the BARN Challenge at ICRA 2022. arXiv:2208.10473, 2022.","identifier":"arXiv:2208.10473","url":"https://arxiv.org/abs/2208.10473"},{"text":"X. Xiao, B. Liu, G. Warnell, and P. Stone. Motion Planning and Control for Mobile Robot Navigation Using Machine Learning: A Survey. arXiv:2011.13112, 2020.","identifier":"arXiv:2011.13112","url":"https://arxiv.org/abs/2011.13112"},{"text":"M. Mittal et al.. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. arXiv:2301.04195, 2023.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"P. Henderson, R. Islam, P. Bachman, J. Pineau, D. Precup, and D. Meger. Deep Reinforcement Learning that Matters. arXiv:1709.06560, 2017.","identifier":"arXiv:1709.06560","url":"https://arxiv.org/abs/1709.06560"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2610.00001","metadata":{"title":"A Unified Evaluation Harness for Quadruped Navigation across Interfaces, Terrain, and Sensors","authors":[{"name":"Marcus Hill","affiliation":"Robotics Institute, Georgia Institute of Technology, Atlanta, GA, USA"},{"name":"Rachel Kim","affiliation":"Robotics Institute, Georgia Institute of Technology, Atlanta, GA, USA"},{"name":"Yating Feng","affiliation":"Robotics Institute, Georgia Institute of Technology, Atlanta, GA, USA"},{"name":"Ivan Sokolov","affiliation":"Robotics Institute, Georgia Institute of Technology, Atlanta, GA, USA"}],"publication_date":"2026-10-01","abstract":"Reported performance of quadruped navigation planners is rarely comparable across papers, because each evaluation couples the planner with a particular velocity interface, terrain definition, and sensor rig. This paper presents a unified evaluation harness that isolates these three axes. The interface axis spans an ideal velocity tracker, a calibrated-opaque interface identified by planner-conditioned active calibration, and two native platform interfaces; the terrain axis comprises six standardized courses (flat, mixed, stairs, narrow, cluttered, rough); and the sensor axis comprises LiDAR-only, LiDAR+vision, and degraded rigs. A fixed set of four planners, with a configuration-driven cross-embodiment planner as the reference stack, is executed on Go2-class, B2-class, and ANYmal-class assets, and every run emits a standardized 214-field results record. Across 6480 standardized runs, the harness reaches 95.7% cross-platform run parity and 100.0% record schema completeness at 2.3 s of overhead per run and measures nine confounding factors, compared with 82.9% parity for raw-score ranking and 87.0% when only the interface axis is held fixed, both of which measure eight. Ablations show that removing the interface, terrain, or sensor-rig axis lowers parity to 74.5%, 80.3%, and 86.4%, respectively. A case study shows that a planner ranked first on raw scores falls to third once the interface axis is held fixed, because its advantage stemmed from a favorable interface on one platform. Recorded failure cases, including sanity-check quarantines and gait-mode switching delays, are reported as harness observations rather than discarded.","references":[{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"M. Tranzatto et al.. CERBERUS: Autonomous Legged and Aerial Robotic Exploration in the Tunnel and Urban Circuits of the DARPA Subterranean Challenge. arXiv:2201.07067, 2022.","identifier":"arXiv:2201.07067","url":"https://arxiv.org/abs/2201.07067"},{"text":"L. Wellhausen and M. Hutter. ArtPlanner: Robust Legged Robot Navigation in the Field. arXiv:2303.01420, 2023.","identifier":"arXiv:2303.01420","url":"https://arxiv.org/abs/2303.01420"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"G. B. Margolis and P. Agrawal. Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior. arXiv:2212.03238, 2022.","identifier":"arXiv:2212.03238","url":"https://arxiv.org/abs/2212.03238"},{"text":"T. Zang, H. Huang, S. Cheng, and W. Zhang. GAUGE: Planner-Conditioned Active Calibration of Opaque Quadruped Velocity Interfaces. arXiv:2607.02841, 2026.","identifier":"arXiv:2607.02841","url":"https://arxiv.org/abs/2607.02841"},{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"V. Makoviychuk et al.. Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. arXiv:2211.07638, 2022.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. arXiv:2309.05665, 2023.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. arXiv:2309.14341, 2023.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"R. Grandia, F. Jenelten, S. Yang, F. Farshidian, and M. Hutter. Perceptive Locomotion through Nonlinear Model Predictive Control. arXiv:2208.08373, 2022.","identifier":"arXiv:2208.08373","url":"https://arxiv.org/abs/2208.08373"},{"text":"S. Ha, J. Lee, M. van de Panne, Z. Xie, W. Yu, and M. Khadiv. Learning-based Legged Locomotion: State of the Art and Future Perspectives. arXiv:2406.01152, 2024.","identifier":"arXiv:2406.01152","url":"https://arxiv.org/abs/2406.01152"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. arXiv:1710.06537, 2017.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. arXiv:1703.06907, 2017.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"},{"text":"J. Tan, T. Zhang, E. Coumans, A. Iscen, Y. Bai, D. Hafner, S. Bohez, and V. Vanhoucke. Sim-to-Real: Learning Agile Locomotion For Quadruped Robots. arXiv:1804.10332, 2018.","identifier":"arXiv:1804.10332","url":"https://arxiv.org/abs/1804.10332"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"J. Truong, D. Yarats, T. Li, F. Meier, S. Chernova, D. Batra, and A. Rai. Learning Navigation Skills for Legged Robots with Learned Robot Embodiments. arXiv:2011.12255, 2020.","identifier":"arXiv:2011.12255","url":"https://arxiv.org/abs/2011.12255"},{"text":"Z. Fu, A. Kumar, A. Agarwal, H. Qi, J. Malik, and D. Pathak. Coupling Vision and Proprioception for Navigation of Legged Robots. arXiv:2112.02094, 2021.","identifier":"arXiv:2112.02094","url":"https://arxiv.org/abs/2112.02094"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Sridhar, D. Shah, C. Glossop, and S. Levine. NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration. arXiv:2310.07896, 2023.","identifier":"arXiv:2310.07896","url":"https://arxiv.org/abs/2310.07896"},{"text":"T. Miki, L. Wellhausen, R. Grandia, F. Jenelten, T. Homberger, and M. Hutter. Elevation Mapping for Locomotion and Navigation using GPU. arXiv:2204.12876, 2022.","identifier":"arXiv:2204.12876","url":"https://arxiv.org/abs/2204.12876"},{"text":"W. Xu, Y. Cai, D. He, J. Lin, and F. Zhang. FAST-LIO2: Fast Direct LiDAR-Inertial Odometry. arXiv:2107.06829, 2021.","identifier":"arXiv:2107.06829","url":"https://arxiv.org/abs/2107.06829"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. arXiv:2305.08510, 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"M. Mattamala, J. Frey, P. Libera, N. Chebrolu, G. Martius, C. Cadena, M. Hutter, and M. Fallon. Wild Visual Navigation: Fast Traversability Learning via Pre-Trained Models and Online Self-Supervision. arXiv:2404.07110, 2024.","identifier":"arXiv:2404.07110","url":"https://arxiv.org/abs/2404.07110"},{"text":"M. Savva et al.. Habitat: A Platform for Embodied AI Research. arXiv:1904.01201, 2019.","identifier":"arXiv:1904.01201","url":"https://arxiv.org/abs/1904.01201"},{"text":"P. Anderson et al.. On Evaluation of Embodied Navigation Agents. arXiv:1807.06757, 2018.","identifier":"arXiv:1807.06757","url":"https://arxiv.org/abs/1807.06757"},{"text":"N. Yokoyama, S. Ha, and D. Batra. Success Weighted by Completion Time: A Dynamics-Aware Evaluation Criteria for Embodied Navigation. arXiv:2103.08022, 2021.","identifier":"arXiv:2103.08022","url":"https://arxiv.org/abs/2103.08022"},{"text":"A. Kadian et al.. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance? arXiv:1912.06321, 2019.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"},{"text":"D. Perille, A. Truong, X. Xiao, and P. Stone. Benchmarking Metric Ground Navigation. arXiv:2008.13315, 2020.","identifier":"arXiv:2008.13315","url":"https://arxiv.org/abs/2008.13315"},{"text":"X. Xiao et al.. Autonomous Ground Navigation in Highly Constrained Spaces: Lessons Learned from the BARN Challenge at ICRA 2022. arXiv:2208.10473, 2022.","identifier":"arXiv:2208.10473","url":"https://arxiv.org/abs/2208.10473"},{"text":"X. Xiao, B. Liu, G. Warnell, and P. Stone. Motion Planning and Control for Mobile Robot Navigation Using Machine Learning: A Survey. arXiv:2011.13112, 2020.","identifier":"arXiv:2011.13112","url":"https://arxiv.org/abs/2011.13112"},{"text":"M. Mittal et al.. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. arXiv:2301.04195, 2023.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"P. Henderson, R. Islam, P. Bachman, J. Pineau, D. Precup, and D. Meger. Deep Reinforcement Learning that Matters. arXiv:1709.06560, 2017.","identifier":"arXiv:1709.06560","url":"https://arxiv.org/abs/1709.06560"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2610.00001","slug":"2610.00001","registered_at":"2026-10-06T18:31:00Z","updated_at":"2026-10-06T21:31:00Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2610.00001","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2610.00001","tex_path":"articles/2610.00001/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-10-06T20:48:00Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2610.00001"},{"kagg_id":"KAGG:2609.00009","slug":"2609.00009","title":"Adaptive Safety Margins for Quadruped Navigation with Calibrated Velocity Interfaces","authors":[{"name":"Junho Kang","affiliation":"Legged Locomotion and Control Lab, KAIST, Daejeon, South Korea"},{"name":"Hyesun Bae","affiliation":"Legged Locomotion and Control Lab, KAIST, Daejeon, South Korea"},{"name":"Daewon Ryu","affiliation":"Legged Locomotion and Control Lab, KAIST, Daejeon, South Korea"},{"name":"Minju Seo","affiliation":"Legged Locomotion and Control Lab, KAIST, Daejeon, South Korea"}],"affiliation":null,"abstract":"Fixed safety margins for legged navigation are tuned once, so they ignore both how well the robot's velocity interface is currently calibrated and how much free space the robot is about to enter. This paper presents an adaptive safety margin for quadruped navigation that combines the two signals. A calibrated interface model obtained with GAUGE supplies a high-confidence upper bound on the lateral deviation caused by command lag, and the control-aligned future-range predictor FutureRay supplies a lower confidence bound on the free range along the directions the controller can actually execute. The margin is the sum of a kinematic term and a lag-error term, each blended between its model-based value and its worst-case value by an explicit confidence weight. We prove that the margin is monotone nondecreasing in both uncertainty sources, that it never falls below the model-based bounds, and that it reduces exactly to the worst-case fixed margin when both confidences vanish. On a Unitree Go2 evaluated on 12 obstacle-adjacent segments at speeds from 1.0 to 4.0 m/s, with matching Isaac Lab simulation, the adaptive margin raises the collision-free speed from 3.10 m/s for the worst-case fixed margin to 3.40 m/s and task success from 78.4 to 90.8 percent, while lowering the 95th-percentile margin from 42.0 to 38.2 cm. When the interface model becomes stale after a payload change, the margin inflates toward the worst-case value without manual intervention.","publication_date":"2026-09-28","registered_at":"2026-10-01T13:00:00Z","updated_at":"2026-10-03T16:57:00Z","status":"published","html_path":null,"pdf_path":"articles/2609.00009/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. Science Robotics, 2019. arXiv:1901.08652.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. Science Robotics, 2020. arXiv:2010.11251.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. Science Robotics, 2022. arXiv:2201.08117.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. Robotics: Science and Systems, 2021. arXiv:2107.04034.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. Conference on Robot Learning, 2021. arXiv:2109.11978.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"G. B. Margolis, G. Yang, K. Paigwar, T. Chen, and P. Agrawal. Rapid Locomotion via Reinforcement Learning. Robotics: Science and Systems, 2022. arXiv:2205.02824.","identifier":"arXiv:2205.02824","url":"https://arxiv.org/abs/2205.02824"},{"text":"G. B. Margolis and P. Agrawal. Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior. Conference on Robot Learning, 2022. arXiv:2212.03238.","identifier":"arXiv:2212.03238","url":"https://arxiv.org/abs/2212.03238"},{"text":"I. M. A. Nahrendra, B. Yu, and H. Myung. DreamWaQ: Learning Robust Quadrupedal Locomotion with Implicit Terrain Imagination via Deep Reinforcement Learning. IEEE International Conference on Robotics and Automation, 2023. arXiv:2301.10602.","identifier":"arXiv:2301.10602","url":"https://arxiv.org/abs/2301.10602"},{"text":"D. Hoeller, L. Wellhausen, F. Farshidian, and M. Hutter. Learning a State Representation and Navigation in Cluttered and Dynamic Environments. IEEE Robotics and Automation Letters, 2021. arXiv:2103.04351.","identifier":"arXiv:2103.04351","url":"https://arxiv.org/abs/2103.04351"},{"text":"L. Wellhausen and M. Hutter. ArtPlanner: Robust Legged Robot Navigation in the Field. Field Robotics, 2023. arXiv:2303.01420.","identifier":"arXiv:2303.01420","url":"https://arxiv.org/abs/2303.01420"},{"text":"M. Gaertner, M. Bjelonic, F. Farshidian, and M. Hutter. Collision-Free MPC for Legged Robots in Static and Dynamic Scenes. IEEE International Conference on Robotics and Automation, 2021. arXiv:2103.13987.","identifier":"arXiv:2103.13987","url":"https://arxiv.org/abs/2103.13987"},{"text":"J. Hu, S. Yuan, G. C. R. Bethala, A. Tzes, and Y. Fang. Learning Adaptive Safety Margins for Visual Navigation. arXiv:2607.18200, 2026.","identifier":"arXiv:2607.18200","url":"https://arxiv.org/abs/2607.18200"},{"text":"T. Zang, H. Huang, S. Cheng, and W. Zhang. GAUGE: Planner-Conditioned Active Calibration of Opaque Quadruped Velocity Interfaces. arXiv:2607.02841, 2026.","identifier":"arXiv:2607.02841","url":"https://arxiv.org/abs/2607.02841"},{"text":"T. Zang, S. Wang, H. Huang, and S. Cheng. FutureRay: Control-Aligned Future Range for Agile Quadruped Navigation. arXiv:2608.01197, 2026.","identifier":"arXiv:2608.01197","url":"https://arxiv.org/abs/2608.01197"},{"text":"T. Miki, L. Wellhausen, R. Grandia, F. Jenelten, T. Homberger, and M. Hutter. Elevation Mapping for Locomotion and Navigation using GPU. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2022. arXiv:2204.12876.","identifier":"arXiv:2204.12876","url":"https://arxiv.org/abs/2204.12876"},{"text":"W. Xu, Y. Cai, D. He, J. Lin, and F. Zhang. FAST-LIO2: Fast Direct LiDAR-Inertial Odometry. IEEE Transactions on Robotics, 2022. arXiv:2107.06829.","identifier":"arXiv:2107.06829","url":"https://arxiv.org/abs/2107.06829"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2022. arXiv:2209.12827.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"F. Yang, C. Wang, C. Cadena, and M. Hutter. iPlanner: Imperative Path Planning. Robotics: Science and Systems, 2023. arXiv:2302.11434.","identifier":"arXiv:2302.11434","url":"https://arxiv.org/abs/2302.11434"},{"text":"P. Roth, J. Nubert, F. Yang, M. Mittal, and M. Hutter. ViPlanner: Visual Semantic Imperative Learning for Local Navigation. IEEE International Conference on Robotics and Automation, 2024. arXiv:2310.00982.","identifier":"arXiv:2310.00982","url":"https://arxiv.org/abs/2310.00982"},{"text":"Z. Fu, A. Kumar, A. Agarwal, H. Qi, J. Malik, and D. Pathak. Coupling Vision and Proprioception for Navigation of Legged Robots. IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022. arXiv:2112.02094.","identifier":"arXiv:2112.02094","url":"https://arxiv.org/abs/2112.02094"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. Robotics: Science and Systems, 2023. arXiv:2305.08510.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"T. Miki, J. Lee, L. Wellhausen, and M. Hutter. Learning to Walk in Confined Spaces Using 3D Representation. IEEE International Conference on Robotics and Automation, 2024. arXiv:2403.00187.","identifier":"arXiv:2403.00187","url":"https://arxiv.org/abs/2403.00187"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. Science Robotics, 2024. arXiv:2306.14874.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. Conference on Robot Learning, 2023. arXiv:2309.05665.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. IEEE International Conference on Robotics and Automation, 2024. arXiv:2309.14341.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"S. L. Herbert, M. Chen, S. Han, S. Bansal, J. F. Fisac, and C. J. Tomlin. FaSTrack: A Modular Framework for Fast and Guaranteed Safe Motion Planning. IEEE Conference on Decision and Control, 2017. arXiv:1703.07373.","identifier":"arXiv:1703.07373","url":"https://arxiv.org/abs/1703.07373"},{"text":"S. Bansal, M. Chen, S. Herbert, and C. J. Tomlin. Hamilton-Jacobi Reachability: A Brief Overview and Recent Advances. IEEE Conference on Decision and Control, 2017. arXiv:1709.07523.","identifier":"arXiv:1709.07523","url":"https://arxiv.org/abs/1709.07523"},{"text":"A. Majumdar and R. Tedrake. Funnel Libraries for Real-Time Robust Feedback Motion Planning. International Journal of Robotics Research, 2017. arXiv:1601.04037.","identifier":"arXiv:1601.04037","url":"https://arxiv.org/abs/1601.04037"},{"text":"A. D. Ames, X. Xu, J. W. Grizzle, and P. Tabuada. Control Barrier Function Based Quadratic Programs for Safety Critical Systems. IEEE Transactions on Automatic Control, 2017. arXiv:1609.06408.","identifier":"arXiv:1609.06408","url":"https://arxiv.org/abs/1609.06408"},{"text":"K. P. Wabersich and M. N. Zeilinger. A Predictive Safety Filter for Learning-Based Control of Constrained Nonlinear Dynamical Systems. Automatica, 2021. arXiv:1812.05506.","identifier":"arXiv:1812.05506","url":"https://arxiv.org/abs/1812.05506"},{"text":"J. F. Fisac, A. K. Akametalu, M. N. Zeilinger, S. Kaynama, J. Gillula, and C. J. Tomlin. A General Safety Framework for Learning-Based Control in Uncertain Robotic Systems. IEEE Transactions on Automatic Control, 2019. arXiv:1705.01292.","identifier":"arXiv:1705.01292","url":"https://arxiv.org/abs/1705.01292"},{"text":"T. He, C. Zhang, W. Xiao, G. He, C. Liu, and G. Shi. Agile But Safe: Learning Collision-Free High-Speed Legged Locomotion. Robotics: Science and Systems, 2024. arXiv:2401.17583.","identifier":"arXiv:2401.17583","url":"https://arxiv.org/abs/2401.17583"},{"text":"G. Kahn, A. Villaflor, V. Pong, P. Abbeel, and S. Levine. Uncertainty-Aware Reinforcement Learning for Collision Avoidance. arXiv:1702.01182, 2017.","identifier":"arXiv:1702.01182","url":"https://arxiv.org/abs/1702.01182"},{"text":"F. Berkenkamp, M. Turchetta, A. P. Schoellig, and A. Krause. Safe Model-Based Reinforcement Learning with Stability Guarantees. Advances in Neural Information Processing Systems, 2017. arXiv:1705.08551.","identifier":"arXiv:1705.08551","url":"https://arxiv.org/abs/1705.08551"},{"text":"G. Williams, A. Aldrich, and E. A. Theodorou. Model Predictive Path Integral Control using Covariance Variable Importance Sampling. arXiv:1509.01149, 2015.","identifier":"arXiv:1509.01149","url":"https://arxiv.org/abs/1509.01149"},{"text":"A. N. Angelopoulos and S. Bates. A Gentle Introduction to Conformal Prediction and Distribution-Free Uncertainty Quantification. arXiv:2107.07511, 2021.","identifier":"arXiv:2107.07511","url":"https://arxiv.org/abs/2107.07511"},{"text":"L. Lindemann, M. Cleaveland, G. Shim, and G. J. Pappas. Safe Planning in Dynamic Environments Using Conformal Prediction. IEEE Robotics and Automation Letters, 2023. arXiv:2210.10254.","identifier":"arXiv:2210.10254","url":"https://arxiv.org/abs/2210.10254"},{"text":"A. Dixit, L. Lindemann, S. X. Wei, M. Cleaveland, G. J. Pappas, and J. W. Burdick. Adaptive Conformal Prediction for Motion Planning among Dynamic Agents. Learning for Dynamics and Control Conference, 2023. arXiv:2212.00278.","identifier":"arXiv:2212.00278","url":"https://arxiv.org/abs/2212.00278"},{"text":"A. Loquercio, E. Kaufmann, R. Ranftl, M. Müller, V. Koltun, and D. Scaramuzza. Learning High-Speed Flight in the Wild. Science Robotics, 2021. arXiv:2110.05113.","identifier":"arXiv:2110.05113","url":"https://arxiv.org/abs/2110.05113"},{"text":"V. Makoviychuk et al. Isaac Gym: High Performance GPU-Based Physics Simulation for Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal et al. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. IEEE Robotics and Automation Letters, 2023. arXiv:2301.04195.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2017. arXiv:1703.06907.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. IEEE International Conference on Robotics and Automation, 2018. arXiv:1710.06537.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2609.00009","metadata":{"title":"Adaptive Safety Margins for Quadruped Navigation with Calibrated Velocity Interfaces","authors":[{"name":"Junho Kang","affiliation":"Legged Locomotion and Control Lab, KAIST, Daejeon, South Korea"},{"name":"Hyesun Bae","affiliation":"Legged Locomotion and Control Lab, KAIST, Daejeon, South Korea"},{"name":"Daewon Ryu","affiliation":"Legged Locomotion and Control Lab, KAIST, Daejeon, South Korea"},{"name":"Minju Seo","affiliation":"Legged Locomotion and Control Lab, KAIST, Daejeon, South Korea"}],"publication_date":"2026-09-28","abstract":"Fixed safety margins for legged navigation are tuned once, so they ignore both how well the robot's velocity interface is currently calibrated and how much free space the robot is about to enter. This paper presents an adaptive safety margin for quadruped navigation that combines the two signals. A calibrated interface model obtained with GAUGE supplies a high-confidence upper bound on the lateral deviation caused by command lag, and the control-aligned future-range predictor FutureRay supplies a lower confidence bound on the free range along the directions the controller can actually execute. The margin is the sum of a kinematic term and a lag-error term, each blended between its model-based value and its worst-case value by an explicit confidence weight. We prove that the margin is monotone nondecreasing in both uncertainty sources, that it never falls below the model-based bounds, and that it reduces exactly to the worst-case fixed margin when both confidences vanish. On a Unitree Go2 evaluated on 12 obstacle-adjacent segments at speeds from 1.0 to 4.0 m/s, with matching Isaac Lab simulation, the adaptive margin raises the collision-free speed from 3.10 m/s for the worst-case fixed margin to 3.40 m/s and task success from 78.4 to 90.8 percent, while lowering the 95th-percentile margin from 42.0 to 38.2 cm. When the interface model becomes stale after a payload change, the margin inflates toward the worst-case value without manual intervention.","references":[{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. Science Robotics, 2019. arXiv:1901.08652.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. Science Robotics, 2020. arXiv:2010.11251.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. Science Robotics, 2022. arXiv:2201.08117.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. Robotics: Science and Systems, 2021. arXiv:2107.04034.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. Conference on Robot Learning, 2021. arXiv:2109.11978.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"G. B. Margolis, G. Yang, K. Paigwar, T. Chen, and P. Agrawal. Rapid Locomotion via Reinforcement Learning. Robotics: Science and Systems, 2022. arXiv:2205.02824.","identifier":"arXiv:2205.02824","url":"https://arxiv.org/abs/2205.02824"},{"text":"G. B. Margolis and P. Agrawal. Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior. Conference on Robot Learning, 2022. arXiv:2212.03238.","identifier":"arXiv:2212.03238","url":"https://arxiv.org/abs/2212.03238"},{"text":"I. M. A. Nahrendra, B. Yu, and H. Myung. DreamWaQ: Learning Robust Quadrupedal Locomotion with Implicit Terrain Imagination via Deep Reinforcement Learning. IEEE International Conference on Robotics and Automation, 2023. arXiv:2301.10602.","identifier":"arXiv:2301.10602","url":"https://arxiv.org/abs/2301.10602"},{"text":"D. Hoeller, L. Wellhausen, F. Farshidian, and M. Hutter. Learning a State Representation and Navigation in Cluttered and Dynamic Environments. IEEE Robotics and Automation Letters, 2021. arXiv:2103.04351.","identifier":"arXiv:2103.04351","url":"https://arxiv.org/abs/2103.04351"},{"text":"L. Wellhausen and M. Hutter. ArtPlanner: Robust Legged Robot Navigation in the Field. Field Robotics, 2023. arXiv:2303.01420.","identifier":"arXiv:2303.01420","url":"https://arxiv.org/abs/2303.01420"},{"text":"M. Gaertner, M. Bjelonic, F. Farshidian, and M. Hutter. Collision-Free MPC for Legged Robots in Static and Dynamic Scenes. IEEE International Conference on Robotics and Automation, 2021. arXiv:2103.13987.","identifier":"arXiv:2103.13987","url":"https://arxiv.org/abs/2103.13987"},{"text":"J. Hu, S. Yuan, G. C. R. Bethala, A. Tzes, and Y. Fang. Learning Adaptive Safety Margins for Visual Navigation. arXiv:2607.18200, 2026.","identifier":"arXiv:2607.18200","url":"https://arxiv.org/abs/2607.18200"},{"text":"T. Zang, H. Huang, S. Cheng, and W. Zhang. GAUGE: Planner-Conditioned Active Calibration of Opaque Quadruped Velocity Interfaces. arXiv:2607.02841, 2026.","identifier":"arXiv:2607.02841","url":"https://arxiv.org/abs/2607.02841"},{"text":"T. Zang, S. Wang, H. Huang, and S. Cheng. FutureRay: Control-Aligned Future Range for Agile Quadruped Navigation. arXiv:2608.01197, 2026.","identifier":"arXiv:2608.01197","url":"https://arxiv.org/abs/2608.01197"},{"text":"T. Miki, L. Wellhausen, R. Grandia, F. Jenelten, T. Homberger, and M. Hutter. Elevation Mapping for Locomotion and Navigation using GPU. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2022. arXiv:2204.12876.","identifier":"arXiv:2204.12876","url":"https://arxiv.org/abs/2204.12876"},{"text":"W. Xu, Y. Cai, D. He, J. Lin, and F. Zhang. FAST-LIO2: Fast Direct LiDAR-Inertial Odometry. IEEE Transactions on Robotics, 2022. arXiv:2107.06829.","identifier":"arXiv:2107.06829","url":"https://arxiv.org/abs/2107.06829"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2022. arXiv:2209.12827.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"F. Yang, C. Wang, C. Cadena, and M. Hutter. iPlanner: Imperative Path Planning. Robotics: Science and Systems, 2023. arXiv:2302.11434.","identifier":"arXiv:2302.11434","url":"https://arxiv.org/abs/2302.11434"},{"text":"P. Roth, J. Nubert, F. Yang, M. Mittal, and M. Hutter. ViPlanner: Visual Semantic Imperative Learning for Local Navigation. IEEE International Conference on Robotics and Automation, 2024. arXiv:2310.00982.","identifier":"arXiv:2310.00982","url":"https://arxiv.org/abs/2310.00982"},{"text":"Z. Fu, A. Kumar, A. Agarwal, H. Qi, J. Malik, and D. Pathak. Coupling Vision and Proprioception for Navigation of Legged Robots. IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022. arXiv:2112.02094.","identifier":"arXiv:2112.02094","url":"https://arxiv.org/abs/2112.02094"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. Robotics: Science and Systems, 2023. arXiv:2305.08510.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"T. Miki, J. Lee, L. Wellhausen, and M. Hutter. Learning to Walk in Confined Spaces Using 3D Representation. IEEE International Conference on Robotics and Automation, 2024. arXiv:2403.00187.","identifier":"arXiv:2403.00187","url":"https://arxiv.org/abs/2403.00187"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. Science Robotics, 2024. arXiv:2306.14874.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. Conference on Robot Learning, 2023. arXiv:2309.05665.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. IEEE International Conference on Robotics and Automation, 2024. arXiv:2309.14341.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"S. L. Herbert, M. Chen, S. Han, S. Bansal, J. F. Fisac, and C. J. Tomlin. FaSTrack: A Modular Framework for Fast and Guaranteed Safe Motion Planning. IEEE Conference on Decision and Control, 2017. arXiv:1703.07373.","identifier":"arXiv:1703.07373","url":"https://arxiv.org/abs/1703.07373"},{"text":"S. Bansal, M. Chen, S. Herbert, and C. J. Tomlin. Hamilton-Jacobi Reachability: A Brief Overview and Recent Advances. IEEE Conference on Decision and Control, 2017. arXiv:1709.07523.","identifier":"arXiv:1709.07523","url":"https://arxiv.org/abs/1709.07523"},{"text":"A. Majumdar and R. Tedrake. Funnel Libraries for Real-Time Robust Feedback Motion Planning. International Journal of Robotics Research, 2017. arXiv:1601.04037.","identifier":"arXiv:1601.04037","url":"https://arxiv.org/abs/1601.04037"},{"text":"A. D. Ames, X. Xu, J. W. Grizzle, and P. Tabuada. Control Barrier Function Based Quadratic Programs for Safety Critical Systems. IEEE Transactions on Automatic Control, 2017. arXiv:1609.06408.","identifier":"arXiv:1609.06408","url":"https://arxiv.org/abs/1609.06408"},{"text":"K. P. Wabersich and M. N. Zeilinger. A Predictive Safety Filter for Learning-Based Control of Constrained Nonlinear Dynamical Systems. Automatica, 2021. arXiv:1812.05506.","identifier":"arXiv:1812.05506","url":"https://arxiv.org/abs/1812.05506"},{"text":"J. F. Fisac, A. K. Akametalu, M. N. Zeilinger, S. Kaynama, J. Gillula, and C. J. Tomlin. A General Safety Framework for Learning-Based Control in Uncertain Robotic Systems. IEEE Transactions on Automatic Control, 2019. arXiv:1705.01292.","identifier":"arXiv:1705.01292","url":"https://arxiv.org/abs/1705.01292"},{"text":"T. He, C. Zhang, W. Xiao, G. He, C. Liu, and G. Shi. Agile But Safe: Learning Collision-Free High-Speed Legged Locomotion. Robotics: Science and Systems, 2024. arXiv:2401.17583.","identifier":"arXiv:2401.17583","url":"https://arxiv.org/abs/2401.17583"},{"text":"G. Kahn, A. Villaflor, V. Pong, P. Abbeel, and S. Levine. Uncertainty-Aware Reinforcement Learning for Collision Avoidance. arXiv:1702.01182, 2017.","identifier":"arXiv:1702.01182","url":"https://arxiv.org/abs/1702.01182"},{"text":"F. Berkenkamp, M. Turchetta, A. P. Schoellig, and A. Krause. Safe Model-Based Reinforcement Learning with Stability Guarantees. Advances in Neural Information Processing Systems, 2017. arXiv:1705.08551.","identifier":"arXiv:1705.08551","url":"https://arxiv.org/abs/1705.08551"},{"text":"G. Williams, A. Aldrich, and E. A. Theodorou. Model Predictive Path Integral Control using Covariance Variable Importance Sampling. arXiv:1509.01149, 2015.","identifier":"arXiv:1509.01149","url":"https://arxiv.org/abs/1509.01149"},{"text":"A. N. Angelopoulos and S. Bates. A Gentle Introduction to Conformal Prediction and Distribution-Free Uncertainty Quantification. arXiv:2107.07511, 2021.","identifier":"arXiv:2107.07511","url":"https://arxiv.org/abs/2107.07511"},{"text":"L. Lindemann, M. Cleaveland, G. Shim, and G. J. Pappas. Safe Planning in Dynamic Environments Using Conformal Prediction. IEEE Robotics and Automation Letters, 2023. arXiv:2210.10254.","identifier":"arXiv:2210.10254","url":"https://arxiv.org/abs/2210.10254"},{"text":"A. Dixit, L. Lindemann, S. X. Wei, M. Cleaveland, G. J. Pappas, and J. W. Burdick. Adaptive Conformal Prediction for Motion Planning among Dynamic Agents. Learning for Dynamics and Control Conference, 2023. arXiv:2212.00278.","identifier":"arXiv:2212.00278","url":"https://arxiv.org/abs/2212.00278"},{"text":"A. Loquercio, E. Kaufmann, R. Ranftl, M. Müller, V. Koltun, and D. Scaramuzza. Learning High-Speed Flight in the Wild. Science Robotics, 2021. arXiv:2110.05113.","identifier":"arXiv:2110.05113","url":"https://arxiv.org/abs/2110.05113"},{"text":"V. Makoviychuk et al. Isaac Gym: High Performance GPU-Based Physics Simulation for Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal et al. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. IEEE Robotics and Automation Letters, 2023. arXiv:2301.04195.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2017. arXiv:1703.06907.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. IEEE International Conference on Robotics and Automation, 2018. arXiv:1710.06537.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2609.00009","slug":"2609.00009","registered_at":"2026-10-01T13:00:00Z","updated_at":"2026-10-03T16:57:00Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2609.00009","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2609.00009","tex_path":"articles/2609.00009/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-10-01T15:17:00Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2609.00009"},{"kagg_id":"KAGG:2609.00008","slug":"2609.00008","title":"Local Planning under Asymmetric Collision Geometries Across Embodiments","authors":[{"name":"Anika Roth","affiliation":"Robot Learning Lab, University of Zurich, Zurich, Switzerland"},{"name":"Selin Aydin","affiliation":"Robot Learning Lab, University of Zurich, Zurich, Switzerland"},{"name":"Lukas Brandt","affiliation":"Robot Learning Lab, University of Zurich, Zurich, Switzerland"}],"affiliation":null,"abstract":"Local planners that run across quadrupeds, wheeled robots, and humanoids commonly represent every body by a symmetrically inflated disk or a static footprint, discarding the swing envelope of the legs, humanoid headroom, and the directional structure of clearance. This paper builds an embodiment-accurate collision layer on the AgniNav configuration-driven backbone. The layer replaces the static envelope with a configuration-dependent collision set, formed by the footprint and a phase-indexed leg-swing volume for quadrupeds, a body envelope for wheeled rovers, and a full-body capsule with arm reach and a headroom term for humanoids, and it scores candidate trajectories with a clearance cost that is anisotropic in each body frame. The planner is evaluated in 15 Isaac Lab mixed-embodiment scenes with diagonal gaps of 45 to 75 degrees, cluttered corridors, and multi-robot proximity, across four embodiments, five seeds, and 30 episodes each. Relative to symmetric inflation, the method raises the feasible passage width gain from 14.2 to 14.8 percent, lowers the collision rate from 0.19 to 0.18 per km, and lowers the 95th-percentile planner runtime from 6.9 to 6.4 ms, while the aggregate spurious-stoppage rate is unchanged at 0.3 per episode. The aggregate gains are modest and concentrate in specific situations: in a 60 degree, 71 cm diagonal gap where the symmetric-inflation quadruped fails 12 of 30 attempts, the swing-aware quadruped passes at 1.2 m/s with 8 cm clearance, and in one corridor the full-body humanoid model reduces spurious stoppages near a wheeled robot from 3.1 to 0.4 per episode.","publication_date":"2026-09-22","registered_at":"2026-09-25T10:45:00Z","updated_at":"2026-09-27T16:26:00Z","status":"published","html_path":null,"pdf_path":"articles/2609.00008/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"J. Zhang, Y. Du, X. Guo, S. Sun, X. Liu, Y. Sun, G. Lu, W. Sui, and J. Li. EA-Nav: Learning Safe Visual Navigation Policies with Embodiment Awareness. arXiv:2607.19880, 2026.","identifier":"arXiv:2607.19880","url":"https://arxiv.org/abs/2607.19880"},{"text":"J. Hu, S. Yuan, G. C. R. Bethala, A. Tzes, and Y. Fang. Learning Adaptive Safety Margins for Visual Navigation. arXiv:2607.18200, 2026.","identifier":"arXiv:2607.18200","url":"https://arxiv.org/abs/2607.18200"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Sridhar, D. Shah, C. Glossop, and S. Levine. NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration. arXiv:2310.07896, 2023.","identifier":"arXiv:2310.07896","url":"https://arxiv.org/abs/2310.07896"},{"text":"A. Eftekhar et al.. The One RING: a Robotic Indoor Navigation Generalist. arXiv:2412.14401, 2024.","identifier":"arXiv:2412.14401","url":"https://arxiv.org/abs/2412.14401"},{"text":"J. Zhang et al.. Embodied Navigation Foundation Model. arXiv:2509.12129, 2025.","identifier":"arXiv:2509.12129","url":"https://arxiv.org/abs/2509.12129"},{"text":"D. Fox, W. Burgard, and S. Thrun. The Dynamic Window Approach to Collision Avoidance. IEEE Robotics and Automation Magazine, 4(1):23-33, doi:10.1109/100.580977, 1997.","identifier":"doi:10.1109/100.580977","url":"https://doi.org/10.1109/100.580977"},{"text":"X. Zhou, Z. Wang, H. Ye, C. Xu, and F. Gao. EGO-Planner: An ESDF-free Gradient-based Local Planner for Quadrotors. arXiv:2008.08835, 2020.","identifier":"arXiv:2008.08835","url":"https://arxiv.org/abs/2008.08835"},{"text":"X. Zhang, A. Liniger, and F. Borrelli. Optimization-Based Collision Avoidance. arXiv:1711.03449, 2017.","identifier":"arXiv:1711.03449","url":"https://arxiv.org/abs/1711.03449"},{"text":"H. Oleynikova, Z. Taylor, M. Fehr, R. Siegwart, and J. Nieto. Voxblox: Incremental 3D Euclidean Signed Distance Fields for On-Board MAV Planning. arXiv:1611.03631, 2016.","identifier":"arXiv:1611.03631","url":"https://arxiv.org/abs/1611.03631"},{"text":"M. Gaertner, M. Bjelonic, F. Farshidian, and M. Hutter. Collision-Free MPC for Legged Robots in Static and Dynamic Scenes. arXiv:2103.13987, 2021.","identifier":"arXiv:2103.13987","url":"https://arxiv.org/abs/2103.13987"},{"text":"R. Grandia, F. Jenelten, S. Yang, F. Farshidian, and M. Hutter. Perceptive Locomotion Through Nonlinear Model-Predictive Control. arXiv:2208.08373, 2022.","identifier":"arXiv:2208.08373","url":"https://arxiv.org/abs/2208.08373"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. arXiv:2211.07638, 2022.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. arXiv:2309.14341, 2023.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"I. Radosavovic, T. Xiao, B. Zhang, T. Darrell, J. Malik, and K. Sreenath. Real-World Humanoid Locomotion with Reinforcement Learning. arXiv:2303.03381, 2023.","identifier":"arXiv:2303.03381","url":"https://arxiv.org/abs/2303.03381"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, L. Wellhausen, F. Farshidian, and M. Hutter. Learning a State Representation and Navigation in Cluttered and Dynamic Environments. arXiv:2103.04351, 2021.","identifier":"arXiv:2103.04351","url":"https://arxiv.org/abs/2103.04351"},{"text":"T. He, C. Zhang, W. Xiao, G. He, C. Liu, and G. Shi. Agile But Safe: Learning Collision-Free High-Speed Legged Locomotion. arXiv:2401.17583, 2024.","identifier":"arXiv:2401.17583","url":"https://arxiv.org/abs/2401.17583"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. arXiv:2305.08510, 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"H.-T. L. Chiang, A. Faust, M. Fiser, and A. Francis. Learning Navigation Behaviors End-to-End with AutoRL. arXiv:1809.10124, 2018.","identifier":"arXiv:1809.10124","url":"https://arxiv.org/abs/1809.10124"},{"text":"Y. F. Chen, M. Liu, M. Everett, and J. P. How. Decentralized Non-communicating Multiagent Collision Avoidance with Deep Reinforcement Learning. arXiv:1609.07845, 2016.","identifier":"arXiv:1609.07845","url":"https://arxiv.org/abs/1609.07845"},{"text":"M. Everett, Y. F. Chen, and J. P. How. Motion Planning Among Dynamic, Decision-Making Agents with Deep Reinforcement Learning. arXiv:1805.01956, 2018.","identifier":"arXiv:1805.01956","url":"https://arxiv.org/abs/1805.01956"},{"text":"P. Long, T. Fan, X. Liao, W. Liu, H. Zhang, and J. Pan. Towards Optimally Decentralized Multi-Robot Collision Avoidance via Deep Reinforcement Learning. arXiv:1709.10082, 2017.","identifier":"arXiv:1709.10082","url":"https://arxiv.org/abs/1709.10082"},{"text":"C. Chen, Y. Liu, S. Kreiss, and A. Alahi. Crowd-Robot Interaction: Crowd-aware Robot Navigation with Attention-based Deep Reinforcement Learning. arXiv:1809.08835, 2018.","identifier":"arXiv:1809.08835","url":"https://arxiv.org/abs/1809.08835"},{"text":"C. Mavrogiannis et al.. Core Challenges of Social Robot Navigation: A Survey. arXiv:2103.05668, 2021.","identifier":"arXiv:2103.05668","url":"https://arxiv.org/abs/2103.05668"},{"text":"A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada. Control Barrier Functions: Theory and Applications. arXiv:1903.11199, 2019.","identifier":"arXiv:1903.11199","url":"https://arxiv.org/abs/1903.11199"},{"text":"S. Bansal, M. Chen, S. Herbert, and C. J. Tomlin. Hamilton-Jacobi Reachability: A Brief Overview and Recent Advances. arXiv:1709.07523, 2017.","identifier":"arXiv:1709.07523","url":"https://arxiv.org/abs/1709.07523"},{"text":"J. F. Fisac, A. K. Akametalu, M. N. Zeilinger, S. Kaynama, J. Gillula, and C. J. Tomlin. A General Safety Framework for Learning-Based Control in Uncertain Robotic Systems. arXiv:1705.01292, 2017.","identifier":"arXiv:1705.01292","url":"https://arxiv.org/abs/1705.01292"},{"text":"M. Mittal et al.. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. arXiv:2301.04195, 2023.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"V. Makoviychuk et al.. Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. arXiv:1703.06907, 2017.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"},{"text":"A. Kadian et al.. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance? arXiv:1912.06321, 2019.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"},{"text":"P. Anderson et al.. On Evaluation of Embodied Navigation Agents. arXiv:1807.06757, 2018.","identifier":"arXiv:1807.06757","url":"https://arxiv.org/abs/1807.06757"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2609.00008","metadata":{"title":"Local Planning under Asymmetric Collision Geometries Across Embodiments","authors":[{"name":"Anika Roth","affiliation":"Robot Learning Lab, University of Zurich, Zurich, Switzerland"},{"name":"Selin Aydin","affiliation":"Robot Learning Lab, University of Zurich, Zurich, Switzerland"},{"name":"Lukas Brandt","affiliation":"Robot Learning Lab, University of Zurich, Zurich, Switzerland"}],"publication_date":"2026-09-22","abstract":"Local planners that run across quadrupeds, wheeled robots, and humanoids commonly represent every body by a symmetrically inflated disk or a static footprint, discarding the swing envelope of the legs, humanoid headroom, and the directional structure of clearance. This paper builds an embodiment-accurate collision layer on the AgniNav configuration-driven backbone. The layer replaces the static envelope with a configuration-dependent collision set, formed by the footprint and a phase-indexed leg-swing volume for quadrupeds, a body envelope for wheeled rovers, and a full-body capsule with arm reach and a headroom term for humanoids, and it scores candidate trajectories with a clearance cost that is anisotropic in each body frame. The planner is evaluated in 15 Isaac Lab mixed-embodiment scenes with diagonal gaps of 45 to 75 degrees, cluttered corridors, and multi-robot proximity, across four embodiments, five seeds, and 30 episodes each. Relative to symmetric inflation, the method raises the feasible passage width gain from 14.2 to 14.8 percent, lowers the collision rate from 0.19 to 0.18 per km, and lowers the 95th-percentile planner runtime from 6.9 to 6.4 ms, while the aggregate spurious-stoppage rate is unchanged at 0.3 per episode. The aggregate gains are modest and concentrate in specific situations: in a 60 degree, 71 cm diagonal gap where the symmetric-inflation quadruped fails 12 of 30 attempts, the swing-aware quadruped passes at 1.2 m/s with 8 cm clearance, and in one corridor the full-body humanoid model reduces spurious stoppages near a wheeled robot from 3.1 to 0.4 per episode.","references":[{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"J. Zhang, Y. Du, X. Guo, S. Sun, X. Liu, Y. Sun, G. Lu, W. Sui, and J. Li. EA-Nav: Learning Safe Visual Navigation Policies with Embodiment Awareness. arXiv:2607.19880, 2026.","identifier":"arXiv:2607.19880","url":"https://arxiv.org/abs/2607.19880"},{"text":"J. Hu, S. Yuan, G. C. R. Bethala, A. Tzes, and Y. Fang. Learning Adaptive Safety Margins for Visual Navigation. arXiv:2607.18200, 2026.","identifier":"arXiv:2607.18200","url":"https://arxiv.org/abs/2607.18200"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Sridhar, D. Shah, C. Glossop, and S. Levine. NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration. arXiv:2310.07896, 2023.","identifier":"arXiv:2310.07896","url":"https://arxiv.org/abs/2310.07896"},{"text":"A. Eftekhar et al.. The One RING: a Robotic Indoor Navigation Generalist. arXiv:2412.14401, 2024.","identifier":"arXiv:2412.14401","url":"https://arxiv.org/abs/2412.14401"},{"text":"J. Zhang et al.. Embodied Navigation Foundation Model. arXiv:2509.12129, 2025.","identifier":"arXiv:2509.12129","url":"https://arxiv.org/abs/2509.12129"},{"text":"D. Fox, W. Burgard, and S. Thrun. The Dynamic Window Approach to Collision Avoidance. IEEE Robotics and Automation Magazine, 4(1):23-33, doi:10.1109/100.580977, 1997.","identifier":"doi:10.1109/100.580977","url":"https://doi.org/10.1109/100.580977"},{"text":"X. Zhou, Z. Wang, H. Ye, C. Xu, and F. Gao. EGO-Planner: An ESDF-free Gradient-based Local Planner for Quadrotors. arXiv:2008.08835, 2020.","identifier":"arXiv:2008.08835","url":"https://arxiv.org/abs/2008.08835"},{"text":"X. Zhang, A. Liniger, and F. Borrelli. Optimization-Based Collision Avoidance. arXiv:1711.03449, 2017.","identifier":"arXiv:1711.03449","url":"https://arxiv.org/abs/1711.03449"},{"text":"H. Oleynikova, Z. Taylor, M. Fehr, R. Siegwart, and J. Nieto. Voxblox: Incremental 3D Euclidean Signed Distance Fields for On-Board MAV Planning. arXiv:1611.03631, 2016.","identifier":"arXiv:1611.03631","url":"https://arxiv.org/abs/1611.03631"},{"text":"M. Gaertner, M. Bjelonic, F. Farshidian, and M. Hutter. Collision-Free MPC for Legged Robots in Static and Dynamic Scenes. arXiv:2103.13987, 2021.","identifier":"arXiv:2103.13987","url":"https://arxiv.org/abs/2103.13987"},{"text":"R. Grandia, F. Jenelten, S. Yang, F. Farshidian, and M. Hutter. Perceptive Locomotion Through Nonlinear Model-Predictive Control. arXiv:2208.08373, 2022.","identifier":"arXiv:2208.08373","url":"https://arxiv.org/abs/2208.08373"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. arXiv:2211.07638, 2022.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. arXiv:2309.14341, 2023.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"I. Radosavovic, T. Xiao, B. Zhang, T. Darrell, J. Malik, and K. Sreenath. Real-World Humanoid Locomotion with Reinforcement Learning. arXiv:2303.03381, 2023.","identifier":"arXiv:2303.03381","url":"https://arxiv.org/abs/2303.03381"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, L. Wellhausen, F. Farshidian, and M. Hutter. Learning a State Representation and Navigation in Cluttered and Dynamic Environments. arXiv:2103.04351, 2021.","identifier":"arXiv:2103.04351","url":"https://arxiv.org/abs/2103.04351"},{"text":"T. He, C. Zhang, W. Xiao, G. He, C. Liu, and G. Shi. Agile But Safe: Learning Collision-Free High-Speed Legged Locomotion. arXiv:2401.17583, 2024.","identifier":"arXiv:2401.17583","url":"https://arxiv.org/abs/2401.17583"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. arXiv:2305.08510, 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"H.-T. L. Chiang, A. Faust, M. Fiser, and A. Francis. Learning Navigation Behaviors End-to-End with AutoRL. arXiv:1809.10124, 2018.","identifier":"arXiv:1809.10124","url":"https://arxiv.org/abs/1809.10124"},{"text":"Y. F. Chen, M. Liu, M. Everett, and J. P. How. Decentralized Non-communicating Multiagent Collision Avoidance with Deep Reinforcement Learning. arXiv:1609.07845, 2016.","identifier":"arXiv:1609.07845","url":"https://arxiv.org/abs/1609.07845"},{"text":"M. Everett, Y. F. Chen, and J. P. How. Motion Planning Among Dynamic, Decision-Making Agents with Deep Reinforcement Learning. arXiv:1805.01956, 2018.","identifier":"arXiv:1805.01956","url":"https://arxiv.org/abs/1805.01956"},{"text":"P. Long, T. Fan, X. Liao, W. Liu, H. Zhang, and J. Pan. Towards Optimally Decentralized Multi-Robot Collision Avoidance via Deep Reinforcement Learning. arXiv:1709.10082, 2017.","identifier":"arXiv:1709.10082","url":"https://arxiv.org/abs/1709.10082"},{"text":"C. Chen, Y. Liu, S. Kreiss, and A. Alahi. Crowd-Robot Interaction: Crowd-aware Robot Navigation with Attention-based Deep Reinforcement Learning. arXiv:1809.08835, 2018.","identifier":"arXiv:1809.08835","url":"https://arxiv.org/abs/1809.08835"},{"text":"C. Mavrogiannis et al.. Core Challenges of Social Robot Navigation: A Survey. arXiv:2103.05668, 2021.","identifier":"arXiv:2103.05668","url":"https://arxiv.org/abs/2103.05668"},{"text":"A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada. Control Barrier Functions: Theory and Applications. arXiv:1903.11199, 2019.","identifier":"arXiv:1903.11199","url":"https://arxiv.org/abs/1903.11199"},{"text":"S. Bansal, M. Chen, S. Herbert, and C. J. Tomlin. Hamilton-Jacobi Reachability: A Brief Overview and Recent Advances. arXiv:1709.07523, 2017.","identifier":"arXiv:1709.07523","url":"https://arxiv.org/abs/1709.07523"},{"text":"J. F. Fisac, A. K. Akametalu, M. N. Zeilinger, S. Kaynama, J. Gillula, and C. J. Tomlin. A General Safety Framework for Learning-Based Control in Uncertain Robotic Systems. arXiv:1705.01292, 2017.","identifier":"arXiv:1705.01292","url":"https://arxiv.org/abs/1705.01292"},{"text":"M. Mittal et al.. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. arXiv:2301.04195, 2023.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"V. Makoviychuk et al.. Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. arXiv:1703.06907, 2017.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"},{"text":"A. Kadian et al.. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance? arXiv:1912.06321, 2019.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"},{"text":"P. Anderson et al.. On Evaluation of Embodied Navigation Agents. arXiv:1807.06757, 2018.","identifier":"arXiv:1807.06757","url":"https://arxiv.org/abs/1807.06757"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2609.00008","slug":"2609.00008","registered_at":"2026-09-25T10:45:00Z","updated_at":"2026-09-27T16:26:00Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2609.00008","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2609.00008","tex_path":"articles/2609.00008/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-09-25T13:02:00Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2609.00008"},{"kagg_id":"KAGG:2609.00007","slug":"2609.00007","title":"Future-Range Conditioning for Imitation-Learned Quadruped Navigation","authors":[{"name":"Rachel Kim","affiliation":"Robotics Institute, Georgia Institute of Technology, Atlanta, GA, USA"},{"name":"Ivan Sokolov","affiliation":"Robotics Institute, Georgia Institute of Technology, Atlanta, GA, USA"},{"name":"Marcus Hill","affiliation":"Robotics Institute, Georgia Institute of Technology, Atlanta, GA, USA"}],"affiliation":null,"abstract":"Imitation-learned navigation policies for legged robots inherit the range representation of the data they are trained on: a policy that observes only the present range scan tends to reproduce how a particular demonstrator reacted to that scan rather than what the robot will actually encounter. This paper studies whether conditioning an imitation policy on a predicted, control-aligned future range field reduces this dependence. A transformer policy consumes the present range scan, a short range history, and the future free range predicted by a FutureRay-style module along the directions the robot is expected to travel under its current command. To keep the policy usable when the predictor is wrong, training applies range-representation dropout that masks or corrupts the future input, so that the policy learns to fall back on present-range behavior. The approach is evaluated with a simulated ANYmal quadruped in Isaac Lab, trained on 12 h of teleoperation from three operators plus 2000 simulator rollouts across 20 cluttered scenes, and tested on 10 unseen cluttered scenes. The conditioned policy reaches a success rate of 88.4 percent, against 76.1 percent for present-range-only imitation and 74.3 percent for a rule-based reactive baseline, raises collision-free distance from 3.8 to 4.2 km per collision, and attains an oracle-parity ratio of 0.81 relative to a policy trained on ground-truth future range. Future conditioning without dropout yields no gain over present-range imitation. Hardware validation remains future work.","publication_date":"2026-09-17","registered_at":"2026-09-20T18:31:00Z","updated_at":"2026-09-20T21:31:00Z","status":"published","html_path":null,"pdf_path":"articles/2609.00007/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"T. Zang, S. Wang, H. Huang, and S. Cheng. FutureRay: Control-Aligned Future Range for Agile Quadruped Navigation. arXiv:2608.01197, 2026.","identifier":"arXiv:2608.01197","url":"https://arxiv.org/abs/2608.01197"},{"text":"M. Hutter, C. Gehring, D. Jud, A. Lauber, C. D. Bellicoso, V. Tsounis, J. Hwangbo, K. Bodie, P. Fankhauser, M. Bloesch, R. Diethelm, S. Bachmann, A. Melzer, and M. Hoepflinger. ANYmal - A Highly Mobile and Dynamic Quadrupedal Robot. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2016.","identifier":"doi:10.1109/IROS.2016.7758092","url":"https://doi.org/10.1109/IROS.2016.7758092"},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. Science Robotics, 4(26), 2019. arXiv:1901.08652.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. Science Robotics, 5(47), 2020. arXiv:2010.11251.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. Science Robotics, 7(62), 2022. arXiv:2201.08117.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. Conference on Robot Learning (CoRL), 2021. arXiv:2109.11978.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. Robotics: Science and Systems (RSS), 2021. arXiv:2107.04034.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. Conference on Robot Learning (CoRL), 2022. arXiv:2211.07638.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. Conference on Robot Learning (CoRL), 2023. arXiv:2309.05665.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. IEEE International Conference on Robotics and Automation (ICRA), 2024. arXiv:2309.14341.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2022. arXiv:2209.12827.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, L. Wellhausen, F. Farshidian, and M. Hutter. Learning a State Representation and Navigation in Cluttered and Dynamic Environments. IEEE Robotics and Automation Letters, 6(3), 2021. arXiv:2103.04351.","identifier":"arXiv:2103.04351","url":"https://arxiv.org/abs/2103.04351"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. Science Robotics, 9(88), 2024. arXiv:2306.14874.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"L. Wellhausen and M. Hutter. ArtPlanner: Robust Legged Robot Navigation in the Field. Field Robotics, 3, 2023. arXiv:2303.01420.","identifier":"arXiv:2303.01420","url":"https://arxiv.org/abs/2303.01420"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. Robotics: Science and Systems (RSS), 2023. arXiv:2305.08510.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"J. Truong, D. Yarats, T. Li, F. Meier, S. Chernova, D. Batra, and A. Rai. Learning Navigation Skills for Legged Robots with Learned Robot Embodiments. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2021. arXiv:2011.06188.","identifier":"arXiv:2011.06188","url":"https://arxiv.org/abs/2011.06188"},{"text":"S. Kareer, N. Yokoyama, D. Batra, S. Ha, and J. Truong. ViNL: Visual Navigation and Locomotion Over Obstacles. IEEE International Conference on Robotics and Automation (ICRA), 2023. arXiv:2210.14791.","identifier":"arXiv:2210.14791","url":"https://arxiv.org/abs/2210.14791"},{"text":"J. Lee, M. Bjelonic, A. Reske, L. Wellhausen, T. Miki, and M. Hutter. Learning Robust Autonomous Navigation and Locomotion for Wheeled-Legged Robots. Science Robotics, 9(89), 2024. arXiv:2405.01792.","identifier":"arXiv:2405.01792","url":"https://arxiv.org/abs/2405.01792"},{"text":"O. Khatib. Real-Time Obstacle Avoidance for Manipulators and Mobile Robots. The International Journal of Robotics Research, 5(1):90-98, 1986.","identifier":"doi:10.1177/027836498600500106","url":"https://doi.org/10.1177/027836498600500106"},{"text":"D. Fox, W. Burgard, and S. Thrun. The Dynamic Window Approach to Collision Avoidance. IEEE Robotics and Automation Magazine, 4(1):23-33, 1997.","identifier":"doi:10.1109/100.580977","url":"https://doi.org/10.1109/100.580977"},{"text":"M. Pfeiffer, M. Schaeuble, J. Nieto, R. Siegwart, and C. Cadena. From Perception to Decision: A Data-Driven Approach to End-to-End Motion Planning for Autonomous Ground Robots. IEEE International Conference on Robotics and Automation (ICRA), 2017. arXiv:1609.07910.","identifier":"arXiv:1609.07910","url":"https://arxiv.org/abs/1609.07910"},{"text":"L. Tai, G. Paolo, and M. Liu. Virtual-to-Real Deep Reinforcement Learning: Continuous Control of Mobile Robots for Mapless Navigation. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2017. arXiv:1703.00420.","identifier":"arXiv:1703.00420","url":"https://arxiv.org/abs/1703.00420"},{"text":"G. Kahn, A. Villaflor, B. Ding, P. Abbeel, and S. Levine. Self-Supervised Deep Reinforcement Learning with Generalized Computation Graphs for Robot Navigation. IEEE International Conference on Robotics and Automation (ICRA), 2018. arXiv:1709.10489.","identifier":"arXiv:1709.10489","url":"https://arxiv.org/abs/1709.10489"},{"text":"G. Kahn, P. Abbeel, and S. Levine. BADGR: An Autonomous Self-Supervised Learning-Based Navigation System. IEEE Robotics and Automation Letters, 6(2), 2021. arXiv:2002.05700.","identifier":"arXiv:2002.05700","url":"https://arxiv.org/abs/2002.05700"},{"text":"C. Finn and S. Levine. Deep Visual Foresight for Planning Robot Motion. IEEE International Conference on Robotics and Automation (ICRA), 2017. arXiv:1610.00696.","identifier":"arXiv:1610.00696","url":"https://arxiv.org/abs/1610.00696"},{"text":"S. Ross, G. J. Gordon, and J. A. Bagnell. A Reduction of Imitation Learning and Structured Prediction to No-Regret Online Learning. International Conference on Artificial Intelligence and Statistics (AISTATS), 2011. arXiv:1011.0686.","identifier":"arXiv:1011.0686","url":"https://arxiv.org/abs/1011.0686"},{"text":"M. Bojarski, D. Del Testa, D. Dworakowski, B. Firner, B. Flepp, P. Goyal, L. D. Jackel, M. Monfort, U. Muller, J. Zhang, X. Zhang, J. Zhao, and K. Zieba. End to End Learning for Self-Driving Cars. arXiv:1604.07316, 2016.","identifier":"arXiv:1604.07316","url":"https://arxiv.org/abs/1604.07316"},{"text":"F. Codevilla, M. Mueller, A. Lopez, V. Koltun, and A. Dosovitskiy. End-to-End Driving via Conditional Imitation Learning. IEEE International Conference on Robotics and Automation (ICRA), 2018. arXiv:1710.02410.","identifier":"arXiv:1710.02410","url":"https://arxiv.org/abs/1710.02410"},{"text":"F. Codevilla, E. Santana, A. M. Lopez, and A. Gaidon. Exploring the Limitations of Behavior Cloning for Autonomous Driving. IEEE/CVF International Conference on Computer Vision (ICCV), 2019. arXiv:1904.08980.","identifier":"arXiv:1904.08980","url":"https://arxiv.org/abs/1904.08980"},{"text":"P. de Haan, D. Jayaraman, and S. Levine. Causal Confusion in Imitation Learning. Advances in Neural Information Processing Systems (NeurIPS), 2019. arXiv:1905.11979.","identifier":"arXiv:1905.11979","url":"https://arxiv.org/abs/1905.11979"},{"text":"M. Bansal, A. Krizhevsky, and A. Ogale. ChauffeurNet: Learning to Drive by Imitating the Best and Synthesizing the Worst. Robotics: Science and Systems (RSS), 2019. arXiv:1812.03079.","identifier":"arXiv:1812.03079","url":"https://arxiv.org/abs/1812.03079"},{"text":"A. Mandlekar, D. Xu, J. Wong, S. Nasiriany, C. Wang, R. Kulkarni, L. Fei-Fei, S. Savarese, Y. Zhu, and R. Martin-Martin. What Matters in Learning from Offline Human Demonstrations for Robot Manipulation. Conference on Robot Learning (CoRL), 2021. arXiv:2108.03298.","identifier":"arXiv:2108.03298","url":"https://arxiv.org/abs/2108.03298"},{"text":"T. Z. Zhao, V. Kumar, S. Levine, and C. Finn. Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware. Robotics: Science and Systems (RSS), 2023. arXiv:2304.13705.","identifier":"arXiv:2304.13705","url":"https://arxiv.org/abs/2304.13705"},{"text":"C. Chi, S. Feng, Y. Du, Z. Xu, E. Cousineau, B. Burchfiel, and S. Song. Diffusion Policy: Visuomotor Policy Learning via Action Diffusion. Robotics: Science and Systems (RSS), 2023. arXiv:2303.04137.","identifier":"arXiv:2303.04137","url":"https://arxiv.org/abs/2303.04137"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Sridhar, D. Shah, C. Glossop, and S. Levine. NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration. arXiv:2310.07896, 2023.","identifier":"arXiv:2310.07896","url":"https://arxiv.org/abs/2310.07896"},{"text":"D. Chen, B. Zhou, V. Koltun, and P. Kraehenbuehl. Learning by Cheating. Conference on Robot Learning (CoRL), 2019. arXiv:1912.12294.","identifier":"arXiv:1912.12294","url":"https://arxiv.org/abs/1912.12294"},{"text":"G. E. Hinton, N. Srivastava, A. Krizhevsky, I. Sutskever, and R. R. Salakhutdinov. Improving Neural Networks by Preventing Co-Adaptation of Feature Detectors. arXiv:1207.0580, 2012.","identifier":"arXiv:1207.0580","url":"https://arxiv.org/abs/1207.0580"},{"text":"N. Neverova, C. Wolf, G. Taylor, and F. Nebout. ModDrop: Adaptive Multi-Modal Gesture Recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence, 38(8), 2016. arXiv:1501.00102.","identifier":"arXiv:1501.00102","url":"https://arxiv.org/abs/1501.00102"},{"text":"A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin. Attention Is All You Need. Advances in Neural Information Processing Systems (NeurIPS), 2017. arXiv:1706.03762.","identifier":"arXiv:1706.03762","url":"https://arxiv.org/abs/1706.03762"},{"text":"V. Makoviychuk, L. Wawrzyniak, Y. Guo, M. Lu, K. Storey, M. Macklin, D. Hoeller, N. Rudin, A. Allshire, A. Handa, and G. State. Isaac Gym: High Performance GPU-Based Physics Simulation for Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal, C. Yu, Q. Yu, J. Liu, N. Rudin, D. Hoeller, J. L. Yuan, R. Singh, Y. Guo, H. Mazhar, A. Mandlekar, B. Babich, G. State, M. Hutter, and A. Garg. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. IEEE Robotics and Automation Letters, 8(6), 2023. arXiv:2301.04195.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"P. Anderson, A. Chang, D. S. Chaplot, A. Dosovitskiy, S. Gupta, V. Koltun, J. Kosecka, J. Malik, R. Mottaghi, M. Savva, and A. R. Zamir. On Evaluation of Embodied Navigation Agents. arXiv:1807.06757, 2018.","identifier":"arXiv:1807.06757","url":"https://arxiv.org/abs/1807.06757"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2609.00007","metadata":{"title":"Future-Range Conditioning for Imitation-Learned Quadruped Navigation","authors":[{"name":"Rachel Kim","affiliation":"Robotics Institute, Georgia Institute of Technology, Atlanta, GA, USA"},{"name":"Ivan Sokolov","affiliation":"Robotics Institute, Georgia Institute of Technology, Atlanta, GA, USA"},{"name":"Marcus Hill","affiliation":"Robotics Institute, Georgia Institute of Technology, Atlanta, GA, USA"}],"publication_date":"2026-09-17","abstract":"Imitation-learned navigation policies for legged robots inherit the range representation of the data they are trained on: a policy that observes only the present range scan tends to reproduce how a particular demonstrator reacted to that scan rather than what the robot will actually encounter. This paper studies whether conditioning an imitation policy on a predicted, control-aligned future range field reduces this dependence. A transformer policy consumes the present range scan, a short range history, and the future free range predicted by a FutureRay-style module along the directions the robot is expected to travel under its current command. To keep the policy usable when the predictor is wrong, training applies range-representation dropout that masks or corrupts the future input, so that the policy learns to fall back on present-range behavior. The approach is evaluated with a simulated ANYmal quadruped in Isaac Lab, trained on 12 h of teleoperation from three operators plus 2000 simulator rollouts across 20 cluttered scenes, and tested on 10 unseen cluttered scenes. The conditioned policy reaches a success rate of 88.4 percent, against 76.1 percent for present-range-only imitation and 74.3 percent for a rule-based reactive baseline, raises collision-free distance from 3.8 to 4.2 km per collision, and attains an oracle-parity ratio of 0.81 relative to a policy trained on ground-truth future range. Future conditioning without dropout yields no gain over present-range imitation. Hardware validation remains future work.","references":[{"text":"T. Zang, S. Wang, H. Huang, and S. Cheng. FutureRay: Control-Aligned Future Range for Agile Quadruped Navigation. arXiv:2608.01197, 2026.","identifier":"arXiv:2608.01197","url":"https://arxiv.org/abs/2608.01197"},{"text":"M. Hutter, C. Gehring, D. Jud, A. Lauber, C. D. Bellicoso, V. Tsounis, J. Hwangbo, K. Bodie, P. Fankhauser, M. Bloesch, R. Diethelm, S. Bachmann, A. Melzer, and M. Hoepflinger. ANYmal - A Highly Mobile and Dynamic Quadrupedal Robot. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2016.","identifier":"doi:10.1109/IROS.2016.7758092","url":"https://doi.org/10.1109/IROS.2016.7758092"},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. Science Robotics, 4(26), 2019. arXiv:1901.08652.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. Science Robotics, 5(47), 2020. arXiv:2010.11251.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. Science Robotics, 7(62), 2022. arXiv:2201.08117.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. Conference on Robot Learning (CoRL), 2021. arXiv:2109.11978.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. Robotics: Science and Systems (RSS), 2021. arXiv:2107.04034.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. Conference on Robot Learning (CoRL), 2022. arXiv:2211.07638.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. Conference on Robot Learning (CoRL), 2023. arXiv:2309.05665.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. IEEE International Conference on Robotics and Automation (ICRA), 2024. arXiv:2309.14341.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2022. arXiv:2209.12827.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, L. Wellhausen, F. Farshidian, and M. Hutter. Learning a State Representation and Navigation in Cluttered and Dynamic Environments. IEEE Robotics and Automation Letters, 6(3), 2021. arXiv:2103.04351.","identifier":"arXiv:2103.04351","url":"https://arxiv.org/abs/2103.04351"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. Science Robotics, 9(88), 2024. arXiv:2306.14874.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"L. Wellhausen and M. Hutter. ArtPlanner: Robust Legged Robot Navigation in the Field. Field Robotics, 3, 2023. arXiv:2303.01420.","identifier":"arXiv:2303.01420","url":"https://arxiv.org/abs/2303.01420"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. Robotics: Science and Systems (RSS), 2023. arXiv:2305.08510.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"J. Truong, D. Yarats, T. Li, F. Meier, S. Chernova, D. Batra, and A. Rai. Learning Navigation Skills for Legged Robots with Learned Robot Embodiments. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2021. arXiv:2011.06188.","identifier":"arXiv:2011.06188","url":"https://arxiv.org/abs/2011.06188"},{"text":"S. Kareer, N. Yokoyama, D. Batra, S. Ha, and J. Truong. ViNL: Visual Navigation and Locomotion Over Obstacles. IEEE International Conference on Robotics and Automation (ICRA), 2023. arXiv:2210.14791.","identifier":"arXiv:2210.14791","url":"https://arxiv.org/abs/2210.14791"},{"text":"J. Lee, M. Bjelonic, A. Reske, L. Wellhausen, T. Miki, and M. Hutter. Learning Robust Autonomous Navigation and Locomotion for Wheeled-Legged Robots. Science Robotics, 9(89), 2024. arXiv:2405.01792.","identifier":"arXiv:2405.01792","url":"https://arxiv.org/abs/2405.01792"},{"text":"O. Khatib. Real-Time Obstacle Avoidance for Manipulators and Mobile Robots. The International Journal of Robotics Research, 5(1):90-98, 1986.","identifier":"doi:10.1177/027836498600500106","url":"https://doi.org/10.1177/027836498600500106"},{"text":"D. Fox, W. Burgard, and S. Thrun. The Dynamic Window Approach to Collision Avoidance. IEEE Robotics and Automation Magazine, 4(1):23-33, 1997.","identifier":"doi:10.1109/100.580977","url":"https://doi.org/10.1109/100.580977"},{"text":"M. Pfeiffer, M. Schaeuble, J. Nieto, R. Siegwart, and C. Cadena. From Perception to Decision: A Data-Driven Approach to End-to-End Motion Planning for Autonomous Ground Robots. IEEE International Conference on Robotics and Automation (ICRA), 2017. arXiv:1609.07910.","identifier":"arXiv:1609.07910","url":"https://arxiv.org/abs/1609.07910"},{"text":"L. Tai, G. Paolo, and M. Liu. Virtual-to-Real Deep Reinforcement Learning: Continuous Control of Mobile Robots for Mapless Navigation. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2017. arXiv:1703.00420.","identifier":"arXiv:1703.00420","url":"https://arxiv.org/abs/1703.00420"},{"text":"G. Kahn, A. Villaflor, B. Ding, P. Abbeel, and S. Levine. Self-Supervised Deep Reinforcement Learning with Generalized Computation Graphs for Robot Navigation. IEEE International Conference on Robotics and Automation (ICRA), 2018. arXiv:1709.10489.","identifier":"arXiv:1709.10489","url":"https://arxiv.org/abs/1709.10489"},{"text":"G. Kahn, P. Abbeel, and S. Levine. BADGR: An Autonomous Self-Supervised Learning-Based Navigation System. IEEE Robotics and Automation Letters, 6(2), 2021. arXiv:2002.05700.","identifier":"arXiv:2002.05700","url":"https://arxiv.org/abs/2002.05700"},{"text":"C. Finn and S. Levine. Deep Visual Foresight for Planning Robot Motion. IEEE International Conference on Robotics and Automation (ICRA), 2017. arXiv:1610.00696.","identifier":"arXiv:1610.00696","url":"https://arxiv.org/abs/1610.00696"},{"text":"S. Ross, G. J. Gordon, and J. A. Bagnell. A Reduction of Imitation Learning and Structured Prediction to No-Regret Online Learning. International Conference on Artificial Intelligence and Statistics (AISTATS), 2011. arXiv:1011.0686.","identifier":"arXiv:1011.0686","url":"https://arxiv.org/abs/1011.0686"},{"text":"M. Bojarski, D. Del Testa, D. Dworakowski, B. Firner, B. Flepp, P. Goyal, L. D. Jackel, M. Monfort, U. Muller, J. Zhang, X. Zhang, J. Zhao, and K. Zieba. End to End Learning for Self-Driving Cars. arXiv:1604.07316, 2016.","identifier":"arXiv:1604.07316","url":"https://arxiv.org/abs/1604.07316"},{"text":"F. Codevilla, M. Mueller, A. Lopez, V. Koltun, and A. Dosovitskiy. End-to-End Driving via Conditional Imitation Learning. IEEE International Conference on Robotics and Automation (ICRA), 2018. arXiv:1710.02410.","identifier":"arXiv:1710.02410","url":"https://arxiv.org/abs/1710.02410"},{"text":"F. Codevilla, E. Santana, A. M. Lopez, and A. Gaidon. Exploring the Limitations of Behavior Cloning for Autonomous Driving. IEEE/CVF International Conference on Computer Vision (ICCV), 2019. arXiv:1904.08980.","identifier":"arXiv:1904.08980","url":"https://arxiv.org/abs/1904.08980"},{"text":"P. de Haan, D. Jayaraman, and S. Levine. Causal Confusion in Imitation Learning. Advances in Neural Information Processing Systems (NeurIPS), 2019. arXiv:1905.11979.","identifier":"arXiv:1905.11979","url":"https://arxiv.org/abs/1905.11979"},{"text":"M. Bansal, A. Krizhevsky, and A. Ogale. ChauffeurNet: Learning to Drive by Imitating the Best and Synthesizing the Worst. Robotics: Science and Systems (RSS), 2019. arXiv:1812.03079.","identifier":"arXiv:1812.03079","url":"https://arxiv.org/abs/1812.03079"},{"text":"A. Mandlekar, D. Xu, J. Wong, S. Nasiriany, C. Wang, R. Kulkarni, L. Fei-Fei, S. Savarese, Y. Zhu, and R. Martin-Martin. What Matters in Learning from Offline Human Demonstrations for Robot Manipulation. Conference on Robot Learning (CoRL), 2021. arXiv:2108.03298.","identifier":"arXiv:2108.03298","url":"https://arxiv.org/abs/2108.03298"},{"text":"T. Z. Zhao, V. Kumar, S. Levine, and C. Finn. Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware. Robotics: Science and Systems (RSS), 2023. arXiv:2304.13705.","identifier":"arXiv:2304.13705","url":"https://arxiv.org/abs/2304.13705"},{"text":"C. Chi, S. Feng, Y. Du, Z. Xu, E. Cousineau, B. Burchfiel, and S. Song. Diffusion Policy: Visuomotor Policy Learning via Action Diffusion. Robotics: Science and Systems (RSS), 2023. arXiv:2303.04137.","identifier":"arXiv:2303.04137","url":"https://arxiv.org/abs/2303.04137"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Sridhar, D. Shah, C. Glossop, and S. Levine. NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration. arXiv:2310.07896, 2023.","identifier":"arXiv:2310.07896","url":"https://arxiv.org/abs/2310.07896"},{"text":"D. Chen, B. Zhou, V. Koltun, and P. Kraehenbuehl. Learning by Cheating. Conference on Robot Learning (CoRL), 2019. arXiv:1912.12294.","identifier":"arXiv:1912.12294","url":"https://arxiv.org/abs/1912.12294"},{"text":"G. E. Hinton, N. Srivastava, A. Krizhevsky, I. Sutskever, and R. R. Salakhutdinov. Improving Neural Networks by Preventing Co-Adaptation of Feature Detectors. arXiv:1207.0580, 2012.","identifier":"arXiv:1207.0580","url":"https://arxiv.org/abs/1207.0580"},{"text":"N. Neverova, C. Wolf, G. Taylor, and F. Nebout. ModDrop: Adaptive Multi-Modal Gesture Recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence, 38(8), 2016. arXiv:1501.00102.","identifier":"arXiv:1501.00102","url":"https://arxiv.org/abs/1501.00102"},{"text":"A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin. Attention Is All You Need. Advances in Neural Information Processing Systems (NeurIPS), 2017. arXiv:1706.03762.","identifier":"arXiv:1706.03762","url":"https://arxiv.org/abs/1706.03762"},{"text":"V. Makoviychuk, L. Wawrzyniak, Y. Guo, M. Lu, K. Storey, M. Macklin, D. Hoeller, N. Rudin, A. Allshire, A. Handa, and G. State. Isaac Gym: High Performance GPU-Based Physics Simulation for Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal, C. Yu, Q. Yu, J. Liu, N. Rudin, D. Hoeller, J. L. Yuan, R. Singh, Y. Guo, H. Mazhar, A. Mandlekar, B. Babich, G. State, M. Hutter, and A. Garg. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. IEEE Robotics and Automation Letters, 8(6), 2023. arXiv:2301.04195.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"P. Anderson, A. Chang, D. S. Chaplot, A. Dosovitskiy, S. Gupta, V. Koltun, J. Kosecka, J. Malik, R. Mottaghi, M. Savva, and A. R. Zamir. On Evaluation of Embodied Navigation Agents. arXiv:1807.06757, 2018.","identifier":"arXiv:1807.06757","url":"https://arxiv.org/abs/1807.06757"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2609.00007","slug":"2609.00007","registered_at":"2026-09-20T18:31:00Z","updated_at":"2026-09-20T21:31:00Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2609.00007","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2609.00007","tex_path":"articles/2609.00007/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-09-20T20:48:00Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2609.00007"},{"kagg_id":"KAGG:2609.00006","slug":"2609.00006","title":"System Identification versus Learned Models for Commodity Quadruped Velocity Interfaces","authors":[{"name":"Lin Zhao","affiliation":"Perception and Control Group, Technical University of Munich, Munich, Germany"},{"name":"Maximilian Weber","affiliation":"Perception and Control Group, Technical University of Munich, Munich, Germany"},{"name":"Julia Hoffmann","affiliation":"Perception and Control Group, Technical University of Munich, Munich, Germany"},{"name":"Omar Farouk","affiliation":"Perception and Control Group, Technical University of Munich, Munich, Germany"}],"affiliation":null,"abstract":"Navigation stacks for commodity quadrupeds command the robot through a vendor velocity interface whose response to commands, including delay, lag, cross-axis coupling, and saturation, is undocumented and differs between platforms. Planners therefore need a calibrated model of this interface, but it is unclear whether classical system identification or a learned model should provide it. This paper compares, under matched probe budgets and on the same planner-conditioned probe data, structured identification of a state-delay-plus-saturation model, a Gaussian-process surrogate, and a multilayer perceptron with one million parameters. Scoring uses navigation-relevant criteria: worst-direction tracking error, extrapolation error on unseen command regimes, calibration time, and sensitivity to command noise. On a Unitree Go2 and a Unitree B2, with 60 probe sessions per platform, 12 m of probe motion per session, and 100 held-out command regimes per platform, the structured model attains 9.2% worst-direction tracking error and 14.6% extrapolation error, against 9.7% and 16.3% for the Gaussian process and 9.5% and 15.6% for the perceptron, with the shortest calibration time (44 s) and the lowest noise sensitivity (2.4 points at 3% noise). The margins are modest, but the failure cases are informative: at 3% command noise the perceptron's extrapolation error on the B2 rises to 21%, whereas the structured model stays at 16.1%; conversely, the perceptron captures the B2's saturation knee 9% better in-regime. The resulting decision rule is to identify the structured model by default, and adopt a learned model only when the saturation shape is the binding constraint.","publication_date":"2026-09-14","registered_at":"2026-09-17T14:45:00Z","updated_at":"2026-09-17T18:09:00Z","status":"published","html_path":null,"pdf_path":"articles/2609.00006/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"T. Zang, H. Huang, S. Cheng, and W. Zhang. GAUGE: Planner-Conditioned Active Calibration of Opaque Quadruped Velocity Interfaces. arXiv:2607.02841, 2026.","identifier":"arXiv:2607.02841","url":"https://arxiv.org/abs/2607.02841"},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. Science Robotics, arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. Science Robotics, arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. Robotics: Science and Systems, arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. Conference on Robot Learning, arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. Science Robotics, arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"G. B. Margolis, G. Yang, K. Paigwar, T. Chen, and P. Agrawal. Rapid Locomotion via Reinforcement Learning. Robotics: Science and Systems, arXiv:2205.02824, 2022.","identifier":"arXiv:2205.02824","url":"https://arxiv.org/abs/2205.02824"},{"text":"G. B. Margolis and P. Agrawal. Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior. Conference on Robot Learning, arXiv:2212.03238, 2022.","identifier":"arXiv:2212.03238","url":"https://arxiv.org/abs/2212.03238"},{"text":"J. Tan, T. Zhang, E. Coumans, A. Iscen, Y. Bai, D. Hafner, S. Bohez, and V. Vanhoucke. Sim-to-Real: Learning Agile Locomotion for Quadruped Robots. Robotics: Science and Systems, arXiv:1804.10332, 2018.","identifier":"arXiv:1804.10332","url":"https://arxiv.org/abs/1804.10332"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. IEEE International Conference on Robotics and Automation, arXiv:1710.06537, 2018.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"},{"text":"W. Yu, J. Tan, C. K. Liu, and G. Turk. Preparing for the Unknown: Learning a Universal Policy with Online System Identification. Robotics: Science and Systems, arXiv:1702.02453, 2017.","identifier":"arXiv:1702.02453","url":"https://arxiv.org/abs/1702.02453"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. IEEE/RSJ International Conference on Intelligent Robots and Systems, arXiv:1703.06907, 2017.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"},{"text":"J. Schoukens and L. Ljung. Nonlinear System Identification: A User-Oriented Road Map. IEEE Control Systems Magazine, arXiv:1902.00683, 2019.","identifier":"arXiv:1902.00683","url":"https://arxiv.org/abs/1902.00683"},{"text":"G. Pillonetto, A. Aravkin, D. Gedon, L. Ljung, A. H. Ribeiro, and T. B. Schön. Deep Networks for System Identification: A Survey. arXiv:2301.12832, 2023.","identifier":"arXiv:2301.12832","url":"https://arxiv.org/abs/2301.12832"},{"text":"S. L. Brunton, J. L. Proctor, and J. N. Kutz. Discovering Governing Equations from Data by Sparse Identification of Nonlinear Dynamical Systems. Proceedings of the National Academy of Sciences, arXiv:1509.03580, 2016.","identifier":"arXiv:1509.03580","url":"https://arxiv.org/abs/1509.03580"},{"text":"H. Mania, M. I. Jordan, and B. Recht. Active Learning for Nonlinear System Identification with Guarantees. Journal of Machine Learning Research, arXiv:2006.10277, 2022.","identifier":"arXiv:2006.10277","url":"https://arxiv.org/abs/2006.10277"},{"text":"M. Buisson-Fenet, F. Solowjow, and S. Trimpe. Actively Learning Gaussian Process Dynamics. Learning for Dynamics and Control, arXiv:1911.09946, 2020.","identifier":"arXiv:1911.09946","url":"https://arxiv.org/abs/1911.09946"},{"text":"T. Lew, A. Sharma, J. Harrison, A. Bylard, and M. Pavone. Safe Active Dynamics Learning and Control: A Sequential Exploration-Exploitation Framework. IEEE Transactions on Robotics, arXiv:2008.11700, 2022.","identifier":"arXiv:2008.11700","url":"https://arxiv.org/abs/2008.11700"},{"text":"M. P. Deisenroth, D. Fox, and C. E. Rasmussen. Gaussian Processes for Data-Efficient Learning in Robotics and Control. IEEE Transactions on Pattern Analysis and Machine Intelligence, arXiv:1502.02860, 2015.","identifier":"arXiv:1502.02860","url":"https://arxiv.org/abs/1502.02860"},{"text":"C. E. Rasmussen and C. K. I. Williams. Gaussian Processes for Machine Learning. MIT Press, ISBN 0-262-18253-X, 2006.","identifier":"ISBN:0-262-18253-X","url":"http://gaussianprocess.org/gpml/"},{"text":"L. Hewing, J. Kabzan, and M. N. Zeilinger. Cautious Model Predictive Control Using Gaussian Process Regression. IEEE Transactions on Control Systems Technology, arXiv:1705.10702, 2017.","identifier":"arXiv:1705.10702","url":"https://arxiv.org/abs/1705.10702"},{"text":"F. Berkenkamp, M. Turchetta, A. P. Schoellig, and A. Krause. Safe Model-based Reinforcement Learning with Stability Guarantees. Advances in Neural Information Processing Systems, arXiv:1705.08551, 2017.","identifier":"arXiv:1705.08551","url":"https://arxiv.org/abs/1705.08551"},{"text":"K. Chua, R. Calandra, R. McAllister, and S. Levine. Deep Reinforcement Learning in a Handful of Trials Using Probabilistic Dynamics Models. Advances in Neural Information Processing Systems, arXiv:1805.12114, 2018.","identifier":"arXiv:1805.12114","url":"https://arxiv.org/abs/1805.12114"},{"text":"A. Nagabandi, G. Kahn, R. S. Fearing, and S. Levine. Neural Network Dynamics for Model-Based Deep Reinforcement Learning with Model-Free Fine-Tuning. IEEE International Conference on Robotics and Automation, arXiv:1708.02596, 2018.","identifier":"arXiv:1708.02596","url":"https://arxiv.org/abs/1708.02596"},{"text":"A. Nagabandi, I. Clavera, S. Liu, R. S. Fearing, P. Abbeel, S. Levine, and C. Finn. Learning to Adapt in Dynamic, Real-World Environments through Meta-Reinforcement Learning. International Conference on Learning Representations, arXiv:1803.11347, 2019.","identifier":"arXiv:1803.11347","url":"https://arxiv.org/abs/1803.11347"},{"text":"G. Torrente, E. Kaufmann, P. Föhn, and D. Scaramuzza. Data-Driven MPC for Quadrotors. IEEE Robotics and Automation Letters, arXiv:2102.05773, 2021.","identifier":"arXiv:2102.05773","url":"https://arxiv.org/abs/2102.05773"},{"text":"L. Bauersfeld, E. Kaufmann, P. Foehn, S. Sun, and D. Scaramuzza. NeuroBEM: Hybrid Aerodynamic Quadrotor Model. Robotics: Science and Systems, arXiv:2106.08015, 2021.","identifier":"arXiv:2106.08015","url":"https://arxiv.org/abs/2106.08015"},{"text":"M. O'Connell, G. Shi, X. Shi, K. Azizzadenesheli, A. Anandkumar, Y. Yue, and S.-J. Chung. Neural-Fly Enables Rapid Learning for Agile Flight in Strong Winds. Science Robotics, arXiv:2205.06908, 2022.","identifier":"arXiv:2205.06908","url":"https://arxiv.org/abs/2205.06908"},{"text":"D. Kim, J. Di Carlo, B. Katz, G. Bledt, and S. Kim. Highly Dynamic Quadruped Locomotion via Whole-Body Impulse Control and Model Predictive Control. arXiv:1909.06586, 2019.","identifier":"arXiv:1909.06586","url":"https://arxiv.org/abs/1909.06586"},{"text":"R. Grandia, F. Jenelten, S. Yang, F. Farshidian, and M. Hutter. Perceptive Locomotion through Nonlinear Model-Predictive Control. IEEE Transactions on Robotics, arXiv:2208.08373, 2023.","identifier":"arXiv:2208.08373","url":"https://arxiv.org/abs/2208.08373"},{"text":"X. Xiao, B. Liu, G. Warnell, and P. Stone. Motion Planning and Control for Mobile Robot Navigation Using Machine Learning: A Survey. Autonomous Robots, arXiv:2011.13112, 2022.","identifier":"arXiv:2011.13112","url":"https://arxiv.org/abs/2011.13112"},{"text":"X. Xiao, J. Biswas, and P. Stone. Learning Inverse Kinodynamics for Accurate High-Speed Off-Road Navigation on Unstructured Terrain. IEEE Robotics and Automation Letters, arXiv:2102.12667, 2021.","identifier":"arXiv:2102.12667","url":"https://arxiv.org/abs/2102.12667"},{"text":"H. Karnan, K. S. Sikand, P. Atreya, S. Rabiee, X. Xiao, G. Warnell, P. Stone, and J. Biswas. VI-IKD: High-Speed Accurate Off-Road Navigation Using Learned Visual-Inertial Inverse Kinodynamics. IEEE International Conference on Robotics and Automation, arXiv:2203.15983, 2022.","identifier":"arXiv:2203.15983","url":"https://arxiv.org/abs/2203.15983"},{"text":"L. Wellhausen and M. Hutter. ArtPlanner: Robust Legged Robot Navigation in the Field. Field Robotics, arXiv:2303.01420, 2023.","identifier":"arXiv:2303.01420","url":"https://arxiv.org/abs/2303.01420"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. IEEE/RSJ International Conference on Intelligent Robots and Systems, arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"J. Truong, D. Yarats, T. Li, F. Meier, S. Chernova, D. Batra, and A. Rai. Learning Navigation Skills for Legged Robots with Learned Robot Embeddings. IEEE/RSJ International Conference on Intelligent Robots and Systems, arXiv:2011.12255, 2021.","identifier":"arXiv:2011.12255","url":"https://arxiv.org/abs/2011.12255"},{"text":"Z. Fu, A. Kumar, A. Agarwal, H. Qi, J. Malik, and D. Pathak. Coupling Vision and Proprioception for Navigation of Legged Robots. IEEE/CVF Conference on Computer Vision and Pattern Recognition, arXiv:2112.02094, 2022.","identifier":"arXiv:2112.02094","url":"https://arxiv.org/abs/2112.02094"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"B. Lakshminarayanan, A. Pritzel, and C. Blundell. Simple and Scalable Predictive Uncertainty Estimation Using Deep Ensembles. Advances in Neural Information Processing Systems, arXiv:1612.01474, 2017.","identifier":"arXiv:1612.01474","url":"https://arxiv.org/abs/1612.01474"},{"text":"C. Guo, G. Pleiss, Y. Sun, and K. Q. Weinberger. On Calibration of Modern Neural Networks. International Conference on Machine Learning, arXiv:1706.04599, 2017.","identifier":"arXiv:1706.04599","url":"https://arxiv.org/abs/1706.04599"},{"text":"J. R. Gardner, G. Pleiss, D. Bindel, K. Q. Weinberger, and A. G. Wilson. GPyTorch: Blackbox Matrix-Matrix Gaussian Process Inference with GPU Acceleration. Advances in Neural Information Processing Systems, arXiv:1809.11165, 2018.","identifier":"arXiv:1809.11165","url":"https://arxiv.org/abs/1809.11165"},{"text":"D. P. Kingma and J. Ba. Adam: A Method for Stochastic Optimization. International Conference on Learning Representations, arXiv:1412.6980, 2015.","identifier":"arXiv:1412.6980","url":"https://arxiv.org/abs/1412.6980"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2609.00006","metadata":{"title":"System Identification versus Learned Models for Commodity Quadruped Velocity Interfaces","authors":[{"name":"Lin Zhao","affiliation":"Perception and Control Group, Technical University of Munich, Munich, Germany"},{"name":"Maximilian Weber","affiliation":"Perception and Control Group, Technical University of Munich, Munich, Germany"},{"name":"Julia Hoffmann","affiliation":"Perception and Control Group, Technical University of Munich, Munich, Germany"},{"name":"Omar Farouk","affiliation":"Perception and Control Group, Technical University of Munich, Munich, Germany"}],"publication_date":"2026-09-14","abstract":"Navigation stacks for commodity quadrupeds command the robot through a vendor velocity interface whose response to commands, including delay, lag, cross-axis coupling, and saturation, is undocumented and differs between platforms. Planners therefore need a calibrated model of this interface, but it is unclear whether classical system identification or a learned model should provide it. This paper compares, under matched probe budgets and on the same planner-conditioned probe data, structured identification of a state-delay-plus-saturation model, a Gaussian-process surrogate, and a multilayer perceptron with one million parameters. Scoring uses navigation-relevant criteria: worst-direction tracking error, extrapolation error on unseen command regimes, calibration time, and sensitivity to command noise. On a Unitree Go2 and a Unitree B2, with 60 probe sessions per platform, 12 m of probe motion per session, and 100 held-out command regimes per platform, the structured model attains 9.2% worst-direction tracking error and 14.6% extrapolation error, against 9.7% and 16.3% for the Gaussian process and 9.5% and 15.6% for the perceptron, with the shortest calibration time (44 s) and the lowest noise sensitivity (2.4 points at 3% noise). The margins are modest, but the failure cases are informative: at 3% command noise the perceptron's extrapolation error on the B2 rises to 21%, whereas the structured model stays at 16.1%; conversely, the perceptron captures the B2's saturation knee 9% better in-regime. The resulting decision rule is to identify the structured model by default, and adopt a learned model only when the saturation shape is the binding constraint.","references":[{"text":"T. Zang, H. Huang, S. Cheng, and W. Zhang. GAUGE: Planner-Conditioned Active Calibration of Opaque Quadruped Velocity Interfaces. arXiv:2607.02841, 2026.","identifier":"arXiv:2607.02841","url":"https://arxiv.org/abs/2607.02841"},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. Science Robotics, arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. Science Robotics, arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. Robotics: Science and Systems, arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. Conference on Robot Learning, arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. Science Robotics, arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"G. B. Margolis, G. Yang, K. Paigwar, T. Chen, and P. Agrawal. Rapid Locomotion via Reinforcement Learning. Robotics: Science and Systems, arXiv:2205.02824, 2022.","identifier":"arXiv:2205.02824","url":"https://arxiv.org/abs/2205.02824"},{"text":"G. B. Margolis and P. Agrawal. Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior. Conference on Robot Learning, arXiv:2212.03238, 2022.","identifier":"arXiv:2212.03238","url":"https://arxiv.org/abs/2212.03238"},{"text":"J. Tan, T. Zhang, E. Coumans, A. Iscen, Y. Bai, D. Hafner, S. Bohez, and V. Vanhoucke. Sim-to-Real: Learning Agile Locomotion for Quadruped Robots. Robotics: Science and Systems, arXiv:1804.10332, 2018.","identifier":"arXiv:1804.10332","url":"https://arxiv.org/abs/1804.10332"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. IEEE International Conference on Robotics and Automation, arXiv:1710.06537, 2018.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"},{"text":"W. Yu, J. Tan, C. K. Liu, and G. Turk. Preparing for the Unknown: Learning a Universal Policy with Online System Identification. Robotics: Science and Systems, arXiv:1702.02453, 2017.","identifier":"arXiv:1702.02453","url":"https://arxiv.org/abs/1702.02453"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. IEEE/RSJ International Conference on Intelligent Robots and Systems, arXiv:1703.06907, 2017.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"},{"text":"J. Schoukens and L. Ljung. Nonlinear System Identification: A User-Oriented Road Map. IEEE Control Systems Magazine, arXiv:1902.00683, 2019.","identifier":"arXiv:1902.00683","url":"https://arxiv.org/abs/1902.00683"},{"text":"G. Pillonetto, A. Aravkin, D. Gedon, L. Ljung, A. H. Ribeiro, and T. B. Schön. Deep Networks for System Identification: A Survey. arXiv:2301.12832, 2023.","identifier":"arXiv:2301.12832","url":"https://arxiv.org/abs/2301.12832"},{"text":"S. L. Brunton, J. L. Proctor, and J. N. Kutz. Discovering Governing Equations from Data by Sparse Identification of Nonlinear Dynamical Systems. Proceedings of the National Academy of Sciences, arXiv:1509.03580, 2016.","identifier":"arXiv:1509.03580","url":"https://arxiv.org/abs/1509.03580"},{"text":"H. Mania, M. I. Jordan, and B. Recht. Active Learning for Nonlinear System Identification with Guarantees. Journal of Machine Learning Research, arXiv:2006.10277, 2022.","identifier":"arXiv:2006.10277","url":"https://arxiv.org/abs/2006.10277"},{"text":"M. Buisson-Fenet, F. Solowjow, and S. Trimpe. Actively Learning Gaussian Process Dynamics. Learning for Dynamics and Control, arXiv:1911.09946, 2020.","identifier":"arXiv:1911.09946","url":"https://arxiv.org/abs/1911.09946"},{"text":"T. Lew, A. Sharma, J. Harrison, A. Bylard, and M. Pavone. Safe Active Dynamics Learning and Control: A Sequential Exploration-Exploitation Framework. IEEE Transactions on Robotics, arXiv:2008.11700, 2022.","identifier":"arXiv:2008.11700","url":"https://arxiv.org/abs/2008.11700"},{"text":"M. P. Deisenroth, D. Fox, and C. E. Rasmussen. Gaussian Processes for Data-Efficient Learning in Robotics and Control. IEEE Transactions on Pattern Analysis and Machine Intelligence, arXiv:1502.02860, 2015.","identifier":"arXiv:1502.02860","url":"https://arxiv.org/abs/1502.02860"},{"text":"C. E. Rasmussen and C. K. I. Williams. Gaussian Processes for Machine Learning. MIT Press, ISBN 0-262-18253-X, 2006.","identifier":"ISBN:0-262-18253-X","url":"http://gaussianprocess.org/gpml/"},{"text":"L. Hewing, J. Kabzan, and M. N. Zeilinger. Cautious Model Predictive Control Using Gaussian Process Regression. IEEE Transactions on Control Systems Technology, arXiv:1705.10702, 2017.","identifier":"arXiv:1705.10702","url":"https://arxiv.org/abs/1705.10702"},{"text":"F. Berkenkamp, M. Turchetta, A. P. Schoellig, and A. Krause. Safe Model-based Reinforcement Learning with Stability Guarantees. Advances in Neural Information Processing Systems, arXiv:1705.08551, 2017.","identifier":"arXiv:1705.08551","url":"https://arxiv.org/abs/1705.08551"},{"text":"K. Chua, R. Calandra, R. McAllister, and S. Levine. Deep Reinforcement Learning in a Handful of Trials Using Probabilistic Dynamics Models. Advances in Neural Information Processing Systems, arXiv:1805.12114, 2018.","identifier":"arXiv:1805.12114","url":"https://arxiv.org/abs/1805.12114"},{"text":"A. Nagabandi, G. Kahn, R. S. Fearing, and S. Levine. Neural Network Dynamics for Model-Based Deep Reinforcement Learning with Model-Free Fine-Tuning. IEEE International Conference on Robotics and Automation, arXiv:1708.02596, 2018.","identifier":"arXiv:1708.02596","url":"https://arxiv.org/abs/1708.02596"},{"text":"A. Nagabandi, I. Clavera, S. Liu, R. S. Fearing, P. Abbeel, S. Levine, and C. Finn. Learning to Adapt in Dynamic, Real-World Environments through Meta-Reinforcement Learning. International Conference on Learning Representations, arXiv:1803.11347, 2019.","identifier":"arXiv:1803.11347","url":"https://arxiv.org/abs/1803.11347"},{"text":"G. Torrente, E. Kaufmann, P. Föhn, and D. Scaramuzza. Data-Driven MPC for Quadrotors. IEEE Robotics and Automation Letters, arXiv:2102.05773, 2021.","identifier":"arXiv:2102.05773","url":"https://arxiv.org/abs/2102.05773"},{"text":"L. Bauersfeld, E. Kaufmann, P. Foehn, S. Sun, and D. Scaramuzza. NeuroBEM: Hybrid Aerodynamic Quadrotor Model. Robotics: Science and Systems, arXiv:2106.08015, 2021.","identifier":"arXiv:2106.08015","url":"https://arxiv.org/abs/2106.08015"},{"text":"M. O'Connell, G. Shi, X. Shi, K. Azizzadenesheli, A. Anandkumar, Y. Yue, and S.-J. Chung. Neural-Fly Enables Rapid Learning for Agile Flight in Strong Winds. Science Robotics, arXiv:2205.06908, 2022.","identifier":"arXiv:2205.06908","url":"https://arxiv.org/abs/2205.06908"},{"text":"D. Kim, J. Di Carlo, B. Katz, G. Bledt, and S. Kim. Highly Dynamic Quadruped Locomotion via Whole-Body Impulse Control and Model Predictive Control. arXiv:1909.06586, 2019.","identifier":"arXiv:1909.06586","url":"https://arxiv.org/abs/1909.06586"},{"text":"R. Grandia, F. Jenelten, S. Yang, F. Farshidian, and M. Hutter. Perceptive Locomotion through Nonlinear Model-Predictive Control. IEEE Transactions on Robotics, arXiv:2208.08373, 2023.","identifier":"arXiv:2208.08373","url":"https://arxiv.org/abs/2208.08373"},{"text":"X. Xiao, B. Liu, G. Warnell, and P. Stone. Motion Planning and Control for Mobile Robot Navigation Using Machine Learning: A Survey. Autonomous Robots, arXiv:2011.13112, 2022.","identifier":"arXiv:2011.13112","url":"https://arxiv.org/abs/2011.13112"},{"text":"X. Xiao, J. Biswas, and P. Stone. Learning Inverse Kinodynamics for Accurate High-Speed Off-Road Navigation on Unstructured Terrain. IEEE Robotics and Automation Letters, arXiv:2102.12667, 2021.","identifier":"arXiv:2102.12667","url":"https://arxiv.org/abs/2102.12667"},{"text":"H. Karnan, K. S. Sikand, P. Atreya, S. Rabiee, X. Xiao, G. Warnell, P. Stone, and J. Biswas. VI-IKD: High-Speed Accurate Off-Road Navigation Using Learned Visual-Inertial Inverse Kinodynamics. IEEE International Conference on Robotics and Automation, arXiv:2203.15983, 2022.","identifier":"arXiv:2203.15983","url":"https://arxiv.org/abs/2203.15983"},{"text":"L. Wellhausen and M. Hutter. ArtPlanner: Robust Legged Robot Navigation in the Field. Field Robotics, arXiv:2303.01420, 2023.","identifier":"arXiv:2303.01420","url":"https://arxiv.org/abs/2303.01420"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. IEEE/RSJ International Conference on Intelligent Robots and Systems, arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"J. Truong, D. Yarats, T. Li, F. Meier, S. Chernova, D. Batra, and A. Rai. Learning Navigation Skills for Legged Robots with Learned Robot Embeddings. IEEE/RSJ International Conference on Intelligent Robots and Systems, arXiv:2011.12255, 2021.","identifier":"arXiv:2011.12255","url":"https://arxiv.org/abs/2011.12255"},{"text":"Z. Fu, A. Kumar, A. Agarwal, H. Qi, J. Malik, and D. Pathak. Coupling Vision and Proprioception for Navigation of Legged Robots. IEEE/CVF Conference on Computer Vision and Pattern Recognition, arXiv:2112.02094, 2022.","identifier":"arXiv:2112.02094","url":"https://arxiv.org/abs/2112.02094"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"B. Lakshminarayanan, A. Pritzel, and C. Blundell. Simple and Scalable Predictive Uncertainty Estimation Using Deep Ensembles. Advances in Neural Information Processing Systems, arXiv:1612.01474, 2017.","identifier":"arXiv:1612.01474","url":"https://arxiv.org/abs/1612.01474"},{"text":"C. Guo, G. Pleiss, Y. Sun, and K. Q. Weinberger. On Calibration of Modern Neural Networks. International Conference on Machine Learning, arXiv:1706.04599, 2017.","identifier":"arXiv:1706.04599","url":"https://arxiv.org/abs/1706.04599"},{"text":"J. R. Gardner, G. Pleiss, D. Bindel, K. Q. Weinberger, and A. G. Wilson. GPyTorch: Blackbox Matrix-Matrix Gaussian Process Inference with GPU Acceleration. Advances in Neural Information Processing Systems, arXiv:1809.11165, 2018.","identifier":"arXiv:1809.11165","url":"https://arxiv.org/abs/1809.11165"},{"text":"D. P. Kingma and J. Ba. Adam: A Method for Stochastic Optimization. International Conference on Learning Representations, arXiv:1412.6980, 2015.","identifier":"arXiv:1412.6980","url":"https://arxiv.org/abs/1412.6980"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2609.00006","slug":"2609.00006","registered_at":"2026-09-17T14:45:00Z","updated_at":"2026-09-17T18:09:00Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2609.00006","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2609.00006","tex_path":"articles/2609.00006/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-09-17T17:02:00Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2609.00006"},{"kagg_id":"KAGG:2609.00005","slug":"2609.00005","title":"Shared Configuration Abstractions for Heterogeneous Fleet Navigation","authors":[{"name":"Hana Sato","affiliation":"Advanced Robotics Group, Toyota Research Institute, Palo Alto, CA, USA"},{"name":"Tomas Keller","affiliation":"Advanced Robotics Group, Toyota Research Institute, Palo Alto, CA, USA"},{"name":"Mei-Lin Chao","affiliation":"Advanced Robotics Group, Toyota Research Institute, Palo Alto, CA, USA"}],"affiliation":null,"abstract":"Heterogeneous robot fleets that must follow, avoid, or rendezvous with one another need more than per-robot planning: each robot must reason about how its peers are shaped, how quickly they can stop, and how these properties change during operation. Configuration-driven local planners such as AgniNav describe a single robot through an abstract configuration of body geometry, sensing, and velocity limits, but they treat other robots as generic obstacles. This paper extends the configuration descriptor into a fleet-shared abstraction. Each robot publishes a versioned configuration handle on the fleet bus, and the unchanged local planner evaluates interaction-critical derived quantities, namely a predicted clearance envelope and a stop distance, as functions of the peer's live configuration, with conservative inflation when a handle is stale. The approach is evaluated in simulation on a mixed fleet of a Go2-class quadruped, a B2-class quadruped, and a wheeled robot across 12 corridor and intersection scenarios with 5 seeds each. Relative to a configuration-blind planner that models peers as footprint blobs, interaction success increases from 77.7 to 91.6 percent, envelope prediction error decreases from 12.2 cm to 10.8 cm, and stop-distance error decreases from 15.3 cm to 14.3 cm, while the 95th-percentile planner overhead is 3.1 ms compared with 3.5 ms. Ablations indicate that the stop-distance derivation, the clearance envelope, and live configuration updates each account for 9.0 to 10.1 percentage points of interaction success. Recorded failure cases illustrate behavior under stale handles and mid-episode payload changes.","publication_date":"2026-09-10","registered_at":"2026-09-17T18:00:00Z","updated_at":"2026-09-17T19:09:00Z","status":"published","html_path":null,"pdf_path":"articles/2609.00005/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Sridhar, D. Shah, C. Glossop, and S. Levine. NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration. arXiv:2310.07896, 2023.","identifier":"arXiv:2310.07896","url":"https://arxiv.org/abs/2310.07896"},{"text":"A. Eftekhar et al.. The One RING: a Robotic Indoor Navigation Generalist. arXiv:2412.14401, 2024.","identifier":"arXiv:2412.14401","url":"https://arxiv.org/abs/2412.14401"},{"text":"J. Zhang et al.. Embodied Navigation Foundation Model. arXiv:2509.12129, 2025.","identifier":"arXiv:2509.12129","url":"https://arxiv.org/abs/2509.12129"},{"text":"J. Zhang, Y. Du, X. Guo, S. Sun, X. Liu, Y. Sun, G. Lu, W. Sui, and J. Li. EA-Nav: Learning Safe Visual Navigation Policies with Embodiment Awareness. arXiv:2607.19880, 2026.","identifier":"arXiv:2607.19880","url":"https://arxiv.org/abs/2607.19880"},{"text":"J. Hu, S. Yuan, G. C. R. Bethala, A. Tzes, and Y. Fang. Learning Adaptive Safety Margins for Visual Navigation. arXiv:2607.18200, 2026.","identifier":"arXiv:2607.18200","url":"https://arxiv.org/abs/2607.18200"},{"text":"D. Fox, W. Burgard, and S. Thrun. The Dynamic Window Approach to Collision Avoidance. IEEE Robotics and Automation Magazine, 4(1), 1997.","identifier":"doi:10.1109/100.580977","url":"https://doi.org/10.1109/100.580977"},{"text":"C. Rösmann, F. Hoffmann, and T. Bertram. Integrated Online Trajectory Planning and Optimization in Distinctive Topologies. Robotics and Autonomous Systems, 88, 2017.","identifier":"doi:10.1016/j.robot.2016.11.007","url":"https://doi.org/10.1016/j.robot.2016.11.007"},{"text":"G. Williams, N. Wagener, B. Goldfain, P. Drews, J. M. Rehg, B. Boots, and E. A. Theodorou. Information Theoretic MPC for Model-Based Reinforcement Learning. IEEE International Conference on Robotics and Automation (ICRA), 2017.","identifier":"doi:10.1109/ICRA.2017.7989202","url":"https://doi.org/10.1109/ICRA.2017.7989202"},{"text":"S. Macenski, F. Martín, R. White, and J. Ginés Clavero. The Marathon 2: A Navigation System. arXiv:2003.00368, 2020.","identifier":"arXiv:2003.00368","url":"https://arxiv.org/abs/2003.00368"},{"text":"S. Macenski, T. Foote, B. Gerkey, C. Lalancette, and W. Woodall. Robot Operating System 2: Design, Architecture, and Uses in the Wild. arXiv:2211.07752, 2022.","identifier":"arXiv:2211.07752","url":"https://arxiv.org/abs/2211.07752"},{"text":"P. Fiorini and Z. Shiller. Motion Planning in Dynamic Environments Using Velocity Obstacles. The International Journal of Robotics Research, 17(7), 1998.","identifier":"doi:10.1177/027836499801700706","url":"https://doi.org/10.1177/027836499801700706"},{"text":"J. van den Berg, M. Lin, and D. Manocha. Reciprocal Velocity Obstacles for Real-Time Multi-Agent Navigation. IEEE International Conference on Robotics and Automation (ICRA), 2008.","identifier":"doi:10.1109/ROBOT.2008.4543489","url":"https://doi.org/10.1109/ROBOT.2008.4543489"},{"text":"J. van den Berg, S. J. Guy, M. Lin, and D. Manocha. Reciprocal n-Body Collision Avoidance. Robotics Research (ISRR), Springer Tracts in Advanced Robotics, vol. 70, 2011.","identifier":"doi:10.1007/978-3-642-19457-3_1","url":"https://doi.org/10.1007/978-3-642-19457-3_1"},{"text":"J. Alonso-Mora, A. Breitenmoser, M. Rufli, P. Beardsley, and R. Siegwart. Optimal Reciprocal Collision Avoidance for Multiple Non-Holonomic Robots. Distributed Autonomous Robotic Systems, Springer Tracts in Advanced Robotics, vol. 83, 2013.","identifier":"doi:10.1007/978-3-642-32723-0_15","url":"https://doi.org/10.1007/978-3-642-32723-0_15"},{"text":"A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada. Control Barrier Functions: Theory and Applications. arXiv:1903.11199, 2019.","identifier":"arXiv:1903.11199","url":"https://arxiv.org/abs/1903.11199"},{"text":"L. Wang, A. D. Ames, and M. Egerstedt. Safety Barrier Certificates for Collisions-Free Multirobot Systems. IEEE Transactions on Robotics, 33(3), 2017.","identifier":"doi:10.1109/TRO.2017.2659727","url":"https://doi.org/10.1109/TRO.2017.2659727"},{"text":"S. Shalev-Shwartz, S. Shammah, and A. Shashua. On a Formal Model of Safe and Scalable Self-driving Cars. arXiv:1708.06374, 2017.","identifier":"arXiv:1708.06374","url":"https://arxiv.org/abs/1708.06374"},{"text":"Y. F. Chen, M. Liu, M. Everett, and J. P. How. Decentralized Non-communicating Multiagent Collision Avoidance with Deep Reinforcement Learning. arXiv:1609.07845, 2016.","identifier":"arXiv:1609.07845","url":"https://arxiv.org/abs/1609.07845"},{"text":"P. Long, T. Fan, X. Liao, W. Liu, H. Zhang, and J. Pan. Towards Optimally Decentralized Multi-Robot Collision Avoidance via Deep Reinforcement Learning. arXiv:1709.10082, 2017.","identifier":"arXiv:1709.10082","url":"https://arxiv.org/abs/1709.10082"},{"text":"M. Everett, Y. F. Chen, and J. P. How. Motion Planning Among Dynamic, Decision-Making Agents with Deep Reinforcement Learning. arXiv:1805.01956, 2018.","identifier":"arXiv:1805.01956","url":"https://arxiv.org/abs/1805.01956"},{"text":"X. Zhou, J. Zhu, H. Zhou, C. Xu, and F. Gao. EGO-Swarm: A Fully Autonomous and Decentralized Quadrotor Swarm System in Cluttered Environments. arXiv:2011.04183, 2020.","identifier":"arXiv:2011.04183","url":"https://arxiv.org/abs/2011.04183"},{"text":"J. Tordesillas and J. P. How. MADER: Trajectory Planner in Multi-Agent and Dynamic Environments. arXiv:2010.04466, 2020.","identifier":"arXiv:2010.04466","url":"https://arxiv.org/abs/2010.04466"},{"text":"C. E. Luis and A. P. Schoellig. Trajectory Generation for Multiagent Point-To-Point Transitions via Distributed Model Predictive Control. arXiv:1809.04230, 2018.","identifier":"arXiv:1809.04230","url":"https://arxiv.org/abs/1809.04230"},{"text":"G. Sharon, R. Stern, A. Felner, and N. R. Sturtevant. Conflict-Based Search for Optimal Multi-Agent Pathfinding. Artificial Intelligence, 219, 2015.","identifier":"doi:10.1016/j.artint.2014.11.006","url":"https://doi.org/10.1016/j.artint.2014.11.006"},{"text":"L. E. Parker. ALLIANCE: An Architecture for Fault Tolerant Multirobot Cooperation. IEEE Transactions on Robotics and Automation, 14(2), 1998.","identifier":"doi:10.1109/70.681242","url":"https://doi.org/10.1109/70.681242"},{"text":"B. P. Gerkey and M. J. Matarić. A Formal Analysis and Taxonomy of Task Allocation in Multi-Robot Systems. The International Journal of Robotics Research, 23(9), 2004.","identifier":"doi:10.1177/0278364904045564","url":"https://doi.org/10.1177/0278364904045564"},{"text":"Y. Rizk, M. Awad, and E. W. Tunstel. Cooperative Heterogeneous Multi-Robot Systems: A Survey. ACM Computing Surveys, 52(2), 2019.","identifier":"doi:10.1145/3303848","url":"https://doi.org/10.1145/3303848"},{"text":"P. Trautman and A. Krause. Unfreezing the Robot: Navigation in Dense, Interacting Crowds. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2010.","identifier":"doi:10.1109/IROS.2010.5654369","url":"https://doi.org/10.1109/IROS.2010.5654369"},{"text":"C. Mavrogiannis, F. Baldini, A. Wang, D. Zhao, P. Trautman, A. Steinfeld, and J. Oh. Core Challenges of Social Robot Navigation: A Survey. arXiv:2103.05668, 2021.","identifier":"arXiv:2103.05668","url":"https://arxiv.org/abs/2103.05668"},{"text":"A. Francis et al.. Principles and Guidelines for Evaluating Social Robot Navigation Algorithms. arXiv:2306.16740, 2023.","identifier":"arXiv:2306.16740","url":"https://arxiv.org/abs/2306.16740"},{"text":"M. Hutter et al.. ANYmal - a Highly Mobile and Dynamic Quadrupedal Robot. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2016.","identifier":"doi:10.1109/IROS.2016.7758092","url":"https://doi.org/10.1109/IROS.2016.7758092"},{"text":"J. Di Carlo, P. M. Wensing, B. Katz, G. Bledt, and S. Kim. Dynamic Locomotion in the MIT Cheetah 3 Through Convex Model-Predictive Control. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2018.","identifier":"doi:10.1109/IROS.2018.8594448","url":"https://doi.org/10.1109/IROS.2018.8594448"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"V. Makoviychuk et al.. Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"A. Kadian et al.. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance? arXiv:1912.06321, 2019.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2609.00005","metadata":{"title":"Shared Configuration Abstractions for Heterogeneous Fleet Navigation","authors":[{"name":"Hana Sato","affiliation":"Advanced Robotics Group, Toyota Research Institute, Palo Alto, CA, USA"},{"name":"Tomas Keller","affiliation":"Advanced Robotics Group, Toyota Research Institute, Palo Alto, CA, USA"},{"name":"Mei-Lin Chao","affiliation":"Advanced Robotics Group, Toyota Research Institute, Palo Alto, CA, USA"}],"publication_date":"2026-09-10","abstract":"Heterogeneous robot fleets that must follow, avoid, or rendezvous with one another need more than per-robot planning: each robot must reason about how its peers are shaped, how quickly they can stop, and how these properties change during operation. Configuration-driven local planners such as AgniNav describe a single robot through an abstract configuration of body geometry, sensing, and velocity limits, but they treat other robots as generic obstacles. This paper extends the configuration descriptor into a fleet-shared abstraction. Each robot publishes a versioned configuration handle on the fleet bus, and the unchanged local planner evaluates interaction-critical derived quantities, namely a predicted clearance envelope and a stop distance, as functions of the peer's live configuration, with conservative inflation when a handle is stale. The approach is evaluated in simulation on a mixed fleet of a Go2-class quadruped, a B2-class quadruped, and a wheeled robot across 12 corridor and intersection scenarios with 5 seeds each. Relative to a configuration-blind planner that models peers as footprint blobs, interaction success increases from 77.7 to 91.6 percent, envelope prediction error decreases from 12.2 cm to 10.8 cm, and stop-distance error decreases from 15.3 cm to 14.3 cm, while the 95th-percentile planner overhead is 3.1 ms compared with 3.5 ms. Ablations indicate that the stop-distance derivation, the clearance envelope, and live configuration updates each account for 9.0 to 10.1 percentage points of interaction success. Recorded failure cases illustrate behavior under stale handles and mid-episode payload changes.","references":[{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Sridhar, D. Shah, C. Glossop, and S. Levine. NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration. arXiv:2310.07896, 2023.","identifier":"arXiv:2310.07896","url":"https://arxiv.org/abs/2310.07896"},{"text":"A. Eftekhar et al.. The One RING: a Robotic Indoor Navigation Generalist. arXiv:2412.14401, 2024.","identifier":"arXiv:2412.14401","url":"https://arxiv.org/abs/2412.14401"},{"text":"J. Zhang et al.. Embodied Navigation Foundation Model. arXiv:2509.12129, 2025.","identifier":"arXiv:2509.12129","url":"https://arxiv.org/abs/2509.12129"},{"text":"J. Zhang, Y. Du, X. Guo, S. Sun, X. Liu, Y. Sun, G. Lu, W. Sui, and J. Li. EA-Nav: Learning Safe Visual Navigation Policies with Embodiment Awareness. arXiv:2607.19880, 2026.","identifier":"arXiv:2607.19880","url":"https://arxiv.org/abs/2607.19880"},{"text":"J. Hu, S. Yuan, G. C. R. Bethala, A. Tzes, and Y. Fang. Learning Adaptive Safety Margins for Visual Navigation. arXiv:2607.18200, 2026.","identifier":"arXiv:2607.18200","url":"https://arxiv.org/abs/2607.18200"},{"text":"D. Fox, W. Burgard, and S. Thrun. The Dynamic Window Approach to Collision Avoidance. IEEE Robotics and Automation Magazine, 4(1), 1997.","identifier":"doi:10.1109/100.580977","url":"https://doi.org/10.1109/100.580977"},{"text":"C. Rösmann, F. Hoffmann, and T. Bertram. Integrated Online Trajectory Planning and Optimization in Distinctive Topologies. Robotics and Autonomous Systems, 88, 2017.","identifier":"doi:10.1016/j.robot.2016.11.007","url":"https://doi.org/10.1016/j.robot.2016.11.007"},{"text":"G. Williams, N. Wagener, B. Goldfain, P. Drews, J. M. Rehg, B. Boots, and E. A. Theodorou. Information Theoretic MPC for Model-Based Reinforcement Learning. IEEE International Conference on Robotics and Automation (ICRA), 2017.","identifier":"doi:10.1109/ICRA.2017.7989202","url":"https://doi.org/10.1109/ICRA.2017.7989202"},{"text":"S. Macenski, F. Martín, R. White, and J. Ginés Clavero. The Marathon 2: A Navigation System. arXiv:2003.00368, 2020.","identifier":"arXiv:2003.00368","url":"https://arxiv.org/abs/2003.00368"},{"text":"S. Macenski, T. Foote, B. Gerkey, C. Lalancette, and W. Woodall. Robot Operating System 2: Design, Architecture, and Uses in the Wild. arXiv:2211.07752, 2022.","identifier":"arXiv:2211.07752","url":"https://arxiv.org/abs/2211.07752"},{"text":"P. Fiorini and Z. Shiller. Motion Planning in Dynamic Environments Using Velocity Obstacles. The International Journal of Robotics Research, 17(7), 1998.","identifier":"doi:10.1177/027836499801700706","url":"https://doi.org/10.1177/027836499801700706"},{"text":"J. van den Berg, M. Lin, and D. Manocha. Reciprocal Velocity Obstacles for Real-Time Multi-Agent Navigation. IEEE International Conference on Robotics and Automation (ICRA), 2008.","identifier":"doi:10.1109/ROBOT.2008.4543489","url":"https://doi.org/10.1109/ROBOT.2008.4543489"},{"text":"J. van den Berg, S. J. Guy, M. Lin, and D. Manocha. Reciprocal n-Body Collision Avoidance. Robotics Research (ISRR), Springer Tracts in Advanced Robotics, vol. 70, 2011.","identifier":"doi:10.1007/978-3-642-19457-3_1","url":"https://doi.org/10.1007/978-3-642-19457-3_1"},{"text":"J. Alonso-Mora, A. Breitenmoser, M. Rufli, P. Beardsley, and R. Siegwart. Optimal Reciprocal Collision Avoidance for Multiple Non-Holonomic Robots. Distributed Autonomous Robotic Systems, Springer Tracts in Advanced Robotics, vol. 83, 2013.","identifier":"doi:10.1007/978-3-642-32723-0_15","url":"https://doi.org/10.1007/978-3-642-32723-0_15"},{"text":"A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada. Control Barrier Functions: Theory and Applications. arXiv:1903.11199, 2019.","identifier":"arXiv:1903.11199","url":"https://arxiv.org/abs/1903.11199"},{"text":"L. Wang, A. D. Ames, and M. Egerstedt. Safety Barrier Certificates for Collisions-Free Multirobot Systems. IEEE Transactions on Robotics, 33(3), 2017.","identifier":"doi:10.1109/TRO.2017.2659727","url":"https://doi.org/10.1109/TRO.2017.2659727"},{"text":"S. Shalev-Shwartz, S. Shammah, and A. Shashua. On a Formal Model of Safe and Scalable Self-driving Cars. arXiv:1708.06374, 2017.","identifier":"arXiv:1708.06374","url":"https://arxiv.org/abs/1708.06374"},{"text":"Y. F. Chen, M. Liu, M. Everett, and J. P. How. Decentralized Non-communicating Multiagent Collision Avoidance with Deep Reinforcement Learning. arXiv:1609.07845, 2016.","identifier":"arXiv:1609.07845","url":"https://arxiv.org/abs/1609.07845"},{"text":"P. Long, T. Fan, X. Liao, W. Liu, H. Zhang, and J. Pan. Towards Optimally Decentralized Multi-Robot Collision Avoidance via Deep Reinforcement Learning. arXiv:1709.10082, 2017.","identifier":"arXiv:1709.10082","url":"https://arxiv.org/abs/1709.10082"},{"text":"M. Everett, Y. F. Chen, and J. P. How. Motion Planning Among Dynamic, Decision-Making Agents with Deep Reinforcement Learning. arXiv:1805.01956, 2018.","identifier":"arXiv:1805.01956","url":"https://arxiv.org/abs/1805.01956"},{"text":"X. Zhou, J. Zhu, H. Zhou, C. Xu, and F. Gao. EGO-Swarm: A Fully Autonomous and Decentralized Quadrotor Swarm System in Cluttered Environments. arXiv:2011.04183, 2020.","identifier":"arXiv:2011.04183","url":"https://arxiv.org/abs/2011.04183"},{"text":"J. Tordesillas and J. P. How. MADER: Trajectory Planner in Multi-Agent and Dynamic Environments. arXiv:2010.04466, 2020.","identifier":"arXiv:2010.04466","url":"https://arxiv.org/abs/2010.04466"},{"text":"C. E. Luis and A. P. Schoellig. Trajectory Generation for Multiagent Point-To-Point Transitions via Distributed Model Predictive Control. arXiv:1809.04230, 2018.","identifier":"arXiv:1809.04230","url":"https://arxiv.org/abs/1809.04230"},{"text":"G. Sharon, R. Stern, A. Felner, and N. R. Sturtevant. Conflict-Based Search for Optimal Multi-Agent Pathfinding. Artificial Intelligence, 219, 2015.","identifier":"doi:10.1016/j.artint.2014.11.006","url":"https://doi.org/10.1016/j.artint.2014.11.006"},{"text":"L. E. Parker. ALLIANCE: An Architecture for Fault Tolerant Multirobot Cooperation. IEEE Transactions on Robotics and Automation, 14(2), 1998.","identifier":"doi:10.1109/70.681242","url":"https://doi.org/10.1109/70.681242"},{"text":"B. P. Gerkey and M. J. Matarić. A Formal Analysis and Taxonomy of Task Allocation in Multi-Robot Systems. The International Journal of Robotics Research, 23(9), 2004.","identifier":"doi:10.1177/0278364904045564","url":"https://doi.org/10.1177/0278364904045564"},{"text":"Y. Rizk, M. Awad, and E. W. Tunstel. Cooperative Heterogeneous Multi-Robot Systems: A Survey. ACM Computing Surveys, 52(2), 2019.","identifier":"doi:10.1145/3303848","url":"https://doi.org/10.1145/3303848"},{"text":"P. Trautman and A. Krause. Unfreezing the Robot: Navigation in Dense, Interacting Crowds. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2010.","identifier":"doi:10.1109/IROS.2010.5654369","url":"https://doi.org/10.1109/IROS.2010.5654369"},{"text":"C. Mavrogiannis, F. Baldini, A. Wang, D. Zhao, P. Trautman, A. Steinfeld, and J. Oh. Core Challenges of Social Robot Navigation: A Survey. arXiv:2103.05668, 2021.","identifier":"arXiv:2103.05668","url":"https://arxiv.org/abs/2103.05668"},{"text":"A. Francis et al.. Principles and Guidelines for Evaluating Social Robot Navigation Algorithms. arXiv:2306.16740, 2023.","identifier":"arXiv:2306.16740","url":"https://arxiv.org/abs/2306.16740"},{"text":"M. Hutter et al.. ANYmal - a Highly Mobile and Dynamic Quadrupedal Robot. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2016.","identifier":"doi:10.1109/IROS.2016.7758092","url":"https://doi.org/10.1109/IROS.2016.7758092"},{"text":"J. Di Carlo, P. M. Wensing, B. Katz, G. Bledt, and S. Kim. Dynamic Locomotion in the MIT Cheetah 3 Through Convex Model-Predictive Control. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2018.","identifier":"doi:10.1109/IROS.2018.8594448","url":"https://doi.org/10.1109/IROS.2018.8594448"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"V. Makoviychuk et al.. Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"A. Kadian et al.. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance? arXiv:1912.06321, 2019.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2609.00005","slug":"2609.00005","registered_at":"2026-09-17T18:00:00Z","updated_at":"2026-09-17T19:09:00Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2609.00005","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2609.00005","tex_path":"articles/2609.00005/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-09-17T20:17:00Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2609.00005"},{"kagg_id":"KAGG:2609.00004","slug":"2609.00004","title":"High-Speed Obstacle Negotiation for Agile Quadrupeds with Range-Predictive Control","authors":[{"name":"Carlos Ramirez","affiliation":"Legged Robotics Lab, Stanford University, Stanford, CA, USA"},{"name":"Felix Wagner","affiliation":"Legged Robotics Lab, Stanford University, Stanford, CA, USA"},{"name":"Ethan Brooks","affiliation":"Legged Robotics Lab, Stanford University, Stanford, CA, USA"},{"name":"Ananya Iyer","affiliation":"Legged Robotics Lab, Stanford University, Stanford, CA, USA"}],"affiliation":null,"abstract":"At running speeds of 4 to 6 m/s, the reaction budget of a quadruped robot is dominated by the lag of its velocity controller rather than by perception, so a controller that reacts to obstacles already present in its range measurements acts too late. This paper presents a range-predictive negotiation controller built on FutureRay, a perception module that predicts future free range along control-aligned directions. The controller uses the predicted range field to pre-shape the forward and yaw velocity commands before a geometric conflict materializes, schedules the pre-shaping lead time from an identified lag model, and applies an explicit hysteretic rule that hands control to a braking policy when the predicted range collapses below the braking envelope. The design space was explored in Isaac Lab with five seeds, and the controller was evaluated on a Unitree Go2 in a controlled course with six obstacle classes, approach speeds from 4 to 6 m/s in 0.5 m/s steps, and 20 trials per class and speed. Compared with a present-range reactive baseline, the controller raises the mean maximum safe speed over obstacle classes from 4.72 to 4.90 m/s, lowers the miss rate from 4.2 to 3.8 percent, and raises braking handoff success from 90.2 to 97.2 percent. Ablations indicate that the handoff rule, the predicted range, and lead-time scheduling each contribute. Failure analysis shows that at 6.0 m/s the occluded gap class is limited by the unobserved range behind the occluder rather than by the controller.","publication_date":"2026-09-08","registered_at":"2026-09-13T12:52:00Z","updated_at":"2026-09-13T18:44:00Z","status":"published","html_path":null,"pdf_path":"articles/2609.00004/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"T. Zang, S. Wang, H. Huang, and S. Cheng. FutureRay: Control-Aligned Future Range for Agile Quadruped Navigation. arXiv:2608.01197, 2026.","identifier":"arXiv:2608.01197","url":"https://arxiv.org/abs/2608.01197"},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"G. Ji, J. Mun, H. Kim, and J. Hwangbo. Concurrent Training of a Control Policy and a State Estimator for Dynamic and Robust Legged Locomotion. arXiv:2202.05481, 2022.","identifier":"arXiv:2202.05481","url":"https://arxiv.org/abs/2202.05481"},{"text":"G. B. Margolis, G. Yang, K. Paigwar, T. Chen, and P. Agrawal. Rapid Locomotion via Reinforcement Learning. arXiv:2205.02824, 2022.","identifier":"arXiv:2205.02824","url":"https://arxiv.org/abs/2205.02824"},{"text":"G. B. Margolis and P. Agrawal. Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior. arXiv:2212.03238, 2022.","identifier":"arXiv:2212.03238","url":"https://arxiv.org/abs/2212.03238"},{"text":"I. M. A. Nahrendra, B. Yu, and H. Myung. DreamWaQ: Learning Robust Quadrupedal Locomotion with Implicit Terrain Imagination via Deep Reinforcement Learning. arXiv:2301.10602, 2023.","identifier":"arXiv:2301.10602","url":"https://arxiv.org/abs/2301.10602"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. arXiv:2211.07638, 2022.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"R. Yang, G. Yang, and X. Wang. Neural Volumetric Memory for Visual Locomotion Control. arXiv:2304.01201, 2023.","identifier":"arXiv:2304.01201","url":"https://arxiv.org/abs/2304.01201"},{"text":"D. Hoeller, N. Rudin, C. Choy, A. Anandkumar, and M. Hutter. Neural Scene Representation for Locomotion on Structured Terrain. arXiv:2206.08077, 2022.","identifier":"arXiv:2206.08077","url":"https://arxiv.org/abs/2206.08077"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. arXiv:2309.05665, 2023.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. arXiv:2309.14341, 2023.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"S. Ha, J. Lee, M. van de Panne, Z. Xie, W. Yu, and M. Khadiv. Learning-based Legged Locomotion: State of the Art and Future Perspectives. arXiv:2406.01152, 2024.","identifier":"arXiv:2406.01152","url":"https://arxiv.org/abs/2406.01152"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, L. Wellhausen, F. Farshidian, and M. Hutter. Learning a State Representation and Navigation in Cluttered and Dynamic Environments. arXiv:2103.04351, 2021.","identifier":"arXiv:2103.04351","url":"https://arxiv.org/abs/2103.04351"},{"text":"J. Truong, D. Yarats, T. Li, F. Meier, S. Chernova, D. Batra, and A. Rai. Learning Navigation Skills for Legged Robots with Learned Robot Embeddings. arXiv:2011.12255, 2020.","identifier":"arXiv:2011.12255","url":"https://arxiv.org/abs/2011.12255"},{"text":"T. He, C. Zhang, W. Xiao, G. He, C. Liu, and G. Shi. Agile But Safe: Learning Collision-Free High-Speed Legged Locomotion. arXiv:2401.17583, 2024.","identifier":"arXiv:2401.17583","url":"https://arxiv.org/abs/2401.17583"},{"text":"Y. Kim, H. Oh, J. Lee, J. Choi, G. Ji, M. Jung, D. Youm, and J. Hwangbo. Not Only Rewards But Also Constraints: Applications on Legged Robot Locomotion. arXiv:2308.12517, 2023.","identifier":"arXiv:2308.12517","url":"https://arxiv.org/abs/2308.12517"},{"text":"M. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. arXiv:2305.08510, 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"G. Kahn, P. Abbeel, and S. Levine. BADGR: An Autonomous Self-Supervised Learning-Based Navigation System. arXiv:2002.05700, 2020.","identifier":"arXiv:2002.05700","url":"https://arxiv.org/abs/2002.05700"},{"text":"D. Fox, W. Burgard, and S. Thrun. The Dynamic Window Approach to Collision Avoidance. IEEE Robotics and Automation Magazine, 4(1):23–33, 1997.","identifier":"doi:10.1109/100.580977","url":"https://doi.org/10.1109/100.580977"},{"text":"P. Fiorini and Z. Shiller. Motion Planning in Dynamic Environments Using Velocity Obstacles. International Journal of Robotics Research, 17(7):760–772, 1998.","identifier":"doi:10.1177/027836499801700706","url":"https://doi.org/10.1177/027836499801700706"},{"text":"A. D. Ames, X. Xu, J. W. Grizzle, and P. Tabuada. Control Barrier Function Based Quadratic Programs for Safety Critical Systems. arXiv:1609.06408, 2016.","identifier":"arXiv:1609.06408","url":"https://arxiv.org/abs/1609.06408"},{"text":"A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada. Control Barrier Functions: Theory and Applications. arXiv:1903.11199, 2019.","identifier":"arXiv:1903.11199","url":"https://arxiv.org/abs/1903.11199"},{"text":"K.-C. Hsu, H. Hu, and J. F. Fisac. The Safety Filter: A Unified View of Safety-Critical Control in Autonomous Systems. arXiv:2309.05837, 2023.","identifier":"arXiv:2309.05837","url":"https://arxiv.org/abs/2309.05837"},{"text":"S. Bansal, M. Chen, S. Herbert, and C. J. Tomlin. Hamilton-Jacobi Reachability: A Brief Overview and Recent Advances. arXiv:1709.07523, 2017.","identifier":"arXiv:1709.07523","url":"https://arxiv.org/abs/1709.07523"},{"text":"A. Bajcsy, S. Bansal, E. Bronstein, V. Tolani, and C. J. Tomlin. An Efficient Reachability-Based Framework for Provably Safe Autonomous Navigation in Unknown Environments. arXiv:1905.00532, 2019.","identifier":"arXiv:1905.00532","url":"https://arxiv.org/abs/1905.00532"},{"text":"K. P. Wabersich and M. N. Zeilinger. A Predictive Safety Filter for Learning-Based Control of Constrained Nonlinear Dynamical Systems. arXiv:1812.05506, 2018.","identifier":"arXiv:1812.05506","url":"https://arxiv.org/abs/1812.05506"},{"text":"S. Kousik, S. Vaskov, F. Bu, M. Johnson-Roberson, and R. Vasudevan. Bridging the Gap Between Safety and Real-Time Performance in Receding-Horizon Trajectory Design for Mobile Robots. arXiv:1809.06746, 2018.","identifier":"arXiv:1809.06746","url":"https://arxiv.org/abs/1809.06746"},{"text":"J. Tordesillas, B. T. Lopez, M. Everett, and J. P. How. FASTER: Fast and Safe Trajectory Planner for Navigation in Unknown Environments. arXiv:2001.04420, 2020.","identifier":"arXiv:2001.04420","url":"https://arxiv.org/abs/2001.04420"},{"text":"X. Zhou, Z. Wang, H. Ye, C. Xu, and F. Gao. EGO-Planner: An ESDF-free Gradient-based Local Planner for Quadrotors. arXiv:2008.08835, 2020.","identifier":"arXiv:2008.08835","url":"https://arxiv.org/abs/2008.08835"},{"text":"A. Loquercio, E. Kaufmann, R. Ranftl, M. Müller, V. Koltun, and D. Scaramuzza. Learning High-Speed Flight in the Wild. arXiv:2110.05113, 2021.","identifier":"arXiv:2110.05113","url":"https://arxiv.org/abs/2110.05113"},{"text":"E. Kaufmann, L. Bauersfeld, A. Loquercio, M. Müller, V. Koltun, and D. Scaramuzza. Champion-Level Drone Racing Using Deep Reinforcement Learning. Nature, 620:982–987, 2023.","identifier":"doi:10.1038/s41586-023-06419-4","url":"https://doi.org/10.1038/s41586-023-06419-4"},{"text":"Y. Song, A. Romero, M. Müller, V. Koltun, and D. Scaramuzza. Reaching the Limit in Autonomous Racing: Optimal Control versus Reinforcement Learning. arXiv:2310.10943, 2023.","identifier":"arXiv:2310.10943","url":"https://arxiv.org/abs/2310.10943"},{"text":"J. Di Carlo, P. M. Wensing, B. Katz, G. Bledt, and S. Kim. Dynamic Locomotion in the MIT Cheetah 3 Through Convex Model-Predictive Control. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2018.","identifier":"doi:10.1109/IROS.2018.8594448","url":"https://doi.org/10.1109/IROS.2018.8594448"},{"text":"R. Grandia, F. Jenelten, S. Yang, F. Farshidian, and M. Hutter. Perceptive Locomotion through Nonlinear Model Predictive Control. arXiv:2208.08373, 2022.","identifier":"arXiv:2208.08373","url":"https://arxiv.org/abs/2208.08373"},{"text":"V. Makoviychuk et al.. Isaac Gym: High Performance GPU-Based Physics Simulation for Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal et al.. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. arXiv:2301.04195, 2023.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"J. Schulman, F. Wolski, P. Dhariwal, A. Radford, and O. Klimov. Proximal Policy Optimization Algorithms. arXiv:1707.06347, 2017.","identifier":"arXiv:1707.06347","url":"https://arxiv.org/abs/1707.06347"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. arXiv:1703.06907, 2017.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2609.00004","metadata":{"title":"High-Speed Obstacle Negotiation for Agile Quadrupeds with Range-Predictive Control","authors":[{"name":"Carlos Ramirez","affiliation":"Legged Robotics Lab, Stanford University, Stanford, CA, USA"},{"name":"Felix Wagner","affiliation":"Legged Robotics Lab, Stanford University, Stanford, CA, USA"},{"name":"Ethan Brooks","affiliation":"Legged Robotics Lab, Stanford University, Stanford, CA, USA"},{"name":"Ananya Iyer","affiliation":"Legged Robotics Lab, Stanford University, Stanford, CA, USA"}],"publication_date":"2026-09-08","abstract":"At running speeds of 4 to 6 m/s, the reaction budget of a quadruped robot is dominated by the lag of its velocity controller rather than by perception, so a controller that reacts to obstacles already present in its range measurements acts too late. This paper presents a range-predictive negotiation controller built on FutureRay, a perception module that predicts future free range along control-aligned directions. The controller uses the predicted range field to pre-shape the forward and yaw velocity commands before a geometric conflict materializes, schedules the pre-shaping lead time from an identified lag model, and applies an explicit hysteretic rule that hands control to a braking policy when the predicted range collapses below the braking envelope. The design space was explored in Isaac Lab with five seeds, and the controller was evaluated on a Unitree Go2 in a controlled course with six obstacle classes, approach speeds from 4 to 6 m/s in 0.5 m/s steps, and 20 trials per class and speed. Compared with a present-range reactive baseline, the controller raises the mean maximum safe speed over obstacle classes from 4.72 to 4.90 m/s, lowers the miss rate from 4.2 to 3.8 percent, and raises braking handoff success from 90.2 to 97.2 percent. Ablations indicate that the handoff rule, the predicted range, and lead-time scheduling each contribute. Failure analysis shows that at 6.0 m/s the occluded gap class is limited by the unobserved range behind the occluder rather than by the controller.","references":[{"text":"T. Zang, S. Wang, H. Huang, and S. Cheng. FutureRay: Control-Aligned Future Range for Agile Quadruped Navigation. arXiv:2608.01197, 2026.","identifier":"arXiv:2608.01197","url":"https://arxiv.org/abs/2608.01197"},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"G. Ji, J. Mun, H. Kim, and J. Hwangbo. Concurrent Training of a Control Policy and a State Estimator for Dynamic and Robust Legged Locomotion. arXiv:2202.05481, 2022.","identifier":"arXiv:2202.05481","url":"https://arxiv.org/abs/2202.05481"},{"text":"G. B. Margolis, G. Yang, K. Paigwar, T. Chen, and P. Agrawal. Rapid Locomotion via Reinforcement Learning. arXiv:2205.02824, 2022.","identifier":"arXiv:2205.02824","url":"https://arxiv.org/abs/2205.02824"},{"text":"G. B. Margolis and P. Agrawal. Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior. arXiv:2212.03238, 2022.","identifier":"arXiv:2212.03238","url":"https://arxiv.org/abs/2212.03238"},{"text":"I. M. A. Nahrendra, B. Yu, and H. Myung. DreamWaQ: Learning Robust Quadrupedal Locomotion with Implicit Terrain Imagination via Deep Reinforcement Learning. arXiv:2301.10602, 2023.","identifier":"arXiv:2301.10602","url":"https://arxiv.org/abs/2301.10602"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. arXiv:2211.07638, 2022.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"R. Yang, G. Yang, and X. Wang. Neural Volumetric Memory for Visual Locomotion Control. arXiv:2304.01201, 2023.","identifier":"arXiv:2304.01201","url":"https://arxiv.org/abs/2304.01201"},{"text":"D. Hoeller, N. Rudin, C. Choy, A. Anandkumar, and M. Hutter. Neural Scene Representation for Locomotion on Structured Terrain. arXiv:2206.08077, 2022.","identifier":"arXiv:2206.08077","url":"https://arxiv.org/abs/2206.08077"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. arXiv:2309.05665, 2023.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. arXiv:2309.14341, 2023.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"S. Ha, J. Lee, M. van de Panne, Z. Xie, W. Yu, and M. Khadiv. Learning-based Legged Locomotion: State of the Art and Future Perspectives. arXiv:2406.01152, 2024.","identifier":"arXiv:2406.01152","url":"https://arxiv.org/abs/2406.01152"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, L. Wellhausen, F. Farshidian, and M. Hutter. Learning a State Representation and Navigation in Cluttered and Dynamic Environments. arXiv:2103.04351, 2021.","identifier":"arXiv:2103.04351","url":"https://arxiv.org/abs/2103.04351"},{"text":"J. Truong, D. Yarats, T. Li, F. Meier, S. Chernova, D. Batra, and A. Rai. Learning Navigation Skills for Legged Robots with Learned Robot Embeddings. arXiv:2011.12255, 2020.","identifier":"arXiv:2011.12255","url":"https://arxiv.org/abs/2011.12255"},{"text":"T. He, C. Zhang, W. Xiao, G. He, C. Liu, and G. Shi. Agile But Safe: Learning Collision-Free High-Speed Legged Locomotion. arXiv:2401.17583, 2024.","identifier":"arXiv:2401.17583","url":"https://arxiv.org/abs/2401.17583"},{"text":"Y. Kim, H. Oh, J. Lee, J. Choi, G. Ji, M. Jung, D. Youm, and J. Hwangbo. Not Only Rewards But Also Constraints: Applications on Legged Robot Locomotion. arXiv:2308.12517, 2023.","identifier":"arXiv:2308.12517","url":"https://arxiv.org/abs/2308.12517"},{"text":"M. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. arXiv:2305.08510, 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"G. Kahn, P. Abbeel, and S. Levine. BADGR: An Autonomous Self-Supervised Learning-Based Navigation System. arXiv:2002.05700, 2020.","identifier":"arXiv:2002.05700","url":"https://arxiv.org/abs/2002.05700"},{"text":"D. Fox, W. Burgard, and S. Thrun. The Dynamic Window Approach to Collision Avoidance. IEEE Robotics and Automation Magazine, 4(1):23–33, 1997.","identifier":"doi:10.1109/100.580977","url":"https://doi.org/10.1109/100.580977"},{"text":"P. Fiorini and Z. Shiller. Motion Planning in Dynamic Environments Using Velocity Obstacles. International Journal of Robotics Research, 17(7):760–772, 1998.","identifier":"doi:10.1177/027836499801700706","url":"https://doi.org/10.1177/027836499801700706"},{"text":"A. D. Ames, X. Xu, J. W. Grizzle, and P. Tabuada. Control Barrier Function Based Quadratic Programs for Safety Critical Systems. arXiv:1609.06408, 2016.","identifier":"arXiv:1609.06408","url":"https://arxiv.org/abs/1609.06408"},{"text":"A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada. Control Barrier Functions: Theory and Applications. arXiv:1903.11199, 2019.","identifier":"arXiv:1903.11199","url":"https://arxiv.org/abs/1903.11199"},{"text":"K.-C. Hsu, H. Hu, and J. F. Fisac. The Safety Filter: A Unified View of Safety-Critical Control in Autonomous Systems. arXiv:2309.05837, 2023.","identifier":"arXiv:2309.05837","url":"https://arxiv.org/abs/2309.05837"},{"text":"S. Bansal, M. Chen, S. Herbert, and C. J. Tomlin. Hamilton-Jacobi Reachability: A Brief Overview and Recent Advances. arXiv:1709.07523, 2017.","identifier":"arXiv:1709.07523","url":"https://arxiv.org/abs/1709.07523"},{"text":"A. Bajcsy, S. Bansal, E. Bronstein, V. Tolani, and C. J. Tomlin. An Efficient Reachability-Based Framework for Provably Safe Autonomous Navigation in Unknown Environments. arXiv:1905.00532, 2019.","identifier":"arXiv:1905.00532","url":"https://arxiv.org/abs/1905.00532"},{"text":"K. P. Wabersich and M. N. Zeilinger. A Predictive Safety Filter for Learning-Based Control of Constrained Nonlinear Dynamical Systems. arXiv:1812.05506, 2018.","identifier":"arXiv:1812.05506","url":"https://arxiv.org/abs/1812.05506"},{"text":"S. Kousik, S. Vaskov, F. Bu, M. Johnson-Roberson, and R. Vasudevan. Bridging the Gap Between Safety and Real-Time Performance in Receding-Horizon Trajectory Design for Mobile Robots. arXiv:1809.06746, 2018.","identifier":"arXiv:1809.06746","url":"https://arxiv.org/abs/1809.06746"},{"text":"J. Tordesillas, B. T. Lopez, M. Everett, and J. P. How. FASTER: Fast and Safe Trajectory Planner for Navigation in Unknown Environments. arXiv:2001.04420, 2020.","identifier":"arXiv:2001.04420","url":"https://arxiv.org/abs/2001.04420"},{"text":"X. Zhou, Z. Wang, H. Ye, C. Xu, and F. Gao. EGO-Planner: An ESDF-free Gradient-based Local Planner for Quadrotors. arXiv:2008.08835, 2020.","identifier":"arXiv:2008.08835","url":"https://arxiv.org/abs/2008.08835"},{"text":"A. Loquercio, E. Kaufmann, R. Ranftl, M. Müller, V. Koltun, and D. Scaramuzza. Learning High-Speed Flight in the Wild. arXiv:2110.05113, 2021.","identifier":"arXiv:2110.05113","url":"https://arxiv.org/abs/2110.05113"},{"text":"E. Kaufmann, L. Bauersfeld, A. Loquercio, M. Müller, V. Koltun, and D. Scaramuzza. Champion-Level Drone Racing Using Deep Reinforcement Learning. Nature, 620:982–987, 2023.","identifier":"doi:10.1038/s41586-023-06419-4","url":"https://doi.org/10.1038/s41586-023-06419-4"},{"text":"Y. Song, A. Romero, M. Müller, V. Koltun, and D. Scaramuzza. Reaching the Limit in Autonomous Racing: Optimal Control versus Reinforcement Learning. arXiv:2310.10943, 2023.","identifier":"arXiv:2310.10943","url":"https://arxiv.org/abs/2310.10943"},{"text":"J. Di Carlo, P. M. Wensing, B. Katz, G. Bledt, and S. Kim. Dynamic Locomotion in the MIT Cheetah 3 Through Convex Model-Predictive Control. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2018.","identifier":"doi:10.1109/IROS.2018.8594448","url":"https://doi.org/10.1109/IROS.2018.8594448"},{"text":"R. Grandia, F. Jenelten, S. Yang, F. Farshidian, and M. Hutter. Perceptive Locomotion through Nonlinear Model Predictive Control. arXiv:2208.08373, 2022.","identifier":"arXiv:2208.08373","url":"https://arxiv.org/abs/2208.08373"},{"text":"V. Makoviychuk et al.. Isaac Gym: High Performance GPU-Based Physics Simulation for Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal et al.. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. arXiv:2301.04195, 2023.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"J. Schulman, F. Wolski, P. Dhariwal, A. Radford, and O. Klimov. Proximal Policy Optimization Algorithms. arXiv:1707.06347, 2017.","identifier":"arXiv:1707.06347","url":"https://arxiv.org/abs/1707.06347"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. arXiv:1703.06907, 2017.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2609.00004","slug":"2609.00004","registered_at":"2026-09-13T12:52:00Z","updated_at":"2026-09-13T18:44:00Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2609.00004","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2609.00004","tex_path":"articles/2609.00004/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-09-13T15:09:00Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2609.00004"},{"kagg_id":"KAGG:2609.00003","slug":"2609.00003","title":"Closed-Loop Field Calibration of Quadruped Velocity Interfaces","authors":[{"name":"Wei Xu","affiliation":"Autonomous Robotics Center, Carnegie Mellon University, Pittsburgh, PA, USA"},{"name":"Arjun Mehta","affiliation":"Autonomous Robotics Center, Carnegie Mellon University, Pittsburgh, PA, USA"},{"name":"Ethan Cole","affiliation":"Autonomous Robotics Center, Carnegie Mellon University, Pittsburgh, PA, USA"}],"affiliation":null,"abstract":"Quadruped navigation stacks usually command a locomotion controller through a velocity interface whose command-to-motion response must be calibrated before the planner's predictions can be trusted. Open-loop active calibration, such as planner-conditioned probing with GAUGE, yields an accurate model at calibration time, but the response drifts as payload, battery state, temperature, and wear change during deployment. This paper closes the calibration loop. The calibrated interface model is embedded in the local planner, and the prediction errors it incurs during normal operation are treated as passive observations that update a Gaussian posterior over the interface parameters. Active probes are issued only when the planner-relevant predictive uncertainty crosses a threshold, and a residual-decomposition damping rule separates genuine parameter change from structural model mismatch so that the estimator does not chatter when the operating point leaves the model class. The analysis gives conditions under which the posterior covariance and the parameter error remain bounded under bounded mismatch. The estimator was evaluated on an ANYmal B during a three-day field campaign with 18 h of operation, payloads of 0, 6, and 12 kg, battery levels from 100 to 30 percent, and 2.1 M observations logged at 50 Hz. At the same probe frequency of 0.4 probes per hour, navigation success was 90.1 percent, compared with 84.1 percent for recalibration every 4 h and 80.5 percent for a static day-0 model, while the improvement in interface drift was small (2.9 versus 3.0 and 3.4 percent per day).","publication_date":"2026-09-05","registered_at":"2026-09-09T16:14:00Z","updated_at":"2026-09-10T17:09:00Z","status":"published","html_path":null,"pdf_path":"articles/2609.00003/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"T. Zang, H. Huang, S. Cheng, and W. Zhang. GAUGE: Planner-Conditioned Active Calibration of Opaque Quadruped Velocity Interfaces. arXiv:2607.02841, 2026.","identifier":"arXiv:2607.02841","url":"https://arxiv.org/abs/2607.02841"},{"text":"M. Hutter et al.. ANYmal - A Highly Mobile and Dynamic Quadrupedal Robot. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2016.","identifier":"doi:10.1109/IROS.2016.7758092","url":"https://doi.org/10.1109/IROS.2016.7758092"},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. Science Robotics, 4(26), 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. Science Robotics, 5(47), 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. Science Robotics, 7(62), 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. Conference on Robot Learning (CoRL), 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"G. B. Margolis, G. Yang, K. Paigwar, T. Chen, and P. Agrawal. Rapid Locomotion via Reinforcement Learning. Robotics: Science and Systems (RSS), 2022.","identifier":"arXiv:2205.02824","url":"https://arxiv.org/abs/2205.02824"},{"text":"G. B. Margolis and P. Agrawal. Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior. Conference on Robot Learning (CoRL), 2022.","identifier":"arXiv:2212.03238","url":"https://arxiv.org/abs/2212.03238"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. Science Robotics, 9(88), 2024.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"S. Ha, J. Lee, M. van de Panne, Z. Xie, W. Yu, and M. Khadiv. Learning-Based Legged Locomotion: State of the Art and Future Perspectives. The International Journal of Robotics Research, 2025.","identifier":"arXiv:2406.01152","url":"https://arxiv.org/abs/2406.01152"},{"text":"N. Seegmiller, F. Rogers-Marcovitz, G. Miller, and A. Kelly. Vehicle Model Identification by Integrated Prediction Error Minimization. The International Journal of Robotics Research, 32(8), 2013.","identifier":"","url":""},{"text":"C. J. Ostafew, A. P. Schoellig, and T. D. Barfoot. Robust Constrained Learning-Based NMPC Enabling Reliable Mobile Robot Path Tracking. The International Journal of Robotics Research, 35(13), 2016.","identifier":"","url":""},{"text":"L. Ljung. System Identification: Theory for the User, 2nd ed.. Prentice Hall, 1999.","identifier":"","url":""},{"text":"M. Gevers. Identification for Control: From the Early Achievements to the Revival of Experiment Design. European Journal of Control, 11(4-5), 2005.","identifier":"","url":""},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. Robotics: Science and Systems (RSS), 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. IEEE International Conference on Robotics and Automation (ICRA), 2018.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"},{"text":"J. Tan et al.. Sim-to-Real: Learning Agile Locomotion for Quadruped Robots. Robotics: Science and Systems (RSS), 2018.","identifier":"arXiv:1804.10332","url":"https://arxiv.org/abs/1804.10332"},{"text":"A. Nagabandi et al.. Learning to Adapt in Dynamic, Real-World Environments through Meta-Reinforcement Learning. International Conference on Learning Representations (ICLR), 2019.","identifier":"arXiv:1803.11347","url":"https://arxiv.org/abs/1803.11347"},{"text":"M. Sombolestan, Y. Chen, and Q. Nguyen. Adaptive Force-Based Control for Legged Robots. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2021.","identifier":"arXiv:2011.06236","url":"https://arxiv.org/abs/2011.06236"},{"text":"R. E. Kalman. A New Approach to Linear Filtering and Prediction Problems. Transactions of the ASME, Journal of Basic Engineering, 82(1), 1960.","identifier":"doi:10.1115/1.3662552","url":"https://doi.org/10.1115/1.3662552"},{"text":"R. K. Mehra. On the Identification of Variances and Adaptive Kalman Filtering. IEEE Transactions on Automatic Control, 15(2), 1970.","identifier":"doi:10.1109/TAC.1970.1099422","url":"https://doi.org/10.1109/TAC.1970.1099422"},{"text":"K. Reif, S. Günther, E. Yaz, and R. Unbehauen. Stochastic Stability of the Discrete-Time Extended Kalman Filter. IEEE Transactions on Automatic Control, 44(4), 1999.","identifier":"doi:10.1109/9.754809","url":"https://doi.org/10.1109/9.754809"},{"text":"Y. Bar-Shalom, X. R. Li, and T. Kirubarajan. Estimation with Applications to Tracking and Navigation. Wiley, 2001.","identifier":"","url":""},{"text":"P. J. Huber. Robust Estimation of a Location Parameter. The Annals of Mathematical Statistics, 35(1), 1964.","identifier":"doi:10.1214/aoms/1177703732","url":"https://doi.org/10.1214/aoms/1177703732"},{"text":"R. Kulhavý and M. B. Zarrop. On a General Concept of Forgetting. International Journal of Control, 58(4), 1993.","identifier":"","url":""},{"text":"P. A. Ioannou and J. Sun. Robust Adaptive Control. Prentice Hall, 1996.","identifier":"","url":""},{"text":"S. Trimpe and R. D'Andrea. Event-Based State Estimation with Variance-Based Triggering. IEEE Transactions on Automatic Control, 59(12), 2014.","identifier":"","url":""},{"text":"A. Mesbah. Stochastic Model Predictive Control with Active Uncertainty Learning: A Survey on Dual Control. Annual Reviews in Control, 45, 2018.","identifier":"","url":""},{"text":"G. Williams, P. Drews, B. Goldfain, J. M. Rehg, and E. A. Theodorou. Information-Theoretic Model Predictive Control: Theory and Applications to Autonomous Driving. IEEE Transactions on Robotics, 34(6), 2018.","identifier":"arXiv:1707.02342","url":"https://arxiv.org/abs/1707.02342"},{"text":"D. Fox, W. Burgard, and S. Thrun. The Dynamic Window Approach to Collision Avoidance. IEEE Robotics and Automation Magazine, 4(1), 1997.","identifier":"doi:10.1109/100.580977","url":"https://doi.org/10.1109/100.580977"},{"text":"L. Hewing, J. Kabzan, and M. N. Zeilinger. Cautious Model Predictive Control Using Gaussian Process Regression. IEEE Transactions on Control Systems Technology, 28(6), 2020.","identifier":"arXiv:1705.10702","url":"https://arxiv.org/abs/1705.10702"},{"text":"F. Berkenkamp, M. Turchetta, A. P. Schoellig, and A. Krause. Safe Model-Based Reinforcement Learning with Stability Guarantees. Advances in Neural Information Processing Systems (NeurIPS), 2017.","identifier":"arXiv:1705.08551","url":"https://arxiv.org/abs/1705.08551"},{"text":"L. Wellhausen and M. Hutter. ArtPlanner: Robust Legged Robot Navigation in the Field. Field Robotics, 3, 2023.","identifier":"arXiv:2303.01420","url":"https://arxiv.org/abs/2303.01420"},{"text":"T. Miki, L. Wellhausen, R. Grandia, F. Jenelten, T. Homberger, and M. Hutter. Elevation Mapping for Locomotion and Navigation Using GPU. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2022.","identifier":"arXiv:2204.12876","url":"https://arxiv.org/abs/2204.12876"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. Robotics: Science and Systems (RSS), 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"M. Tranzatto et al.. CERBERUS: Autonomous Legged and Aerial Robotic Exploration in the Tunnel and Urban Circuits of the DARPA Subterranean Challenge. Field Robotics, 2, 2022.","identifier":"arXiv:2201.07067","url":"https://arxiv.org/abs/2201.07067"},{"text":"A. Agha et al.. NeBula: Quest for Robotic Autonomy in Challenging Environments; TEAM CoSTAR at the DARPA Subterranean Challenge. arXiv:2103.11470, 2021.","identifier":"arXiv:2103.11470","url":"https://arxiv.org/abs/2103.11470"},{"text":"M. Bloesch, M. Hutter, M. A. Hoepflinger, S. Leutenegger, C. Gehring, C. D. Remy, and R. Siegwart. State Estimation for Legged Robots: Consistent Fusion of Leg Kinematics and IMU. Robotics: Science and Systems (RSS), 2012.","identifier":"","url":""},{"text":"D. Wisth, M. Camurri, and M. Fallon. VILENS: Visual, Inertial, Lidar, and Leg Odometry for All-Terrain Legged Robots. IEEE Transactions on Robotics, 39(1), 2023.","identifier":"arXiv:2107.07243","url":"https://arxiv.org/abs/2107.07243"},{"text":"W. Xu, Y. Cai, D. He, J. Lin, and F. Zhang. FAST-LIO2: Fast Direct LiDAR-Inertial Odometry. IEEE Transactions on Robotics, 38(4), 2022.","identifier":"arXiv:2107.06829","url":"https://arxiv.org/abs/2107.06829"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2609.00003","metadata":{"title":"Closed-Loop Field Calibration of Quadruped Velocity Interfaces","authors":[{"name":"Wei Xu","affiliation":"Autonomous Robotics Center, Carnegie Mellon University, Pittsburgh, PA, USA"},{"name":"Arjun Mehta","affiliation":"Autonomous Robotics Center, Carnegie Mellon University, Pittsburgh, PA, USA"},{"name":"Ethan Cole","affiliation":"Autonomous Robotics Center, Carnegie Mellon University, Pittsburgh, PA, USA"}],"publication_date":"2026-09-05","abstract":"Quadruped navigation stacks usually command a locomotion controller through a velocity interface whose command-to-motion response must be calibrated before the planner's predictions can be trusted. Open-loop active calibration, such as planner-conditioned probing with GAUGE, yields an accurate model at calibration time, but the response drifts as payload, battery state, temperature, and wear change during deployment. This paper closes the calibration loop. The calibrated interface model is embedded in the local planner, and the prediction errors it incurs during normal operation are treated as passive observations that update a Gaussian posterior over the interface parameters. Active probes are issued only when the planner-relevant predictive uncertainty crosses a threshold, and a residual-decomposition damping rule separates genuine parameter change from structural model mismatch so that the estimator does not chatter when the operating point leaves the model class. The analysis gives conditions under which the posterior covariance and the parameter error remain bounded under bounded mismatch. The estimator was evaluated on an ANYmal B during a three-day field campaign with 18 h of operation, payloads of 0, 6, and 12 kg, battery levels from 100 to 30 percent, and 2.1 M observations logged at 50 Hz. At the same probe frequency of 0.4 probes per hour, navigation success was 90.1 percent, compared with 84.1 percent for recalibration every 4 h and 80.5 percent for a static day-0 model, while the improvement in interface drift was small (2.9 versus 3.0 and 3.4 percent per day).","references":[{"text":"T. Zang, H. Huang, S. Cheng, and W. Zhang. GAUGE: Planner-Conditioned Active Calibration of Opaque Quadruped Velocity Interfaces. arXiv:2607.02841, 2026.","identifier":"arXiv:2607.02841","url":"https://arxiv.org/abs/2607.02841"},{"text":"M. Hutter et al.. ANYmal - A Highly Mobile and Dynamic Quadrupedal Robot. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2016.","identifier":"doi:10.1109/IROS.2016.7758092","url":"https://doi.org/10.1109/IROS.2016.7758092"},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. Science Robotics, 4(26), 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. Science Robotics, 5(47), 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. Science Robotics, 7(62), 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. Conference on Robot Learning (CoRL), 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"G. B. Margolis, G. Yang, K. Paigwar, T. Chen, and P. Agrawal. Rapid Locomotion via Reinforcement Learning. Robotics: Science and Systems (RSS), 2022.","identifier":"arXiv:2205.02824","url":"https://arxiv.org/abs/2205.02824"},{"text":"G. B. Margolis and P. Agrawal. Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior. Conference on Robot Learning (CoRL), 2022.","identifier":"arXiv:2212.03238","url":"https://arxiv.org/abs/2212.03238"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. Science Robotics, 9(88), 2024.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"S. Ha, J. Lee, M. van de Panne, Z. Xie, W. Yu, and M. Khadiv. Learning-Based Legged Locomotion: State of the Art and Future Perspectives. The International Journal of Robotics Research, 2025.","identifier":"arXiv:2406.01152","url":"https://arxiv.org/abs/2406.01152"},{"text":"N. Seegmiller, F. Rogers-Marcovitz, G. Miller, and A. Kelly. Vehicle Model Identification by Integrated Prediction Error Minimization. The International Journal of Robotics Research, 32(8), 2013.","identifier":"","url":""},{"text":"C. J. Ostafew, A. P. Schoellig, and T. D. Barfoot. Robust Constrained Learning-Based NMPC Enabling Reliable Mobile Robot Path Tracking. The International Journal of Robotics Research, 35(13), 2016.","identifier":"","url":""},{"text":"L. Ljung. System Identification: Theory for the User, 2nd ed.. Prentice Hall, 1999.","identifier":"","url":""},{"text":"M. Gevers. Identification for Control: From the Early Achievements to the Revival of Experiment Design. European Journal of Control, 11(4-5), 2005.","identifier":"","url":""},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. Robotics: Science and Systems (RSS), 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. IEEE International Conference on Robotics and Automation (ICRA), 2018.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"},{"text":"J. Tan et al.. Sim-to-Real: Learning Agile Locomotion for Quadruped Robots. Robotics: Science and Systems (RSS), 2018.","identifier":"arXiv:1804.10332","url":"https://arxiv.org/abs/1804.10332"},{"text":"A. Nagabandi et al.. Learning to Adapt in Dynamic, Real-World Environments through Meta-Reinforcement Learning. International Conference on Learning Representations (ICLR), 2019.","identifier":"arXiv:1803.11347","url":"https://arxiv.org/abs/1803.11347"},{"text":"M. Sombolestan, Y. Chen, and Q. Nguyen. Adaptive Force-Based Control for Legged Robots. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2021.","identifier":"arXiv:2011.06236","url":"https://arxiv.org/abs/2011.06236"},{"text":"R. E. Kalman. A New Approach to Linear Filtering and Prediction Problems. Transactions of the ASME, Journal of Basic Engineering, 82(1), 1960.","identifier":"doi:10.1115/1.3662552","url":"https://doi.org/10.1115/1.3662552"},{"text":"R. K. Mehra. On the Identification of Variances and Adaptive Kalman Filtering. IEEE Transactions on Automatic Control, 15(2), 1970.","identifier":"doi:10.1109/TAC.1970.1099422","url":"https://doi.org/10.1109/TAC.1970.1099422"},{"text":"K. Reif, S. Günther, E. Yaz, and R. Unbehauen. Stochastic Stability of the Discrete-Time Extended Kalman Filter. IEEE Transactions on Automatic Control, 44(4), 1999.","identifier":"doi:10.1109/9.754809","url":"https://doi.org/10.1109/9.754809"},{"text":"Y. Bar-Shalom, X. R. Li, and T. Kirubarajan. Estimation with Applications to Tracking and Navigation. Wiley, 2001.","identifier":"","url":""},{"text":"P. J. Huber. Robust Estimation of a Location Parameter. The Annals of Mathematical Statistics, 35(1), 1964.","identifier":"doi:10.1214/aoms/1177703732","url":"https://doi.org/10.1214/aoms/1177703732"},{"text":"R. Kulhavý and M. B. Zarrop. On a General Concept of Forgetting. International Journal of Control, 58(4), 1993.","identifier":"","url":""},{"text":"P. A. Ioannou and J. Sun. Robust Adaptive Control. Prentice Hall, 1996.","identifier":"","url":""},{"text":"S. Trimpe and R. D'Andrea. Event-Based State Estimation with Variance-Based Triggering. IEEE Transactions on Automatic Control, 59(12), 2014.","identifier":"","url":""},{"text":"A. Mesbah. Stochastic Model Predictive Control with Active Uncertainty Learning: A Survey on Dual Control. Annual Reviews in Control, 45, 2018.","identifier":"","url":""},{"text":"G. Williams, P. Drews, B. Goldfain, J. M. Rehg, and E. A. Theodorou. Information-Theoretic Model Predictive Control: Theory and Applications to Autonomous Driving. IEEE Transactions on Robotics, 34(6), 2018.","identifier":"arXiv:1707.02342","url":"https://arxiv.org/abs/1707.02342"},{"text":"D. Fox, W. Burgard, and S. Thrun. The Dynamic Window Approach to Collision Avoidance. IEEE Robotics and Automation Magazine, 4(1), 1997.","identifier":"doi:10.1109/100.580977","url":"https://doi.org/10.1109/100.580977"},{"text":"L. Hewing, J. Kabzan, and M. N. Zeilinger. Cautious Model Predictive Control Using Gaussian Process Regression. IEEE Transactions on Control Systems Technology, 28(6), 2020.","identifier":"arXiv:1705.10702","url":"https://arxiv.org/abs/1705.10702"},{"text":"F. Berkenkamp, M. Turchetta, A. P. Schoellig, and A. Krause. Safe Model-Based Reinforcement Learning with Stability Guarantees. Advances in Neural Information Processing Systems (NeurIPS), 2017.","identifier":"arXiv:1705.08551","url":"https://arxiv.org/abs/1705.08551"},{"text":"L. Wellhausen and M. Hutter. ArtPlanner: Robust Legged Robot Navigation in the Field. Field Robotics, 3, 2023.","identifier":"arXiv:2303.01420","url":"https://arxiv.org/abs/2303.01420"},{"text":"T. Miki, L. Wellhausen, R. Grandia, F. Jenelten, T. Homberger, and M. Hutter. Elevation Mapping for Locomotion and Navigation Using GPU. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2022.","identifier":"arXiv:2204.12876","url":"https://arxiv.org/abs/2204.12876"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. Robotics: Science and Systems (RSS), 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"M. Tranzatto et al.. CERBERUS: Autonomous Legged and Aerial Robotic Exploration in the Tunnel and Urban Circuits of the DARPA Subterranean Challenge. Field Robotics, 2, 2022.","identifier":"arXiv:2201.07067","url":"https://arxiv.org/abs/2201.07067"},{"text":"A. Agha et al.. NeBula: Quest for Robotic Autonomy in Challenging Environments; TEAM CoSTAR at the DARPA Subterranean Challenge. arXiv:2103.11470, 2021.","identifier":"arXiv:2103.11470","url":"https://arxiv.org/abs/2103.11470"},{"text":"M. Bloesch, M. Hutter, M. A. Hoepflinger, S. Leutenegger, C. Gehring, C. D. Remy, and R. Siegwart. State Estimation for Legged Robots: Consistent Fusion of Leg Kinematics and IMU. Robotics: Science and Systems (RSS), 2012.","identifier":"","url":""},{"text":"D. Wisth, M. Camurri, and M. Fallon. VILENS: Visual, Inertial, Lidar, and Leg Odometry for All-Terrain Legged Robots. IEEE Transactions on Robotics, 39(1), 2023.","identifier":"arXiv:2107.07243","url":"https://arxiv.org/abs/2107.07243"},{"text":"W. Xu, Y. Cai, D. He, J. Lin, and F. Zhang. FAST-LIO2: Fast Direct LiDAR-Inertial Odometry. IEEE Transactions on Robotics, 38(4), 2022.","identifier":"arXiv:2107.06829","url":"https://arxiv.org/abs/2107.06829"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2609.00003","slug":"2609.00003","registered_at":"2026-09-09T16:14:00Z","updated_at":"2026-09-10T17:09:00Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2609.00003","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2609.00003","tex_path":"articles/2609.00003/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-09-09T18:31:00Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2609.00003"},{"kagg_id":"KAGG:2609.00002","slug":"2609.00002","title":"Distilling Cross-Embodiment Planners into Onboard-Deployable Policies","authors":[{"name":"Rongxin Guo","affiliation":"Robotics and Autonomous Systems Lab, Shanghai Jiao Tong University, Shanghai, China"},{"name":"Xiaotong Zhang","affiliation":"Robotics and Autonomous Systems Lab, Shanghai Jiao Tong University, Shanghai, China"},{"name":"Yifan Chen","affiliation":"Robotics and Autonomous Systems Lab, Shanghai Jiao Tong University, Shanghai, China"},{"name":"Daichi Mori","affiliation":"Legged Systems Group, The University of Tokyo, Tokyo, Japan"}],"affiliation":null,"abstract":"Configuration-driven cross-embodiment local planners such as AgniNav allow a single planning stack to serve robots with different body geometry, sensing, and velocity limits, but they pay for this generality with full configuration-space sampling at every replan. On an embedded computer this cost is prohibitive: the full planner requires 41 ms per decision offline, and throttling it onboard degrades agreement with its unthrottled output. This paper distills such a planner into a configuration-conditioned onboard policy through trajectory-level imitation. The teacher produces local trajectories over configuration states; the student maps an observation and a configuration descriptor directly to the next body-motion primitive from a velocity-normalized library. Training combines a soft-target imitation term with a consistency regularizer that penalizes, under perturbed range observations, probability mass on primitives that diverge geometrically from the teacher decision. Because the student remains conditioned on the descriptor, one deployed network serves multiple embodiments. In Isaac Lab experiments with a Unitree Go2 across 18 scenes and 5 seeds, with the student running at 100 Hz on an onboard Orin, the distilled policy reaches a p95 decision latency of 7.8 ms, a teacher parity of 94.3 percent, and a success rate of 91.7 percent, compared with 8.7 ms, 84.2 percent, and 86.3 percent for the throttled full planner. Transfer to a B2-class morphology in simulation retains 89.6 percent parity, and at 10 percent range dropout the regularizer preserves 88.1 percent parity versus 79.4 percent without it.","publication_date":"2026-09-03","registered_at":"2026-09-09T18:00:00Z","updated_at":"2026-09-10T18:17:00Z","status":"published","html_path":null,"pdf_path":"articles/2609.00002/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Sridhar, D. Shah, C. Glossop, and S. Levine. NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration. arXiv:2310.07896, 2023.","identifier":"arXiv:2310.07896","url":"https://arxiv.org/abs/2310.07896"},{"text":"D. Chen, B. Zhou, V. Koltun, and P. Krähenbühl. Learning by Cheating. arXiv:1912.12294, 2019.","identifier":"arXiv:1912.12294","url":"https://arxiv.org/abs/1912.12294"},{"text":"A. Loquercio, E. Kaufmann, R. Ranftl, M. Müller, V. Koltun, and D. Scaramuzza. Learning High-Speed Flight in the Wild. arXiv:2110.05113, 2021.","identifier":"arXiv:2110.05113","url":"https://arxiv.org/abs/2110.05113"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"J. Zhang, Y. Du, X. Guo, S. Sun, X. Liu, Y. Sun, G. Lu, W. Sui, and J. Li. EA-Nav: Learning Safe Visual Navigation Policies with Embodiment Awareness. arXiv:2607.19880, 2026.","identifier":"arXiv:2607.19880","url":"https://arxiv.org/abs/2607.19880"},{"text":"A. Eftekhar et al. The One RING: a Robotic Indoor Navigation Generalist. arXiv:2412.14401, 2024.","identifier":"arXiv:2412.14401","url":"https://arxiv.org/abs/2412.14401"},{"text":"J. Zhang et al. Embodied Navigation Foundation Model. arXiv:2509.12129, 2025.","identifier":"arXiv:2509.12129","url":"https://arxiv.org/abs/2509.12129"},{"text":"J. Yang, C. Glossop, A. Bhorkar, D. Shah, Q. Vuong, C. Finn, D. Sadigh, and S. Levine. Pushing the Limits of Cross-Embodiment Learning for Manipulation and Navigation. arXiv:2402.19432, 2024.","identifier":"arXiv:2402.19432","url":"https://arxiv.org/abs/2402.19432"},{"text":"W. Huang, I. Mordatch, and D. Pathak. One Policy to Control Them All: Shared Modular Policies for Agent-Agnostic Control. arXiv:2007.04976, 2020.","identifier":"arXiv:2007.04976","url":"https://arxiv.org/abs/2007.04976"},{"text":"A. Gupta, L. Fan, S. Ganguli, and L. Fei-Fei. MetaMorph: Learning Universal Controllers with Transformers. arXiv:2203.11931, 2022.","identifier":"arXiv:2203.11931","url":"https://arxiv.org/abs/2203.11931"},{"text":"G. Feng, H. Zhang, Z. Li, X. B. Peng, B. Basireddy, L. Yue, Z. Song, L. Yang, Y. Liu, K. Sreenath, and S. Levine. GenLoco: Generalized Locomotion Controllers for Quadrupedal Robots. arXiv:2209.05309, 2022.","identifier":"arXiv:2209.05309","url":"https://arxiv.org/abs/2209.05309"},{"text":"M. Shafiee, G. Bellegarda, and A. Ijspeert. ManyQuadrupeds: Learning a Single Locomotion Policy for Diverse Quadruped Robots. arXiv:2310.10486, 2023.","identifier":"arXiv:2310.10486","url":"https://arxiv.org/abs/2310.10486"},{"text":"N. Bohlinger, G. Czechmanowski, M. Krupka, P. Kicki, K. Walas, J. Peters, and D. Tateo. One Policy to Run Them All: an End-to-end Learning Approach to Multi-Embodiment Locomotion. arXiv:2409.06366, 2024.","identifier":"arXiv:2409.06366","url":"https://arxiv.org/abs/2409.06366"},{"text":"G. Hinton, O. Vinyals, and J. Dean. Distilling the Knowledge in a Neural Network. arXiv:1503.02531, 2015.","identifier":"arXiv:1503.02531","url":"https://arxiv.org/abs/1503.02531"},{"text":"A. A. Rusu, S. G. Colmenarejo, C. Gulcehre, G. Desjardins, J. Kirkpatrick, R. Pascanu, V. Mnih, K. Kavukcuoglu, and R. Hadsell. Policy Distillation. arXiv:1511.06295, 2015.","identifier":"arXiv:1511.06295","url":"https://arxiv.org/abs/1511.06295"},{"text":"W. M. Czarnecki, R. Pascanu, S. Osindero, S. M. Jayakumar, G. Swirszcz, and M. Jaderberg. Distilling Policy Distillation. arXiv:1902.02186, 2019.","identifier":"arXiv:1902.02186","url":"https://arxiv.org/abs/1902.02186"},{"text":"S. Ross, G. J. Gordon, and J. A. Bagnell. A Reduction of Imitation Learning and Structured Prediction to No-Regret Online Learning. arXiv:1011.0686, 2011.","identifier":"arXiv:1011.0686","url":"https://arxiv.org/abs/1011.0686"},{"text":"S. Levine, C. Finn, T. Darrell, and P. Abbeel. End-to-End Training of Deep Visuomotor Policies. arXiv:1504.00702, 2015.","identifier":"arXiv:1504.00702","url":"https://arxiv.org/abs/1504.00702"},{"text":"T. Zhang, G. Kahn, S. Levine, and P. Abbeel. Learning Deep Control Policies for Autonomous Aerial Vehicles with MPC-Guided Policy Search. arXiv:1509.06791, 2015.","identifier":"arXiv:1509.06791","url":"https://arxiv.org/abs/1509.06791"},{"text":"Y. Pan, C.-A. Cheng, K. Saigol, K. Lee, X. Yan, E. Theodorou, and B. Boots. Agile Autonomous Driving using End-to-End Deep Imitation Learning. arXiv:1709.07174, 2017.","identifier":"arXiv:1709.07174","url":"https://arxiv.org/abs/1709.07174"},{"text":"E. Kaufmann, A. Loquercio, R. Ranftl, M. Müller, V. Koltun, and D. Scaramuzza. Deep Drone Acrobatics. arXiv:2006.05768, 2020.","identifier":"arXiv:2006.05768","url":"https://arxiv.org/abs/2006.05768"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. arXiv:2211.07638, 2022.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. arXiv:2309.05665, 2023.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. arXiv:2309.14341, 2023.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"M. Pfeiffer, M. Schaeuble, J. Nieto, R. Siegwart, and C. Cadena. From Perception to Decision: A Data-driven Approach to End-to-end Motion Planning for Autonomous Ground Robots. arXiv:1609.07910, 2016.","identifier":"arXiv:1609.07910","url":"https://arxiv.org/abs/1609.07910"},{"text":"X. Xiao, B. Liu, G. Warnell, and P. Stone. Motion Planning and Control for Mobile Robot Navigation Using Machine Learning: a Survey. arXiv:2011.13112, 2020.","identifier":"arXiv:2011.13112","url":"https://arxiv.org/abs/2011.13112"},{"text":"X. Xiao, B. Liu, G. Warnell, J. Fink, and P. Stone. APPLD: Adaptive Planner Parameter Learning from Demonstration. arXiv:2004.00116, 2020.","identifier":"arXiv:2004.00116","url":"https://arxiv.org/abs/2004.00116"},{"text":"G. Williams, A. Aldrich, and E. Theodorou. Model Predictive Path Integral Control using Covariance Variable Importance Sampling. arXiv:1509.01149, 2015.","identifier":"arXiv:1509.01149","url":"https://arxiv.org/abs/1509.01149"},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"L. Wellhausen and M. Hutter. ArtPlanner: Robust Legged Robot Navigation in the Field. arXiv:2303.01420, 2023.","identifier":"arXiv:2303.01420","url":"https://arxiv.org/abs/2303.01420"},{"text":"J. Lee, M. Bjelonic, A. Reske, L. Wellhausen, T. Miki, and M. Hutter. Learning Robust Autonomous Navigation and Locomotion for Wheeled-Legged Robots. arXiv:2405.01792, 2024.","identifier":"arXiv:2405.01792","url":"https://arxiv.org/abs/2405.01792"},{"text":"C. Chi, S. Feng, Y. Du, Z. Xu, E. Cousineau, B. Burchfiel, and S. Song. Diffusion Policy: Visuomotor Policy Learning via Action Diffusion. arXiv:2303.04137, 2023.","identifier":"arXiv:2303.04137","url":"https://arxiv.org/abs/2303.04137"},{"text":"E. Perez, F. Strub, H. de Vries, V. Dumoulin, and A. Courville. FiLM: Visual Reasoning with a General Conditioning Layer. arXiv:1709.07871, 2017.","identifier":"arXiv:1709.07871","url":"https://arxiv.org/abs/1709.07871"},{"text":"V. Makoviychuk et al. Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal et al. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. arXiv:2301.04195, 2023.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. arXiv:1710.06537, 2017.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"},{"text":"A. Kadian et al. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance? arXiv:1912.06321, 2019.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2609.00002","metadata":{"title":"Distilling Cross-Embodiment Planners into Onboard-Deployable Policies","authors":[{"name":"Rongxin Guo","affiliation":"Robotics and Autonomous Systems Lab, Shanghai Jiao Tong University, Shanghai, China"},{"name":"Xiaotong Zhang","affiliation":"Robotics and Autonomous Systems Lab, Shanghai Jiao Tong University, Shanghai, China"},{"name":"Yifan Chen","affiliation":"Robotics and Autonomous Systems Lab, Shanghai Jiao Tong University, Shanghai, China"},{"name":"Daichi Mori","affiliation":"Legged Systems Group, The University of Tokyo, Tokyo, Japan"}],"publication_date":"2026-09-03","abstract":"Configuration-driven cross-embodiment local planners such as AgniNav allow a single planning stack to serve robots with different body geometry, sensing, and velocity limits, but they pay for this generality with full configuration-space sampling at every replan. On an embedded computer this cost is prohibitive: the full planner requires 41 ms per decision offline, and throttling it onboard degrades agreement with its unthrottled output. This paper distills such a planner into a configuration-conditioned onboard policy through trajectory-level imitation. The teacher produces local trajectories over configuration states; the student maps an observation and a configuration descriptor directly to the next body-motion primitive from a velocity-normalized library. Training combines a soft-target imitation term with a consistency regularizer that penalizes, under perturbed range observations, probability mass on primitives that diverge geometrically from the teacher decision. Because the student remains conditioned on the descriptor, one deployed network serves multiple embodiments. In Isaac Lab experiments with a Unitree Go2 across 18 scenes and 5 seeds, with the student running at 100 Hz on an onboard Orin, the distilled policy reaches a p95 decision latency of 7.8 ms, a teacher parity of 94.3 percent, and a success rate of 91.7 percent, compared with 8.7 ms, 84.2 percent, and 86.3 percent for the throttled full planner. Transfer to a B2-class morphology in simulation retains 89.6 percent parity, and at 10 percent range dropout the regularizer preserves 88.1 percent parity versus 79.4 percent without it.","references":[{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Sridhar, D. Shah, C. Glossop, and S. Levine. NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration. arXiv:2310.07896, 2023.","identifier":"arXiv:2310.07896","url":"https://arxiv.org/abs/2310.07896"},{"text":"D. Chen, B. Zhou, V. Koltun, and P. Krähenbühl. Learning by Cheating. arXiv:1912.12294, 2019.","identifier":"arXiv:1912.12294","url":"https://arxiv.org/abs/1912.12294"},{"text":"A. Loquercio, E. Kaufmann, R. Ranftl, M. Müller, V. Koltun, and D. Scaramuzza. Learning High-Speed Flight in the Wild. arXiv:2110.05113, 2021.","identifier":"arXiv:2110.05113","url":"https://arxiv.org/abs/2110.05113"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"J. Zhang, Y. Du, X. Guo, S. Sun, X. Liu, Y. Sun, G. Lu, W. Sui, and J. Li. EA-Nav: Learning Safe Visual Navigation Policies with Embodiment Awareness. arXiv:2607.19880, 2026.","identifier":"arXiv:2607.19880","url":"https://arxiv.org/abs/2607.19880"},{"text":"A. Eftekhar et al. The One RING: a Robotic Indoor Navigation Generalist. arXiv:2412.14401, 2024.","identifier":"arXiv:2412.14401","url":"https://arxiv.org/abs/2412.14401"},{"text":"J. Zhang et al. Embodied Navigation Foundation Model. arXiv:2509.12129, 2025.","identifier":"arXiv:2509.12129","url":"https://arxiv.org/abs/2509.12129"},{"text":"J. Yang, C. Glossop, A. Bhorkar, D. Shah, Q. Vuong, C. Finn, D. Sadigh, and S. Levine. Pushing the Limits of Cross-Embodiment Learning for Manipulation and Navigation. arXiv:2402.19432, 2024.","identifier":"arXiv:2402.19432","url":"https://arxiv.org/abs/2402.19432"},{"text":"W. Huang, I. Mordatch, and D. Pathak. One Policy to Control Them All: Shared Modular Policies for Agent-Agnostic Control. arXiv:2007.04976, 2020.","identifier":"arXiv:2007.04976","url":"https://arxiv.org/abs/2007.04976"},{"text":"A. Gupta, L. Fan, S. Ganguli, and L. Fei-Fei. MetaMorph: Learning Universal Controllers with Transformers. arXiv:2203.11931, 2022.","identifier":"arXiv:2203.11931","url":"https://arxiv.org/abs/2203.11931"},{"text":"G. Feng, H. Zhang, Z. Li, X. B. Peng, B. Basireddy, L. Yue, Z. Song, L. Yang, Y. Liu, K. Sreenath, and S. Levine. GenLoco: Generalized Locomotion Controllers for Quadrupedal Robots. arXiv:2209.05309, 2022.","identifier":"arXiv:2209.05309","url":"https://arxiv.org/abs/2209.05309"},{"text":"M. Shafiee, G. Bellegarda, and A. Ijspeert. ManyQuadrupeds: Learning a Single Locomotion Policy for Diverse Quadruped Robots. arXiv:2310.10486, 2023.","identifier":"arXiv:2310.10486","url":"https://arxiv.org/abs/2310.10486"},{"text":"N. Bohlinger, G. Czechmanowski, M. Krupka, P. Kicki, K. Walas, J. Peters, and D. Tateo. One Policy to Run Them All: an End-to-end Learning Approach to Multi-Embodiment Locomotion. arXiv:2409.06366, 2024.","identifier":"arXiv:2409.06366","url":"https://arxiv.org/abs/2409.06366"},{"text":"G. Hinton, O. Vinyals, and J. Dean. Distilling the Knowledge in a Neural Network. arXiv:1503.02531, 2015.","identifier":"arXiv:1503.02531","url":"https://arxiv.org/abs/1503.02531"},{"text":"A. A. Rusu, S. G. Colmenarejo, C. Gulcehre, G. Desjardins, J. Kirkpatrick, R. Pascanu, V. Mnih, K. Kavukcuoglu, and R. Hadsell. Policy Distillation. arXiv:1511.06295, 2015.","identifier":"arXiv:1511.06295","url":"https://arxiv.org/abs/1511.06295"},{"text":"W. M. Czarnecki, R. Pascanu, S. Osindero, S. M. Jayakumar, G. Swirszcz, and M. Jaderberg. Distilling Policy Distillation. arXiv:1902.02186, 2019.","identifier":"arXiv:1902.02186","url":"https://arxiv.org/abs/1902.02186"},{"text":"S. Ross, G. J. Gordon, and J. A. Bagnell. A Reduction of Imitation Learning and Structured Prediction to No-Regret Online Learning. arXiv:1011.0686, 2011.","identifier":"arXiv:1011.0686","url":"https://arxiv.org/abs/1011.0686"},{"text":"S. Levine, C. Finn, T. Darrell, and P. Abbeel. End-to-End Training of Deep Visuomotor Policies. arXiv:1504.00702, 2015.","identifier":"arXiv:1504.00702","url":"https://arxiv.org/abs/1504.00702"},{"text":"T. Zhang, G. Kahn, S. Levine, and P. Abbeel. Learning Deep Control Policies for Autonomous Aerial Vehicles with MPC-Guided Policy Search. arXiv:1509.06791, 2015.","identifier":"arXiv:1509.06791","url":"https://arxiv.org/abs/1509.06791"},{"text":"Y. Pan, C.-A. Cheng, K. Saigol, K. Lee, X. Yan, E. Theodorou, and B. Boots. Agile Autonomous Driving using End-to-End Deep Imitation Learning. arXiv:1709.07174, 2017.","identifier":"arXiv:1709.07174","url":"https://arxiv.org/abs/1709.07174"},{"text":"E. Kaufmann, A. Loquercio, R. Ranftl, M. Müller, V. Koltun, and D. Scaramuzza. Deep Drone Acrobatics. arXiv:2006.05768, 2020.","identifier":"arXiv:2006.05768","url":"https://arxiv.org/abs/2006.05768"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. arXiv:2211.07638, 2022.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. arXiv:2309.05665, 2023.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. arXiv:2309.14341, 2023.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"M. Pfeiffer, M. Schaeuble, J. Nieto, R. Siegwart, and C. Cadena. From Perception to Decision: A Data-driven Approach to End-to-end Motion Planning for Autonomous Ground Robots. arXiv:1609.07910, 2016.","identifier":"arXiv:1609.07910","url":"https://arxiv.org/abs/1609.07910"},{"text":"X. Xiao, B. Liu, G. Warnell, and P. Stone. Motion Planning and Control for Mobile Robot Navigation Using Machine Learning: a Survey. arXiv:2011.13112, 2020.","identifier":"arXiv:2011.13112","url":"https://arxiv.org/abs/2011.13112"},{"text":"X. Xiao, B. Liu, G. Warnell, J. Fink, and P. Stone. APPLD: Adaptive Planner Parameter Learning from Demonstration. arXiv:2004.00116, 2020.","identifier":"arXiv:2004.00116","url":"https://arxiv.org/abs/2004.00116"},{"text":"G. Williams, A. Aldrich, and E. Theodorou. Model Predictive Path Integral Control using Covariance Variable Importance Sampling. arXiv:1509.01149, 2015.","identifier":"arXiv:1509.01149","url":"https://arxiv.org/abs/1509.01149"},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"L. Wellhausen and M. Hutter. ArtPlanner: Robust Legged Robot Navigation in the Field. arXiv:2303.01420, 2023.","identifier":"arXiv:2303.01420","url":"https://arxiv.org/abs/2303.01420"},{"text":"J. Lee, M. Bjelonic, A. Reske, L. Wellhausen, T. Miki, and M. Hutter. Learning Robust Autonomous Navigation and Locomotion for Wheeled-Legged Robots. arXiv:2405.01792, 2024.","identifier":"arXiv:2405.01792","url":"https://arxiv.org/abs/2405.01792"},{"text":"C. Chi, S. Feng, Y. Du, Z. Xu, E. Cousineau, B. Burchfiel, and S. Song. Diffusion Policy: Visuomotor Policy Learning via Action Diffusion. arXiv:2303.04137, 2023.","identifier":"arXiv:2303.04137","url":"https://arxiv.org/abs/2303.04137"},{"text":"E. Perez, F. Strub, H. de Vries, V. Dumoulin, and A. Courville. FiLM: Visual Reasoning with a General Conditioning Layer. arXiv:1709.07871, 2017.","identifier":"arXiv:1709.07871","url":"https://arxiv.org/abs/1709.07871"},{"text":"V. Makoviychuk et al. Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal et al. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. arXiv:2301.04195, 2023.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. arXiv:1710.06537, 2017.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"},{"text":"A. Kadian et al. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance? arXiv:1912.06321, 2019.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2609.00002","slug":"2609.00002","registered_at":"2026-09-09T18:00:00Z","updated_at":"2026-09-10T18:17:00Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2609.00002","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2609.00002","tex_path":"articles/2609.00002/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-09-09T20:17:00Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2609.00002"},{"kagg_id":"KAGG:2609.00001","slug":"2609.00001","title":"Predictive Terrain Range Maps for Energy-Efficient Quadruped Navigation","authors":[{"name":"Jialin Ma","affiliation":"Intelligent Mobility Lab, University of Science and Technology of China, Hefei, China"},{"name":"Mingxuan Li","affiliation":"Intelligent Mobility Lab, University of Science and Technology of China, Hefei, China"},{"name":"Yutong Zhou","affiliation":"Intelligent Mobility Lab, University of Science and Technology of China, Hefei, China"}],"affiliation":null,"abstract":"Energy-efficient quadruped navigation requires knowing not only where obstacles are but also which feasible routes waste energy in rough terrain. Control-aligned future-range prediction estimates free range along the directions the robot's controller can actually execute, but it treats every free meter as equally cheap. This paper extends that representation into a predictive terrain-range map. Each control-aligned ray is annotated with an effort cost obtained by integrating a calibrated interface energy model, which relates terrain class, slope, and speed to electrical power, along the ray's predicted path. The result is a cost-weighted range field that a receding-horizon local planner consumes together with a consistent global cost-to-go, and a single explicit scalar λ sets the exchange rate between distance and energy. For exact route optimization, increasing λ cannot increase route energy or shorten the route. The approach is evaluated on 10 Isaac Lab terrain maps with graded effort cost (5 seeds, 25 goal pairs each) and on a 2.4 km mixed-terrain field loop with a Unitree Go2 and battery telemetry. With λ = 1.0, the planner reaches 1184 J/km, 13.3 percent below a distance-only planner (1365 J/km), with a mean arrival time of 218 s versus 248 s and 96.8 percent goal success versus 90.7 percent. A sweep over λ traces an energy-time frontier from 1412 J/km at 187 s to 1086 J/km at 289 s with diminishing returns beyond λ = 2.0, while energy-only routing is vulnerable to detour loops.","publication_date":"2026-09-01","registered_at":"2026-09-08T15:23:00Z","updated_at":"2026-09-08T18:23:00Z","status":"published","html_path":null,"pdf_path":"articles/2609.00001/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. Science Robotics, 4(26), 2019. arXiv:1901.08652.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. Science Robotics, 5(47), 2020. arXiv:2010.11251.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. Science Robotics, 7(62), 2022. arXiv:2201.08117.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. Conference on Robot Learning (CoRL), 2021. arXiv:2109.11978.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. Robotics: Science and Systems (RSS), 2021. arXiv:2107.04034.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"T. Zang, S. Wang, H. Huang, and S. Cheng. FutureRay: Control-Aligned Future Range for Agile Quadruped Navigation. arXiv:2608.01197, 2026.","identifier":"arXiv:2608.01197","url":"https://arxiv.org/abs/2608.01197"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains Using Egocentric Vision. Conference on Robot Learning (CoRL), 2022. arXiv:2211.07638.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. Conference on Robot Learning (CoRL), 2023. arXiv:2309.05665.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. arXiv:2309.14341, 2023.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. Science Robotics, 9(88), 2024. arXiv:2306.14874.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"G. B. Margolis and P. Agrawal. Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior. Conference on Robot Learning (CoRL), 2022. arXiv:2212.03238.","identifier":"arXiv:2212.03238","url":"https://arxiv.org/abs/2212.03238"},{"text":"Z. Fu, A. Kumar, J. Malik, and D. Pathak. Minimizing Energy Consumption Leads to the Emergence of Gaits in Legged Robots. Conference on Robot Learning (CoRL), 2021. arXiv:2111.01674.","identifier":"arXiv:2111.01674","url":"https://arxiv.org/abs/2111.01674"},{"text":"Y. Yang, T. Zhang, E. Coumans, J. Tan, and B. Boots. Fast and Efficient Locomotion via Learned Gait Transitions. Conference on Robot Learning (CoRL), 2021. arXiv:2104.04644.","identifier":"arXiv:2104.04644","url":"https://arxiv.org/abs/2104.04644"},{"text":"C. Li, T. Zhang, and D. I. Goldman. A Terradynamics of Legged Locomotion on Granular Media. Science, 339(6126):1408–1412, 2013.","identifier":"doi:10.1126/science.1229163","url":"https://doi.org/10.1126/science.1229163"},{"text":"T. Miki, L. Wellhausen, R. Grandia, F. Jenelten, T. Homberger, and M. Hutter. Elevation Mapping for Locomotion and Navigation Using GPU. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2022. arXiv:2204.12876.","identifier":"arXiv:2204.12876","url":"https://arxiv.org/abs/2204.12876"},{"text":"D. Hoeller, L. Wellhausen, F. Farshidian, and M. Hutter. Learning a State Representation and Navigation in Cluttered and Dynamic Environments. IEEE Robotics and Automation Letters, 6(3), 2021. arXiv:2103.04351.","identifier":"arXiv:2103.04351","url":"https://arxiv.org/abs/2103.04351"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2022. arXiv:2209.12827.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"J. Lee, M. Bjelonic, A. Reske, L. Wellhausen, T. Miki, and M. Hutter. Learning Robust Autonomous Navigation and Locomotion for Wheeled-Legged Robots. Science Robotics, 9(89), 2024. arXiv:2405.01792.","identifier":"arXiv:2405.01792","url":"https://arxiv.org/abs/2405.01792"},{"text":"Z. Fu, A. Kumar, A. Agarwal, H. Qi, J. Malik, and D. Pathak. Coupling Vision and Proprioception for Navigation of Legged Robots. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022. arXiv:2112.02094.","identifier":"arXiv:2112.02094","url":"https://arxiv.org/abs/2112.02094"},{"text":"F. Yang, C. Wang, C. Cadena, and M. Hutter. iPlanner: Imperative Path Planning. Robotics: Science and Systems (RSS), 2023. arXiv:2302.11434.","identifier":"arXiv:2302.11434","url":"https://arxiv.org/abs/2302.11434"},{"text":"P. Roth, J. Nubert, F. Yang, M. Mittal, and M. Hutter. ViPlanner: Visual Semantic Imperative Learning for Local Navigation. IEEE International Conference on Robotics and Automation (ICRA), 2024. arXiv:2310.00982.","identifier":"arXiv:2310.00982","url":"https://arxiv.org/abs/2310.00982"},{"text":"A. Loquercio, E. Kaufmann, R. Ranftl, M. Müller, V. Koltun, and D. Scaramuzza. Learning High-Speed Flight in the Wild. Science Robotics, 6(59), 2021. arXiv:2110.05113.","identifier":"arXiv:2110.05113","url":"https://arxiv.org/abs/2110.05113"},{"text":"D. Fox, W. Burgard, and S. Thrun. The Dynamic Window Approach to Collision Avoidance. IEEE Robotics and Automation Magazine, 4(1):23–33, 1997.","identifier":"doi:10.1109/100.580977","url":"https://doi.org/10.1109/100.580977"},{"text":"G. Williams, P. Drews, B. Goldfain, J. M. Rehg, and E. A. Theodorou. Information-Theoretic Model Predictive Control: Theory and Applications to Autonomous Driving. IEEE Transactions on Robotics, 34(6), 2018. arXiv:1707.02342.","identifier":"arXiv:1707.02342","url":"https://arxiv.org/abs/1707.02342"},{"text":"G. Kahn, P. Abbeel, and S. Levine. BADGR: An Autonomous Self-Supervised Learning-Based Navigation System. IEEE Robotics and Automation Letters, 6(2), 2021. arXiv:2002.05700.","identifier":"arXiv:2002.05700","url":"https://arxiv.org/abs/2002.05700"},{"text":"M. G. Castro, S. Triest, W. Wang, J. M. Gregory, F. Sanchez, J. G. Rogers III, and S. Scherer. How Does It Feel? Self-Supervised Costmap Learning for Off-Road Vehicle Traversability. IEEE International Conference on Robotics and Automation (ICRA), 2023. arXiv:2209.10788.","identifier":"arXiv:2209.10788","url":"https://arxiv.org/abs/2209.10788"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. Robotics: Science and Systems (RSS), 2023. arXiv:2305.08510.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"A. J. Sathyamoorthy, K. Weerakoon, T. Guan, J. Liang, and D. Manocha. TerraPN: Unstructured Terrain Navigation Using Online Self-Supervised Learning. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2022. arXiv:2202.12873.","identifier":"arXiv:2202.12873","url":"https://arxiv.org/abs/2202.12873"},{"text":"T. Guan, D. Kothandaraman, R. Chandra, A. J. Sathyamoorthy, K. Weerakoon, and D. Manocha. GA-Nav: Efficient Terrain Segmentation for Robot Navigation in Unstructured Outdoor Environments. IEEE Robotics and Automation Letters, 7(3), 2022. arXiv:2103.04233.","identifier":"arXiv:2103.04233","url":"https://arxiv.org/abs/2103.04233"},{"text":"D. D. Fan, K. Otsu, Y. Kubo, A. Dixit, J. Burdick, and A.-A. Agha-Mohammadi. STEP: Stochastic Traversability Evaluation and Planning for Risk-Aware Off-Road Navigation. Robotics: Science and Systems (RSS), 2021. arXiv:2103.02828.","identifier":"arXiv:2103.02828","url":"https://arxiv.org/abs/2103.02828"},{"text":"X. Cai et al. EVORA: Deep Evidential Traversability Learning for Risk-Aware Off-Road Autonomy. arXiv:2311.06234, 2023.","identifier":"arXiv:2311.06234","url":"https://arxiv.org/abs/2311.06234"},{"text":"R. T. Marler and J. S. Arora. Survey of Multi-Objective Optimization Methods for Engineering. Structural and Multidisciplinary Optimization, 26(6):369–395, 2004.","identifier":"doi:10.1007/s00158-003-0368-6","url":"https://doi.org/10.1007/s00158-003-0368-6"},{"text":"I. Das and J. E. Dennis. A Closer Look at Drawbacks of Minimizing Weighted Sums of Objectives for Pareto Set Generation in Multicriteria Optimization Problems. Structural Optimization, 14(1):63–69, 1997.","identifier":"doi:10.1007/BF01197559","url":"https://doi.org/10.1007/BF01197559"},{"text":"P. E. Hart, N. J. Nilsson, and B. Raphael. A Formal Basis for the Heuristic Determination of Minimum Cost Paths. IEEE Transactions on Systems Science and Cybernetics, 4(2):100–107, 1968.","identifier":"doi:10.1109/TSSC.1968.300136","url":"https://doi.org/10.1109/TSSC.1968.300136"},{"text":"V. Makoviychuk et al. Isaac Gym: High Performance GPU-Based Physics Simulation for Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal et al. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. IEEE Robotics and Automation Letters, 8(6), 2023. arXiv:2301.04195.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2017. arXiv:1703.06907.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"},{"text":"A. Kadian et al. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance? IEEE Robotics and Automation Letters, 5(4), 2020. arXiv:1912.06321.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2609.00001","metadata":{"title":"Predictive Terrain Range Maps for Energy-Efficient Quadruped Navigation","authors":[{"name":"Jialin Ma","affiliation":"Intelligent Mobility Lab, University of Science and Technology of China, Hefei, China"},{"name":"Mingxuan Li","affiliation":"Intelligent Mobility Lab, University of Science and Technology of China, Hefei, China"},{"name":"Yutong Zhou","affiliation":"Intelligent Mobility Lab, University of Science and Technology of China, Hefei, China"}],"publication_date":"2026-09-01","abstract":"Energy-efficient quadruped navigation requires knowing not only where obstacles are but also which feasible routes waste energy in rough terrain. Control-aligned future-range prediction estimates free range along the directions the robot's controller can actually execute, but it treats every free meter as equally cheap. This paper extends that representation into a predictive terrain-range map. Each control-aligned ray is annotated with an effort cost obtained by integrating a calibrated interface energy model, which relates terrain class, slope, and speed to electrical power, along the ray's predicted path. The result is a cost-weighted range field that a receding-horizon local planner consumes together with a consistent global cost-to-go, and a single explicit scalar λ sets the exchange rate between distance and energy. For exact route optimization, increasing λ cannot increase route energy or shorten the route. The approach is evaluated on 10 Isaac Lab terrain maps with graded effort cost (5 seeds, 25 goal pairs each) and on a 2.4 km mixed-terrain field loop with a Unitree Go2 and battery telemetry. With λ = 1.0, the planner reaches 1184 J/km, 13.3 percent below a distance-only planner (1365 J/km), with a mean arrival time of 218 s versus 248 s and 96.8 percent goal success versus 90.7 percent. A sweep over λ traces an energy-time frontier from 1412 J/km at 187 s to 1086 J/km at 289 s with diminishing returns beyond λ = 2.0, while energy-only routing is vulnerable to detour loops.","references":[{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. Science Robotics, 4(26), 2019. arXiv:1901.08652.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. Science Robotics, 5(47), 2020. arXiv:2010.11251.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. Science Robotics, 7(62), 2022. arXiv:2201.08117.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. Conference on Robot Learning (CoRL), 2021. arXiv:2109.11978.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. Robotics: Science and Systems (RSS), 2021. arXiv:2107.04034.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"T. Zang, S. Wang, H. Huang, and S. Cheng. FutureRay: Control-Aligned Future Range for Agile Quadruped Navigation. arXiv:2608.01197, 2026.","identifier":"arXiv:2608.01197","url":"https://arxiv.org/abs/2608.01197"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains Using Egocentric Vision. Conference on Robot Learning (CoRL), 2022. arXiv:2211.07638.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. Conference on Robot Learning (CoRL), 2023. arXiv:2309.05665.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. arXiv:2309.14341, 2023.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. Science Robotics, 9(88), 2024. arXiv:2306.14874.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"G. B. Margolis and P. Agrawal. Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior. Conference on Robot Learning (CoRL), 2022. arXiv:2212.03238.","identifier":"arXiv:2212.03238","url":"https://arxiv.org/abs/2212.03238"},{"text":"Z. Fu, A. Kumar, J. Malik, and D. Pathak. Minimizing Energy Consumption Leads to the Emergence of Gaits in Legged Robots. Conference on Robot Learning (CoRL), 2021. arXiv:2111.01674.","identifier":"arXiv:2111.01674","url":"https://arxiv.org/abs/2111.01674"},{"text":"Y. Yang, T. Zhang, E. Coumans, J. Tan, and B. Boots. Fast and Efficient Locomotion via Learned Gait Transitions. Conference on Robot Learning (CoRL), 2021. arXiv:2104.04644.","identifier":"arXiv:2104.04644","url":"https://arxiv.org/abs/2104.04644"},{"text":"C. Li, T. Zhang, and D. I. Goldman. A Terradynamics of Legged Locomotion on Granular Media. Science, 339(6126):1408–1412, 2013.","identifier":"doi:10.1126/science.1229163","url":"https://doi.org/10.1126/science.1229163"},{"text":"T. Miki, L. Wellhausen, R. Grandia, F. Jenelten, T. Homberger, and M. Hutter. Elevation Mapping for Locomotion and Navigation Using GPU. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2022. arXiv:2204.12876.","identifier":"arXiv:2204.12876","url":"https://arxiv.org/abs/2204.12876"},{"text":"D. Hoeller, L. Wellhausen, F. Farshidian, and M. Hutter. Learning a State Representation and Navigation in Cluttered and Dynamic Environments. IEEE Robotics and Automation Letters, 6(3), 2021. arXiv:2103.04351.","identifier":"arXiv:2103.04351","url":"https://arxiv.org/abs/2103.04351"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2022. arXiv:2209.12827.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"J. Lee, M. Bjelonic, A. Reske, L. Wellhausen, T. Miki, and M. Hutter. Learning Robust Autonomous Navigation and Locomotion for Wheeled-Legged Robots. Science Robotics, 9(89), 2024. arXiv:2405.01792.","identifier":"arXiv:2405.01792","url":"https://arxiv.org/abs/2405.01792"},{"text":"Z. Fu, A. Kumar, A. Agarwal, H. Qi, J. Malik, and D. Pathak. Coupling Vision and Proprioception for Navigation of Legged Robots. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022. arXiv:2112.02094.","identifier":"arXiv:2112.02094","url":"https://arxiv.org/abs/2112.02094"},{"text":"F. Yang, C. Wang, C. Cadena, and M. Hutter. iPlanner: Imperative Path Planning. Robotics: Science and Systems (RSS), 2023. arXiv:2302.11434.","identifier":"arXiv:2302.11434","url":"https://arxiv.org/abs/2302.11434"},{"text":"P. Roth, J. Nubert, F. Yang, M. Mittal, and M. Hutter. ViPlanner: Visual Semantic Imperative Learning for Local Navigation. IEEE International Conference on Robotics and Automation (ICRA), 2024. arXiv:2310.00982.","identifier":"arXiv:2310.00982","url":"https://arxiv.org/abs/2310.00982"},{"text":"A. Loquercio, E. Kaufmann, R. Ranftl, M. Müller, V. Koltun, and D. Scaramuzza. Learning High-Speed Flight in the Wild. Science Robotics, 6(59), 2021. arXiv:2110.05113.","identifier":"arXiv:2110.05113","url":"https://arxiv.org/abs/2110.05113"},{"text":"D. Fox, W. Burgard, and S. Thrun. The Dynamic Window Approach to Collision Avoidance. IEEE Robotics and Automation Magazine, 4(1):23–33, 1997.","identifier":"doi:10.1109/100.580977","url":"https://doi.org/10.1109/100.580977"},{"text":"G. Williams, P. Drews, B. Goldfain, J. M. Rehg, and E. A. Theodorou. Information-Theoretic Model Predictive Control: Theory and Applications to Autonomous Driving. IEEE Transactions on Robotics, 34(6), 2018. arXiv:1707.02342.","identifier":"arXiv:1707.02342","url":"https://arxiv.org/abs/1707.02342"},{"text":"G. Kahn, P. Abbeel, and S. Levine. BADGR: An Autonomous Self-Supervised Learning-Based Navigation System. IEEE Robotics and Automation Letters, 6(2), 2021. arXiv:2002.05700.","identifier":"arXiv:2002.05700","url":"https://arxiv.org/abs/2002.05700"},{"text":"M. G. Castro, S. Triest, W. Wang, J. M. Gregory, F. Sanchez, J. G. Rogers III, and S. Scherer. How Does It Feel? Self-Supervised Costmap Learning for Off-Road Vehicle Traversability. IEEE International Conference on Robotics and Automation (ICRA), 2023. arXiv:2209.10788.","identifier":"arXiv:2209.10788","url":"https://arxiv.org/abs/2209.10788"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. Robotics: Science and Systems (RSS), 2023. arXiv:2305.08510.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"A. J. Sathyamoorthy, K. Weerakoon, T. Guan, J. Liang, and D. Manocha. TerraPN: Unstructured Terrain Navigation Using Online Self-Supervised Learning. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2022. arXiv:2202.12873.","identifier":"arXiv:2202.12873","url":"https://arxiv.org/abs/2202.12873"},{"text":"T. Guan, D. Kothandaraman, R. Chandra, A. J. Sathyamoorthy, K. Weerakoon, and D. Manocha. GA-Nav: Efficient Terrain Segmentation for Robot Navigation in Unstructured Outdoor Environments. IEEE Robotics and Automation Letters, 7(3), 2022. arXiv:2103.04233.","identifier":"arXiv:2103.04233","url":"https://arxiv.org/abs/2103.04233"},{"text":"D. D. Fan, K. Otsu, Y. Kubo, A. Dixit, J. Burdick, and A.-A. Agha-Mohammadi. STEP: Stochastic Traversability Evaluation and Planning for Risk-Aware Off-Road Navigation. Robotics: Science and Systems (RSS), 2021. arXiv:2103.02828.","identifier":"arXiv:2103.02828","url":"https://arxiv.org/abs/2103.02828"},{"text":"X. Cai et al. EVORA: Deep Evidential Traversability Learning for Risk-Aware Off-Road Autonomy. arXiv:2311.06234, 2023.","identifier":"arXiv:2311.06234","url":"https://arxiv.org/abs/2311.06234"},{"text":"R. T. Marler and J. S. Arora. Survey of Multi-Objective Optimization Methods for Engineering. Structural and Multidisciplinary Optimization, 26(6):369–395, 2004.","identifier":"doi:10.1007/s00158-003-0368-6","url":"https://doi.org/10.1007/s00158-003-0368-6"},{"text":"I. Das and J. E. Dennis. A Closer Look at Drawbacks of Minimizing Weighted Sums of Objectives for Pareto Set Generation in Multicriteria Optimization Problems. Structural Optimization, 14(1):63–69, 1997.","identifier":"doi:10.1007/BF01197559","url":"https://doi.org/10.1007/BF01197559"},{"text":"P. E. Hart, N. J. Nilsson, and B. Raphael. A Formal Basis for the Heuristic Determination of Minimum Cost Paths. IEEE Transactions on Systems Science and Cybernetics, 4(2):100–107, 1968.","identifier":"doi:10.1109/TSSC.1968.300136","url":"https://doi.org/10.1109/TSSC.1968.300136"},{"text":"V. Makoviychuk et al. Isaac Gym: High Performance GPU-Based Physics Simulation for Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal et al. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. IEEE Robotics and Automation Letters, 8(6), 2023. arXiv:2301.04195.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2017. arXiv:1703.06907.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"},{"text":"A. Kadian et al. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance? IEEE Robotics and Automation Letters, 5(4), 2020. arXiv:1912.06321.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2609.00001","slug":"2609.00001","registered_at":"2026-09-08T15:23:00Z","updated_at":"2026-09-08T18:23:00Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2609.00001","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2609.00001","tex_path":"articles/2609.00001/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-09-08T17:40:00Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2609.00001"},{"kagg_id":"KAGG:2608.00010","slug":"2608.00010","title":"Velocity Interface Contracts: Specifications and Verification for Legged Robot Middleware","authors":[{"name":"Clara Fischer","affiliation":"Robot Dynamics Group, ETH Zurich, Zurich, Switzerland"},{"name":"Diego Navarro","affiliation":"Robot Dynamics Group, ETH Zurich, Zurich, Switzerland"},{"name":"Amelie Laurent","affiliation":"Robot Dynamics Group, ETH Zurich, Zurich, Switzerland"}],"affiliation":null,"abstract":"Navigation stacks for legged robots rely on implicit assumptions about the velocity interface exposed by locomotion middleware: how quickly a command takes effect, where it saturates, and whether the response depends on gait mode or command history. These properties are rarely documented, and their violation surfaces as degraded tracking rather than as an explicit fault. This paper introduces velocity interface contracts, a typed declarative specification of command fidelity bounds, latency envelopes, saturation regions, and mode semantics, together with a verification tool that operates in two stages. Offline, the tool checks a calibrated interface model, obtained by planner-conditioned active calibration, against the contract and returns witness commands for each violated clause. Online, it monitors the command and odometry streams with lightweight sequential statistical tests and separates transient violation events from standing violations. The tool was evaluated on a Unitree B2 and an 8-DoF research quadruped over 10 days and 214 h of logged operation, with three known but undocumented B2 firmware behaviors seeded as detection targets. The model-based monitor recorded 11 true-positive detections, against 10 for fixed-threshold monitoring and for a monitor encoding the stack's implicit assumptions, at a false positive rate of 2.8 percent, a 95th-percentile runtime of 1.9 ms at 50 Hz, and a median time to detection of 3.4 s. It flagged the B2 speed limiter 3.4 s into its first engagement, whereas the fixed-threshold monitor required 41 s. The paper also specifies the contract format and proposes an adoption path for legged robot middleware.","publication_date":"2026-08-26","registered_at":"2026-09-02T11:31:00Z","updated_at":"2026-09-03T10:26:00Z","status":"published","html_path":null,"pdf_path":"articles/2608.00010/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"J. Di Carlo, P. M. Wensing, B. Katz, G. Bledt, and S. Kim. Dynamic Locomotion in the MIT Cheetah 3 Through Convex Model-Predictive Control. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2018.","identifier":"","url":""},{"text":"C. D. Bellicoso, F. Jenelten, C. Gehring, and M. Hutter. Dynamic Locomotion Through Online Nonlinear Motion Optimization for Quadrupedal Robots. IEEE Robotics and Automation Letters, 3(3):2261-2268, 2018.","identifier":"","url":""},{"text":"R. Grandia, F. Jenelten, S. Yang, F. Farshidian, and M. Hutter. Perceptive Locomotion Through Nonlinear Model-Predictive Control. arXiv:2208.08373, 2022.","identifier":"arXiv:2208.08373","url":"https://arxiv.org/abs/2208.08373"},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"M. Quigley, K. Conley, B. Gerkey, J. Faust, T. Foote, J. Leibs, R. Wheeler, and A. Y. Ng. ROS: An Open-Source Robot Operating System. ICRA Workshop on Open Source Software, 2009.","identifier":"","url":""},{"text":"S. Macenski, T. Foote, B. Gerkey, C. Lalancette, and W. Woodall. Robot Operating System 2: Design, Architecture, and Uses in the Wild. arXiv:2211.07752, 2022.","identifier":"arXiv:2211.07752","url":"https://arxiv.org/abs/2211.07752"},{"text":"S. Macenski, F. Martín, R. White, and J. Ginés Clavero. The Marathon 2: A Navigation System. arXiv:2003.00368, 2020.","identifier":"arXiv:2003.00368","url":"https://arxiv.org/abs/2003.00368"},{"text":"D. Fox, W. Burgard, and S. Thrun. The Dynamic Window Approach to Collision Avoidance. IEEE Robotics and Automation Magazine, 4(1):23-33, 1997.","identifier":"","url":""},{"text":"G. Williams, N. Wagener, B. Goldfain, P. Drews, J. M. Rehg, B. Boots, and E. A. Theodorou. Information Theoretic MPC for Model-Based Reinforcement Learning. IEEE International Conference on Robotics and Automation (ICRA), 2017.","identifier":"","url":""},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"T. Zang, H. Huang, S. Cheng, and W. Zhang. GAUGE: Planner-Conditioned Active Calibration of Opaque Quadruped Velocity Interfaces. arXiv:2607.02841, 2026.","identifier":"arXiv:2607.02841","url":"https://arxiv.org/abs/2607.02841"},{"text":"B. Meyer. Applying Design by Contract. IEEE Computer, 25(10):40-51, 1992.","identifier":"","url":""},{"text":"A. Benveniste, B. Caillaud, D. Nickovic, R. Passerone, J.-B. Raclet, P. Reinkemeier, A. Sangiovanni-Vincentelli, W. Damm, T. A. Henzinger, and K. G. Larsen. Contracts for System Design. Foundations and Trends in Electronic Design Automation, 12(2-3):124-400, 2018.","identifier":"","url":""},{"text":"A. Sangiovanni-Vincentelli, W. Damm, and R. Passerone. Taming Dr. Frankenstein: Contract-Based Design for Cyber-Physical Systems. European Journal of Control, 18(3):217-238, 2012.","identifier":"","url":""},{"text":"M. Leucker and C. Schallhart. A Brief Account of Runtime Verification. Journal of Logic and Algebraic Programming, 78(5):293-303, 2009.","identifier":"","url":""},{"text":"O. Maler and D. Nickovic. Monitoring Temporal Properties of Continuous Signals. International Conference on Formal Modeling and Analysis of Timed Systems (FORMATS), 2004.","identifier":"","url":""},{"text":"A. Donzé and O. Maler. Robust Satisfaction of Temporal Logic over Real-Valued Signals. International Conference on Formal Modeling and Analysis of Timed Systems (FORMATS), 2010.","identifier":"","url":""},{"text":"E. S. Page. Continuous Inspection Schemes. Biometrika, 41(1/2):100-115, 1954.","identifier":"","url":""},{"text":"A. Wald. Sequential Tests of Statistical Hypotheses. Annals of Mathematical Statistics, 16(2):117-186, 1945.","identifier":"","url":""},{"text":"M. Basseville and I. V. Nikiforov. Detection of Abrupt Changes: Theory and Application. Prentice Hall, 1993.","identifier":"","url":""},{"text":"M. Hutter, C. Gehring, D. Jud, A. Lauber, C. D. Bellicoso, V. Tsounis, J. Hwangbo, K. Bodie, P. Fankhauser, M. Bloesch, R. Diethelm, S. Bachmann, A. Melzer, and M. Hoepflinger. ANYmal - A Highly Mobile and Dynamic Quadrupedal Robot. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2016.","identifier":"","url":""},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"G. B. Margolis and P. Agrawal. Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior. arXiv:2212.03238, 2022.","identifier":"arXiv:2212.03238","url":"https://arxiv.org/abs/2212.03238"},{"text":"G. B. Margolis, G. Yang, K. Paigwar, T. Chen, and P. Agrawal. Rapid Locomotion via Reinforcement Learning. arXiv:2205.02824, 2022.","identifier":"arXiv:2205.02824","url":"https://arxiv.org/abs/2205.02824"},{"text":"J. Tan, T. Zhang, E. Coumans, A. Iscen, Y. Bai, D. Hafner, S. Bohez, and V. Vanhoucke. Sim-to-Real: Learning Agile Locomotion For Quadruped Robots. arXiv:1804.10332, 2018.","identifier":"arXiv:1804.10332","url":"https://arxiv.org/abs/1804.10332"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. arXiv:1710.06537, 2017.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"},{"text":"P. Fankhauser, M. Bloesch, and M. Hutter. Probabilistic Terrain Mapping for Mobile Robots with Uncertain Localization. IEEE Robotics and Automation Letters, 3(4):3019-3026, 2018.","identifier":"","url":""},{"text":"T. Miki, L. Wellhausen, R. Grandia, F. Jenelten, T. Homberger, and M. Hutter. Elevation Mapping for Locomotion and Navigation using GPU. arXiv:2204.12876, 2022.","identifier":"arXiv:2204.12876","url":"https://arxiv.org/abs/2204.12876"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. arXiv:2305.08510, 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"M. Wermelinger, P. Fankhauser, R. Diethelm, P. Krüsi, R. Siegwart, and M. Hutter. Navigation Planning for Legged Robots in Challenging Terrain. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2016.","identifier":"","url":""},{"text":"D. Hoeller, L. Wellhausen, F. Farshidian, and M. Hutter. Learning a State Representation and Navigation in Cluttered and Dynamic Environments. arXiv:2103.04351, 2021.","identifier":"arXiv:2103.04351","url":"https://arxiv.org/abs/2103.04351"},{"text":"Z. Fu, A. Kumar, A. Agarwal, H. Qi, J. Malik, and D. Pathak. Coupling Vision and Proprioception for Navigation of Legged Robots. arXiv:2112.02094, 2021.","identifier":"arXiv:2112.02094","url":"https://arxiv.org/abs/2112.02094"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"L. Ljung. System Identification: Theory for the User. 2nd edition, Prentice Hall, 1999.","identifier":"","url":""},{"text":"J. Huang, C. Erdogan, Y. Zhang, B. Moore, Q. Luo, A. Sundaresan, and G. Rosu. ROSRV: Runtime Verification for Robots. International Conference on Runtime Verification (RV), 2014.","identifier":"","url":""},{"text":"R. Isermann. Fault-Diagnosis Systems: An Introduction from Fault Detection to Fault Tolerance. Springer, 2006.","identifier":"","url":""},{"text":"A. D. Ames, X. Xu, J. W. Grizzle, and P. Tabuada. Control Barrier Function Based Quadratic Programs for Safety Critical Systems. arXiv:1609.06408, 2016.","identifier":"arXiv:1609.06408","url":"https://arxiv.org/abs/1609.06408"},{"text":"L. Brunke, M. Greeff, A. W. Hall, Z. Yuan, S. Zhou, J. Panerati, and A. P. Schoellig. Safe Learning in Robotics: From Learning-Based Control to Safe Reinforcement Learning. arXiv:2108.06266, 2021.","identifier":"arXiv:2108.06266","url":"https://arxiv.org/abs/2108.06266"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2608.00010","metadata":{"title":"Velocity Interface Contracts: Specifications and Verification for Legged Robot Middleware","authors":[{"name":"Clara Fischer","affiliation":"Robot Dynamics Group, ETH Zurich, Zurich, Switzerland"},{"name":"Diego Navarro","affiliation":"Robot Dynamics Group, ETH Zurich, Zurich, Switzerland"},{"name":"Amelie Laurent","affiliation":"Robot Dynamics Group, ETH Zurich, Zurich, Switzerland"}],"publication_date":"2026-08-26","abstract":"Navigation stacks for legged robots rely on implicit assumptions about the velocity interface exposed by locomotion middleware: how quickly a command takes effect, where it saturates, and whether the response depends on gait mode or command history. These properties are rarely documented, and their violation surfaces as degraded tracking rather than as an explicit fault. This paper introduces velocity interface contracts, a typed declarative specification of command fidelity bounds, latency envelopes, saturation regions, and mode semantics, together with a verification tool that operates in two stages. Offline, the tool checks a calibrated interface model, obtained by planner-conditioned active calibration, against the contract and returns witness commands for each violated clause. Online, it monitors the command and odometry streams with lightweight sequential statistical tests and separates transient violation events from standing violations. The tool was evaluated on a Unitree B2 and an 8-DoF research quadruped over 10 days and 214 h of logged operation, with three known but undocumented B2 firmware behaviors seeded as detection targets. The model-based monitor recorded 11 true-positive detections, against 10 for fixed-threshold monitoring and for a monitor encoding the stack's implicit assumptions, at a false positive rate of 2.8 percent, a 95th-percentile runtime of 1.9 ms at 50 Hz, and a median time to detection of 3.4 s. It flagged the B2 speed limiter 3.4 s into its first engagement, whereas the fixed-threshold monitor required 41 s. The paper also specifies the contract format and proposes an adoption path for legged robot middleware.","references":[{"text":"J. Di Carlo, P. M. Wensing, B. Katz, G. Bledt, and S. Kim. Dynamic Locomotion in the MIT Cheetah 3 Through Convex Model-Predictive Control. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2018.","identifier":"","url":""},{"text":"C. D. Bellicoso, F. Jenelten, C. Gehring, and M. Hutter. Dynamic Locomotion Through Online Nonlinear Motion Optimization for Quadrupedal Robots. IEEE Robotics and Automation Letters, 3(3):2261-2268, 2018.","identifier":"","url":""},{"text":"R. Grandia, F. Jenelten, S. Yang, F. Farshidian, and M. Hutter. Perceptive Locomotion Through Nonlinear Model-Predictive Control. arXiv:2208.08373, 2022.","identifier":"arXiv:2208.08373","url":"https://arxiv.org/abs/2208.08373"},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"M. Quigley, K. Conley, B. Gerkey, J. Faust, T. Foote, J. Leibs, R. Wheeler, and A. Y. Ng. ROS: An Open-Source Robot Operating System. ICRA Workshop on Open Source Software, 2009.","identifier":"","url":""},{"text":"S. Macenski, T. Foote, B. Gerkey, C. Lalancette, and W. Woodall. Robot Operating System 2: Design, Architecture, and Uses in the Wild. arXiv:2211.07752, 2022.","identifier":"arXiv:2211.07752","url":"https://arxiv.org/abs/2211.07752"},{"text":"S. Macenski, F. Martín, R. White, and J. Ginés Clavero. The Marathon 2: A Navigation System. arXiv:2003.00368, 2020.","identifier":"arXiv:2003.00368","url":"https://arxiv.org/abs/2003.00368"},{"text":"D. Fox, W. Burgard, and S. Thrun. The Dynamic Window Approach to Collision Avoidance. IEEE Robotics and Automation Magazine, 4(1):23-33, 1997.","identifier":"","url":""},{"text":"G. Williams, N. Wagener, B. Goldfain, P. Drews, J. M. Rehg, B. Boots, and E. A. Theodorou. Information Theoretic MPC for Model-Based Reinforcement Learning. IEEE International Conference on Robotics and Automation (ICRA), 2017.","identifier":"","url":""},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"T. Zang, H. Huang, S. Cheng, and W. Zhang. GAUGE: Planner-Conditioned Active Calibration of Opaque Quadruped Velocity Interfaces. arXiv:2607.02841, 2026.","identifier":"arXiv:2607.02841","url":"https://arxiv.org/abs/2607.02841"},{"text":"B. Meyer. Applying Design by Contract. IEEE Computer, 25(10):40-51, 1992.","identifier":"","url":""},{"text":"A. Benveniste, B. Caillaud, D. Nickovic, R. Passerone, J.-B. Raclet, P. Reinkemeier, A. Sangiovanni-Vincentelli, W. Damm, T. A. Henzinger, and K. G. Larsen. Contracts for System Design. Foundations and Trends in Electronic Design Automation, 12(2-3):124-400, 2018.","identifier":"","url":""},{"text":"A. Sangiovanni-Vincentelli, W. Damm, and R. Passerone. Taming Dr. Frankenstein: Contract-Based Design for Cyber-Physical Systems. European Journal of Control, 18(3):217-238, 2012.","identifier":"","url":""},{"text":"M. Leucker and C. Schallhart. A Brief Account of Runtime Verification. Journal of Logic and Algebraic Programming, 78(5):293-303, 2009.","identifier":"","url":""},{"text":"O. Maler and D. Nickovic. Monitoring Temporal Properties of Continuous Signals. International Conference on Formal Modeling and Analysis of Timed Systems (FORMATS), 2004.","identifier":"","url":""},{"text":"A. Donzé and O. Maler. Robust Satisfaction of Temporal Logic over Real-Valued Signals. International Conference on Formal Modeling and Analysis of Timed Systems (FORMATS), 2010.","identifier":"","url":""},{"text":"E. S. Page. Continuous Inspection Schemes. Biometrika, 41(1/2):100-115, 1954.","identifier":"","url":""},{"text":"A. Wald. Sequential Tests of Statistical Hypotheses. Annals of Mathematical Statistics, 16(2):117-186, 1945.","identifier":"","url":""},{"text":"M. Basseville and I. V. Nikiforov. Detection of Abrupt Changes: Theory and Application. Prentice Hall, 1993.","identifier":"","url":""},{"text":"M. Hutter, C. Gehring, D. Jud, A. Lauber, C. D. Bellicoso, V. Tsounis, J. Hwangbo, K. Bodie, P. Fankhauser, M. Bloesch, R. Diethelm, S. Bachmann, A. Melzer, and M. Hoepflinger. ANYmal - A Highly Mobile and Dynamic Quadrupedal Robot. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2016.","identifier":"","url":""},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"G. B. Margolis and P. Agrawal. Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior. arXiv:2212.03238, 2022.","identifier":"arXiv:2212.03238","url":"https://arxiv.org/abs/2212.03238"},{"text":"G. B. Margolis, G. Yang, K. Paigwar, T. Chen, and P. Agrawal. Rapid Locomotion via Reinforcement Learning. arXiv:2205.02824, 2022.","identifier":"arXiv:2205.02824","url":"https://arxiv.org/abs/2205.02824"},{"text":"J. Tan, T. Zhang, E. Coumans, A. Iscen, Y. Bai, D. Hafner, S. Bohez, and V. Vanhoucke. Sim-to-Real: Learning Agile Locomotion For Quadruped Robots. arXiv:1804.10332, 2018.","identifier":"arXiv:1804.10332","url":"https://arxiv.org/abs/1804.10332"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. arXiv:1710.06537, 2017.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"},{"text":"P. Fankhauser, M. Bloesch, and M. Hutter. Probabilistic Terrain Mapping for Mobile Robots with Uncertain Localization. IEEE Robotics and Automation Letters, 3(4):3019-3026, 2018.","identifier":"","url":""},{"text":"T. Miki, L. Wellhausen, R. Grandia, F. Jenelten, T. Homberger, and M. Hutter. Elevation Mapping for Locomotion and Navigation using GPU. arXiv:2204.12876, 2022.","identifier":"arXiv:2204.12876","url":"https://arxiv.org/abs/2204.12876"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. arXiv:2305.08510, 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"M. Wermelinger, P. Fankhauser, R. Diethelm, P. Krüsi, R. Siegwart, and M. Hutter. Navigation Planning for Legged Robots in Challenging Terrain. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2016.","identifier":"","url":""},{"text":"D. Hoeller, L. Wellhausen, F. Farshidian, and M. Hutter. Learning a State Representation and Navigation in Cluttered and Dynamic Environments. arXiv:2103.04351, 2021.","identifier":"arXiv:2103.04351","url":"https://arxiv.org/abs/2103.04351"},{"text":"Z. Fu, A. Kumar, A. Agarwal, H. Qi, J. Malik, and D. Pathak. Coupling Vision and Proprioception for Navigation of Legged Robots. arXiv:2112.02094, 2021.","identifier":"arXiv:2112.02094","url":"https://arxiv.org/abs/2112.02094"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"L. Ljung. System Identification: Theory for the User. 2nd edition, Prentice Hall, 1999.","identifier":"","url":""},{"text":"J. Huang, C. Erdogan, Y. Zhang, B. Moore, Q. Luo, A. Sundaresan, and G. Rosu. ROSRV: Runtime Verification for Robots. International Conference on Runtime Verification (RV), 2014.","identifier":"","url":""},{"text":"R. Isermann. Fault-Diagnosis Systems: An Introduction from Fault Detection to Fault Tolerance. Springer, 2006.","identifier":"","url":""},{"text":"A. D. Ames, X. Xu, J. W. Grizzle, and P. Tabuada. Control Barrier Function Based Quadratic Programs for Safety Critical Systems. arXiv:1609.06408, 2016.","identifier":"arXiv:1609.06408","url":"https://arxiv.org/abs/1609.06408"},{"text":"L. Brunke, M. Greeff, A. W. Hall, Z. Yuan, S. Zhou, J. Panerati, and A. P. Schoellig. Safe Learning in Robotics: From Learning-Based Control to Safe Reinforcement Learning. arXiv:2108.06266, 2021.","identifier":"arXiv:2108.06266","url":"https://arxiv.org/abs/2108.06266"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2608.00010","slug":"2608.00010","registered_at":"2026-09-02T11:31:00Z","updated_at":"2026-09-03T10:26:00Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2608.00010","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2608.00010","tex_path":"articles/2608.00010/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-09-02T13:48:00Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2608.00010"},{"kagg_id":"KAGG:2608.00009","slug":"2608.00009","title":"Uncertainty-Aware Local Planning for Quadrupeds in Perceptually Degraded Environments","authors":[{"name":"Jingyao Sun","affiliation":"Legged Locomotion Group, Tsinghua University, Beijing, China"},{"name":"Huiwen Cui","affiliation":"Legged Locomotion Group, Tsinghua University, Beijing, China"},{"name":"Zhaoyu Li","affiliation":"Legged Locomotion Group, Tsinghua University, Beijing, China"},{"name":"Bowen Huang","affiliation":"Legged Locomotion Group, Tsinghua University, Beijing, China"}],"affiliation":null,"abstract":"Airborne dust, fog, and low light corrupt the range measurements on which a legged robot's local planner depends, and they tend to do so precisely when reactive collision avoidance matters most. This paper presents an uncertainty-aware local planner for quadrupeds that couples the configuration-driven AgniNav planning backbone with a FutureRay-style predictor of future free range along control-aligned directions. For every candidate command, the predictor outputs a point estimate and an uncertainty envelope that is recalibrated by degradation-stratified conformal prediction. When the control-aligned envelope width exceeds a threshold, the planner widens its safety margin and caps the commanded speed; below the threshold it reproduces the nominal policy exactly, and every modulated quantity is Lipschitz in the excess uncertainty. The planner was evaluated on an ANYmal in Isaac Lab across 15 scenes with sweeps over dust opacity (0 to 80 percent), illuminance (50 to 800 lx), and range dropout (0 to 40 percent), and in a dusty warehouse bay and a dim parking structure. It reached a success rate of 87.9 percent, compared with 76.9 percent for the nominal planner and 82.5 percent for fixed 1.5x margin inflation, while raising mean speed from 1.82 to 1.94 m/s, extending collision-free distance from 3.4 to 3.8 km per collision, and retaining 99.2 percent clean-regime parity. Removing envelope calibration lowered success to 74.0 percent, below the nominal planner. Recorded failure cases, including a 9 s LiDAR dropout, illustrate both the mechanism and its limits.","publication_date":"2026-08-21","registered_at":"2026-08-24T17:23:00Z","updated_at":"2026-08-25T13:09:00Z","status":"published","html_path":null,"pdf_path":"articles/2608.00009/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. Science Robotics, 2019. arXiv:1901.08652.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. Science Robotics, 2020. arXiv:2010.11251.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. Science Robotics, 2022. arXiv:2201.08117.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"M. Tranzatto et al.. CERBERUS: Autonomous Legged and Aerial Robotic Exploration in the Tunnel and Urban Circuits of the DARPA Subterranean Challenge. Field Robotics, 2022. arXiv:2201.07067.","identifier":"arXiv:2201.07067","url":"https://arxiv.org/abs/2201.07067"},{"text":"A. Agha et al.. NeBula: Quest for Robotic Autonomy in Challenging Environments; TEAM CoSTAR at the DARPA Subterranean Challenge. arXiv:2103.11470, 2021.","identifier":"arXiv:2103.11470","url":"https://arxiv.org/abs/2103.11470"},{"text":"K. Ebadi et al.. Present and Future of SLAM in Extreme Environments: The DARPA SubT Challenge. arXiv:2208.01787, 2022.","identifier":"arXiv:2208.01787","url":"https://arxiv.org/abs/2208.01787"},{"text":"M. Bijelic, T. Gruber, F. Mannan, F. Kraus, W. Ritter, K. Dietmayer, and F. Heide. Seeing Through Fog Without Seeing Fog: Deep Multimodal Sensor Fusion in Unseen Adverse Weather. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020. arXiv:1902.08913.","identifier":"arXiv:1902.08913","url":"https://arxiv.org/abs/1902.08913"},{"text":"M. Hahner, C. Sakaridis, D. Dai, and L. Van Gool. Fog Simulation on Real LiDAR Point Clouds for 3D Object Detection in Adverse Weather. IEEE/CVF International Conference on Computer Vision (ICCV), 2021. arXiv:2108.05249.","identifier":"arXiv:2108.05249","url":"https://arxiv.org/abs/2108.05249"},{"text":"R. Heinzler, F. Piewak, P. Schindler, and W. Stork. CNN-based Lidar Point Cloud De-Noising in Adverse Weather. IEEE Robotics and Automation Letters, 2020. arXiv:1912.03874.","identifier":"arXiv:1912.03874","url":"https://arxiv.org/abs/1912.03874"},{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"T. Zang, S. Wang, H. Huang, and S. Cheng. FutureRay: Control-Aligned Future Range for Agile Quadruped Navigation. arXiv:2608.01197, 2026.","identifier":"arXiv:2608.01197","url":"https://arxiv.org/abs/2608.01197"},{"text":"A. Kendall and Y. Gal. What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision? Advances in Neural Information Processing Systems (NeurIPS), 2017. arXiv:1703.04977.","identifier":"arXiv:1703.04977","url":"https://arxiv.org/abs/1703.04977"},{"text":"B. Lakshminarayanan, A. Pritzel, and C. Blundell. Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles. Advances in Neural Information Processing Systems (NeurIPS), 2017. arXiv:1612.01474.","identifier":"arXiv:1612.01474","url":"https://arxiv.org/abs/1612.01474"},{"text":"C. Guo, G. Pleiss, Y. Sun, and K. Q. Weinberger. On Calibration of Modern Neural Networks. International Conference on Machine Learning (ICML), 2017. arXiv:1706.04599.","identifier":"arXiv:1706.04599","url":"https://arxiv.org/abs/1706.04599"},{"text":"V. Kuleshov, N. Fenner, and S. Ermon. Accurate Uncertainties for Deep Learning Using Calibrated Regression. International Conference on Machine Learning (ICML), 2018. arXiv:1807.00263.","identifier":"arXiv:1807.00263","url":"https://arxiv.org/abs/1807.00263"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Sridhar, D. Shah, C. Glossop, and S. Levine. NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration. arXiv:2310.07896, 2023.","identifier":"arXiv:2310.07896","url":"https://arxiv.org/abs/2310.07896"},{"text":"C. Cao, H. Zhu, F. Yang, Y. Xia, H. Choset, J. Oh, and J. Zhang. Autonomous Exploration Development Environment and the Planning Algorithms. IEEE International Conference on Robotics and Automation (ICRA), 2022. arXiv:2110.14573.","identifier":"arXiv:2110.14573","url":"https://arxiv.org/abs/2110.14573"},{"text":"X. Zhou, Z. Wang, H. Ye, C. Xu, and F. Gao. EGO-Planner: An ESDF-free Gradient-based Local Planner for Quadrotors. IEEE Robotics and Automation Letters, 2021. arXiv:2008.08835.","identifier":"arXiv:2008.08835","url":"https://arxiv.org/abs/2008.08835"},{"text":"J. Hu, S. Yuan, G. C. R. Bethala, A. Tzes, and Y. Fang. Learning Adaptive Safety Margins for Visual Navigation. arXiv:2607.18200, 2026.","identifier":"arXiv:2607.18200","url":"https://arxiv.org/abs/2607.18200"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. Conference on Robot Learning (CoRL), 2021. arXiv:2109.11978.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2022. arXiv:2209.12827.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. Science Robotics, 2024. arXiv:2306.14874.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"L. Wellhausen and M. Hutter. ArtPlanner: Robust Legged Robot Navigation in the Field. Field Robotics, 2023. arXiv:2303.01420.","identifier":"arXiv:2303.01420","url":"https://arxiv.org/abs/2303.01420"},{"text":"R. Grandia, F. Jenelten, S. Yang, F. Farshidian, and M. Hutter. Perceptive Locomotion Through Nonlinear Model-Predictive Control. IEEE Transactions on Robotics, 2023. arXiv:2208.08373.","identifier":"arXiv:2208.08373","url":"https://arxiv.org/abs/2208.08373"},{"text":"T. Miki, L. Wellhausen, R. Grandia, F. Jenelten, T. Homberger, and M. Hutter. Elevation Mapping for Locomotion and Navigation using GPU. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2022. arXiv:2204.12876.","identifier":"arXiv:2204.12876","url":"https://arxiv.org/abs/2204.12876"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. Robotics: Science and Systems (RSS), 2023. arXiv:2305.08510.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"M. Hahner, C. Sakaridis, M. Bijelic, F. Heide, F. Yu, D. Dai, and L. Van Gool. LiDAR Snowfall Simulation for Robust 3D Object Detection. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022. arXiv:2203.15118.","identifier":"arXiv:2203.15118","url":"https://arxiv.org/abs/2203.15118"},{"text":"S. Zhao, H. Zhang, P. Wang, L. Nogueira, and S. Scherer. Super Odometry: IMU-centric LiDAR-Visual-Inertial Estimator for Challenging Environments. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2021. arXiv:2104.14938.","identifier":"arXiv:2104.14938","url":"https://arxiv.org/abs/2104.14938"},{"text":"D. D. Fan, K. Otsu, Y. Kubo, A. Dixit, J. Burdick, and A.-A. Agha-Mohammadi. STEP: Stochastic Traversability Evaluation and Planning for Risk-Aware Off-road Navigation. Robotics: Science and Systems (RSS), 2021. arXiv:2103.02828.","identifier":"arXiv:2103.02828","url":"https://arxiv.org/abs/2103.02828"},{"text":"Y. Gal and Z. Ghahramani. Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning. International Conference on Machine Learning (ICML), 2016. arXiv:1506.02142.","identifier":"arXiv:1506.02142","url":"https://arxiv.org/abs/1506.02142"},{"text":"A. N. Angelopoulos and S. Bates. A Gentle Introduction to Conformal Prediction and Distribution-Free Uncertainty Quantification. arXiv:2107.07511, 2021.","identifier":"arXiv:2107.07511","url":"https://arxiv.org/abs/2107.07511"},{"text":"Y. Romano, E. Patterson, and E. J. Candès. Conformalized Quantile Regression. Advances in Neural Information Processing Systems (NeurIPS), 2019. arXiv:1905.03222.","identifier":"arXiv:1905.03222","url":"https://arxiv.org/abs/1905.03222"},{"text":"L. Lindemann, M. Cleaveland, G. Shim, and G. J. Pappas. Safe Planning in Dynamic Environments using Conformal Prediction. IEEE Robotics and Automation Letters, 2023. arXiv:2210.10254.","identifier":"arXiv:2210.10254","url":"https://arxiv.org/abs/2210.10254"},{"text":"G. Kahn, A. Villaflor, V. Pong, P. Abbeel, and S. Levine. Uncertainty-Aware Reinforcement Learning for Collision Avoidance. arXiv:1702.01182, 2017.","identifier":"arXiv:1702.01182","url":"https://arxiv.org/abs/1702.01182"},{"text":"B. Lütjens, M. Everett, and J. P. How. Safe Reinforcement Learning with Model Uncertainty Estimates. IEEE International Conference on Robotics and Automation (ICRA), 2019. arXiv:1810.08700.","identifier":"arXiv:1810.08700","url":"https://arxiv.org/abs/1810.08700"},{"text":"A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada. Control Barrier Functions: Theory and Applications. European Control Conference (ECC), 2019. arXiv:1903.11199.","identifier":"arXiv:1903.11199","url":"https://arxiv.org/abs/1903.11199"},{"text":"K. P. Wabersich and M. N. Zeilinger. A Predictive Safety Filter for Learning-Based Control of Constrained Nonlinear Dynamical Systems. Automatica, 2021. arXiv:1812.05506.","identifier":"arXiv:1812.05506","url":"https://arxiv.org/abs/1812.05506"},{"text":"K.-C. Hsu, H. Hu, and J. F. Fisac. The Safety Filter: A Unified View of Safety-Critical Control in Autonomous Systems. arXiv:2309.05837, 2023.","identifier":"arXiv:2309.05837","url":"https://arxiv.org/abs/2309.05837"},{"text":"V. Makoviychuk et al.. Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning. NeurIPS Datasets and Benchmarks Track, 2021. arXiv:2108.10470.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal et al.. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. IEEE Robotics and Automation Letters, 2023. arXiv:2301.04195.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2608.00009","metadata":{"title":"Uncertainty-Aware Local Planning for Quadrupeds in Perceptually Degraded Environments","authors":[{"name":"Jingyao Sun","affiliation":"Legged Locomotion Group, Tsinghua University, Beijing, China"},{"name":"Huiwen Cui","affiliation":"Legged Locomotion Group, Tsinghua University, Beijing, China"},{"name":"Zhaoyu Li","affiliation":"Legged Locomotion Group, Tsinghua University, Beijing, China"},{"name":"Bowen Huang","affiliation":"Legged Locomotion Group, Tsinghua University, Beijing, China"}],"publication_date":"2026-08-21","abstract":"Airborne dust, fog, and low light corrupt the range measurements on which a legged robot's local planner depends, and they tend to do so precisely when reactive collision avoidance matters most. This paper presents an uncertainty-aware local planner for quadrupeds that couples the configuration-driven AgniNav planning backbone with a FutureRay-style predictor of future free range along control-aligned directions. For every candidate command, the predictor outputs a point estimate and an uncertainty envelope that is recalibrated by degradation-stratified conformal prediction. When the control-aligned envelope width exceeds a threshold, the planner widens its safety margin and caps the commanded speed; below the threshold it reproduces the nominal policy exactly, and every modulated quantity is Lipschitz in the excess uncertainty. The planner was evaluated on an ANYmal in Isaac Lab across 15 scenes with sweeps over dust opacity (0 to 80 percent), illuminance (50 to 800 lx), and range dropout (0 to 40 percent), and in a dusty warehouse bay and a dim parking structure. It reached a success rate of 87.9 percent, compared with 76.9 percent for the nominal planner and 82.5 percent for fixed 1.5x margin inflation, while raising mean speed from 1.82 to 1.94 m/s, extending collision-free distance from 3.4 to 3.8 km per collision, and retaining 99.2 percent clean-regime parity. Removing envelope calibration lowered success to 74.0 percent, below the nominal planner. Recorded failure cases, including a 9 s LiDAR dropout, illustrate both the mechanism and its limits.","references":[{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. Science Robotics, 2019. arXiv:1901.08652.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. Science Robotics, 2020. arXiv:2010.11251.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. Science Robotics, 2022. arXiv:2201.08117.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"M. Tranzatto et al.. CERBERUS: Autonomous Legged and Aerial Robotic Exploration in the Tunnel and Urban Circuits of the DARPA Subterranean Challenge. Field Robotics, 2022. arXiv:2201.07067.","identifier":"arXiv:2201.07067","url":"https://arxiv.org/abs/2201.07067"},{"text":"A. Agha et al.. NeBula: Quest for Robotic Autonomy in Challenging Environments; TEAM CoSTAR at the DARPA Subterranean Challenge. arXiv:2103.11470, 2021.","identifier":"arXiv:2103.11470","url":"https://arxiv.org/abs/2103.11470"},{"text":"K. Ebadi et al.. Present and Future of SLAM in Extreme Environments: The DARPA SubT Challenge. arXiv:2208.01787, 2022.","identifier":"arXiv:2208.01787","url":"https://arxiv.org/abs/2208.01787"},{"text":"M. Bijelic, T. Gruber, F. Mannan, F. Kraus, W. Ritter, K. Dietmayer, and F. Heide. Seeing Through Fog Without Seeing Fog: Deep Multimodal Sensor Fusion in Unseen Adverse Weather. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020. arXiv:1902.08913.","identifier":"arXiv:1902.08913","url":"https://arxiv.org/abs/1902.08913"},{"text":"M. Hahner, C. Sakaridis, D. Dai, and L. Van Gool. Fog Simulation on Real LiDAR Point Clouds for 3D Object Detection in Adverse Weather. IEEE/CVF International Conference on Computer Vision (ICCV), 2021. arXiv:2108.05249.","identifier":"arXiv:2108.05249","url":"https://arxiv.org/abs/2108.05249"},{"text":"R. Heinzler, F. Piewak, P. Schindler, and W. Stork. CNN-based Lidar Point Cloud De-Noising in Adverse Weather. IEEE Robotics and Automation Letters, 2020. arXiv:1912.03874.","identifier":"arXiv:1912.03874","url":"https://arxiv.org/abs/1912.03874"},{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"T. Zang, S. Wang, H. Huang, and S. Cheng. FutureRay: Control-Aligned Future Range for Agile Quadruped Navigation. arXiv:2608.01197, 2026.","identifier":"arXiv:2608.01197","url":"https://arxiv.org/abs/2608.01197"},{"text":"A. Kendall and Y. Gal. What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision? Advances in Neural Information Processing Systems (NeurIPS), 2017. arXiv:1703.04977.","identifier":"arXiv:1703.04977","url":"https://arxiv.org/abs/1703.04977"},{"text":"B. Lakshminarayanan, A. Pritzel, and C. Blundell. Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles. Advances in Neural Information Processing Systems (NeurIPS), 2017. arXiv:1612.01474.","identifier":"arXiv:1612.01474","url":"https://arxiv.org/abs/1612.01474"},{"text":"C. Guo, G. Pleiss, Y. Sun, and K. Q. Weinberger. On Calibration of Modern Neural Networks. International Conference on Machine Learning (ICML), 2017. arXiv:1706.04599.","identifier":"arXiv:1706.04599","url":"https://arxiv.org/abs/1706.04599"},{"text":"V. Kuleshov, N. Fenner, and S. Ermon. Accurate Uncertainties for Deep Learning Using Calibrated Regression. International Conference on Machine Learning (ICML), 2018. arXiv:1807.00263.","identifier":"arXiv:1807.00263","url":"https://arxiv.org/abs/1807.00263"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Sridhar, D. Shah, C. Glossop, and S. Levine. NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration. arXiv:2310.07896, 2023.","identifier":"arXiv:2310.07896","url":"https://arxiv.org/abs/2310.07896"},{"text":"C. Cao, H. Zhu, F. Yang, Y. Xia, H. Choset, J. Oh, and J. Zhang. Autonomous Exploration Development Environment and the Planning Algorithms. IEEE International Conference on Robotics and Automation (ICRA), 2022. arXiv:2110.14573.","identifier":"arXiv:2110.14573","url":"https://arxiv.org/abs/2110.14573"},{"text":"X. Zhou, Z. Wang, H. Ye, C. Xu, and F. Gao. EGO-Planner: An ESDF-free Gradient-based Local Planner for Quadrotors. IEEE Robotics and Automation Letters, 2021. arXiv:2008.08835.","identifier":"arXiv:2008.08835","url":"https://arxiv.org/abs/2008.08835"},{"text":"J. Hu, S. Yuan, G. C. R. Bethala, A. Tzes, and Y. Fang. Learning Adaptive Safety Margins for Visual Navigation. arXiv:2607.18200, 2026.","identifier":"arXiv:2607.18200","url":"https://arxiv.org/abs/2607.18200"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. Conference on Robot Learning (CoRL), 2021. arXiv:2109.11978.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2022. arXiv:2209.12827.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. Science Robotics, 2024. arXiv:2306.14874.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"L. Wellhausen and M. Hutter. ArtPlanner: Robust Legged Robot Navigation in the Field. Field Robotics, 2023. arXiv:2303.01420.","identifier":"arXiv:2303.01420","url":"https://arxiv.org/abs/2303.01420"},{"text":"R. Grandia, F. Jenelten, S. Yang, F. Farshidian, and M. Hutter. Perceptive Locomotion Through Nonlinear Model-Predictive Control. IEEE Transactions on Robotics, 2023. arXiv:2208.08373.","identifier":"arXiv:2208.08373","url":"https://arxiv.org/abs/2208.08373"},{"text":"T. Miki, L. Wellhausen, R. Grandia, F. Jenelten, T. Homberger, and M. Hutter. Elevation Mapping for Locomotion and Navigation using GPU. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2022. arXiv:2204.12876.","identifier":"arXiv:2204.12876","url":"https://arxiv.org/abs/2204.12876"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. Robotics: Science and Systems (RSS), 2023. arXiv:2305.08510.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"M. Hahner, C. Sakaridis, M. Bijelic, F. Heide, F. Yu, D. Dai, and L. Van Gool. LiDAR Snowfall Simulation for Robust 3D Object Detection. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022. arXiv:2203.15118.","identifier":"arXiv:2203.15118","url":"https://arxiv.org/abs/2203.15118"},{"text":"S. Zhao, H. Zhang, P. Wang, L. Nogueira, and S. Scherer. Super Odometry: IMU-centric LiDAR-Visual-Inertial Estimator for Challenging Environments. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2021. arXiv:2104.14938.","identifier":"arXiv:2104.14938","url":"https://arxiv.org/abs/2104.14938"},{"text":"D. D. Fan, K. Otsu, Y. Kubo, A. Dixit, J. Burdick, and A.-A. Agha-Mohammadi. STEP: Stochastic Traversability Evaluation and Planning for Risk-Aware Off-road Navigation. Robotics: Science and Systems (RSS), 2021. arXiv:2103.02828.","identifier":"arXiv:2103.02828","url":"https://arxiv.org/abs/2103.02828"},{"text":"Y. Gal and Z. Ghahramani. Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning. International Conference on Machine Learning (ICML), 2016. arXiv:1506.02142.","identifier":"arXiv:1506.02142","url":"https://arxiv.org/abs/1506.02142"},{"text":"A. N. Angelopoulos and S. Bates. A Gentle Introduction to Conformal Prediction and Distribution-Free Uncertainty Quantification. arXiv:2107.07511, 2021.","identifier":"arXiv:2107.07511","url":"https://arxiv.org/abs/2107.07511"},{"text":"Y. Romano, E. Patterson, and E. J. Candès. Conformalized Quantile Regression. Advances in Neural Information Processing Systems (NeurIPS), 2019. arXiv:1905.03222.","identifier":"arXiv:1905.03222","url":"https://arxiv.org/abs/1905.03222"},{"text":"L. Lindemann, M. Cleaveland, G. Shim, and G. J. Pappas. Safe Planning in Dynamic Environments using Conformal Prediction. IEEE Robotics and Automation Letters, 2023. arXiv:2210.10254.","identifier":"arXiv:2210.10254","url":"https://arxiv.org/abs/2210.10254"},{"text":"G. Kahn, A. Villaflor, V. Pong, P. Abbeel, and S. Levine. Uncertainty-Aware Reinforcement Learning for Collision Avoidance. arXiv:1702.01182, 2017.","identifier":"arXiv:1702.01182","url":"https://arxiv.org/abs/1702.01182"},{"text":"B. Lütjens, M. Everett, and J. P. How. Safe Reinforcement Learning with Model Uncertainty Estimates. IEEE International Conference on Robotics and Automation (ICRA), 2019. arXiv:1810.08700.","identifier":"arXiv:1810.08700","url":"https://arxiv.org/abs/1810.08700"},{"text":"A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada. Control Barrier Functions: Theory and Applications. European Control Conference (ECC), 2019. arXiv:1903.11199.","identifier":"arXiv:1903.11199","url":"https://arxiv.org/abs/1903.11199"},{"text":"K. P. Wabersich and M. N. Zeilinger. A Predictive Safety Filter for Learning-Based Control of Constrained Nonlinear Dynamical Systems. Automatica, 2021. arXiv:1812.05506.","identifier":"arXiv:1812.05506","url":"https://arxiv.org/abs/1812.05506"},{"text":"K.-C. Hsu, H. Hu, and J. F. Fisac. The Safety Filter: A Unified View of Safety-Critical Control in Autonomous Systems. arXiv:2309.05837, 2023.","identifier":"arXiv:2309.05837","url":"https://arxiv.org/abs/2309.05837"},{"text":"V. Makoviychuk et al.. Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning. NeurIPS Datasets and Benchmarks Track, 2021. arXiv:2108.10470.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal et al.. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. IEEE Robotics and Automation Letters, 2023. arXiv:2301.04195.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2608.00009","slug":"2608.00009","registered_at":"2026-08-24T17:23:00Z","updated_at":"2026-08-25T13:09:00Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2608.00009","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2608.00009","tex_path":"articles/2608.00009/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-08-24T19:40:00Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2608.00009"},{"kagg_id":"KAGG:2608.00008","slug":"2608.00008","title":"Probing Opaque Velocity Interfaces on Production Quadruped Platforms: A Field Study","authors":[{"name":"Andrew Foster","affiliation":"Advanced Robotics Laboratory, Boston Dynamics, Waltham, MA, USA"},{"name":"Lena Vasquez","affiliation":"Advanced Robotics Laboratory, Boston Dynamics, Waltham, MA, USA"},{"name":"Rajiv Menon","affiliation":"Advanced Robotics Laboratory, Boston Dynamics, Waltham, MA, USA"}],"affiliation":null,"abstract":"Production quadruped platforms expose velocity command interfaces whose internals (gait managers, safety layers, terrain modes) are undocumented, yet navigation planners routinely assume that commanded velocities are tracked instantly and uniformly. This paper reports a 14-day field characterization of three commercial quadruped platforms, two 70-class research units and one 45-class work unit, across laboratory, campus-path, and light outdoor terrain using GAUGE-style planner-conditioned active calibration. Each platform received 60 probe sessions (20 per terrain mode) logged at 100 Hz, with 0, 5, and 10 kg payloads rotated across sessions. We report command fidelity, 95th-percentile phase lag, saturation-envelope error, and mode-switch transients. On the primary unit with the full probe suite, the command fidelity RMS error is 9.7 percent with a p95 phase lag of 71 ms, compared with 11.1 percent and 79 ms on Platform B and 10.6 percent and 81 ms on Platform C. Restricting probes to the default mode raises the error to 10.5 percent, and removing disturbance rejection raises it to 10.7 percent. Payload is the dominant degradation factor: from 0 to 10 kg, the fidelity error rises from 8.9 to 12.6 percent and the phase lag from 64 to 88 ms. Probes also exposed three behaviors that break naive planning assumptions: an undocumented 3.2 m/s speed limiter in an outdoor mode, a 0.7 s gait-manager hysteresis that produced 0.4 m overshoot, and an unannounced careful mode under a 10 kg payload that halves the usable envelope. We derive interface assumptions that planners for such platforms should state explicitly.","publication_date":"2026-08-18","registered_at":"2026-08-21T08:45:00Z","updated_at":"2026-08-23T17:57:00Z","status":"published","html_path":null,"pdf_path":"articles/2608.00008/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"G. Williams, A. Aldrich, and E. Theodorou. Model Predictive Path Integral Control using Covariance Variable Importance Sampling. arXiv:1509.01149, 2015.","identifier":"arXiv:1509.01149","url":"https://arxiv.org/abs/1509.01149"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"T. Zang, H. Huang, S. Cheng, and W. Zhang. GAUGE: Planner-Conditioned Active Calibration of Opaque Quadruped Velocity Interfaces. arXiv:2607.02841, 2026.","identifier":"arXiv:2607.02841","url":"https://arxiv.org/abs/2607.02841"},{"text":"W. Yu, J. Tan, C. K. Liu, and G. Turk. Preparing for the Unknown: Learning a Universal Policy with Online System Identification. arXiv:1702.02453, 2017.","identifier":"arXiv:1702.02453","url":"https://arxiv.org/abs/1702.02453"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"A. Nagabandi, I. Clavera, S. Liu, R. S. Fearing, P. Abbeel, S. Levine, and C. Finn. Learning to Adapt in Dynamic, Real-World Environments Through Meta-Reinforcement Learning. arXiv:1803.11347, 2018.","identifier":"arXiv:1803.11347","url":"https://arxiv.org/abs/1803.11347"},{"text":"K. Chua, R. Calandra, R. McAllister, and S. Levine. Deep Reinforcement Learning in a Handful of Trials using Probabilistic Dynamics Models. arXiv:1805.12114, 2018.","identifier":"arXiv:1805.12114","url":"https://arxiv.org/abs/1805.12114"},{"text":"T. Lew, A. Sharma, J. Harrison, A. Bylard, and M. Pavone. Safe Active Dynamics Learning and Control: A Sequential Exploration-Exploitation Framework. arXiv:2008.11700, 2020.","identifier":"arXiv:2008.11700","url":"https://arxiv.org/abs/2008.11700"},{"text":"T. Koller, F. Berkenkamp, M. Turchetta, and A. Krause. Learning-Based Model Predictive Control for Safe Exploration. arXiv:1803.08287, 2018.","identifier":"arXiv:1803.08287","url":"https://arxiv.org/abs/1803.08287"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. arXiv:1703.06907, 2017.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. arXiv:1710.06537, 2017.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"},{"text":"J. Tan, T. Zhang, E. Coumans, A. Iscen, Y. Bai, D. Hafner, S. Bohez, and V. Vanhoucke. Sim-to-Real: Learning Agile Locomotion For Quadruped Robots. arXiv:1804.10332, 2018.","identifier":"arXiv:1804.10332","url":"https://arxiv.org/abs/1804.10332"},{"text":"Z. Xie, X. Da, M. van de Panne, B. Babich, and A. Garg. Dynamics Randomization Revisited: A Case Study for Quadrupedal Locomotion. arXiv:2011.02404, 2020.","identifier":"arXiv:2011.02404","url":"https://arxiv.org/abs/2011.02404"},{"text":"Y. Chebotar, A. Handa, V. Makoviychuk, M. Macklin, J. Issac, N. Ratliff, and D. Fox. Closing the Sim-to-Real Loop: Adapting Simulation Randomization with Real World Experience. arXiv:1810.05687, 2018.","identifier":"arXiv:1810.05687","url":"https://arxiv.org/abs/1810.05687"},{"text":"F. Ramos, R. C. Possas, and D. Fox. BayesSim: Adaptive Domain Randomization via Probabilistic Inference for Robotics Simulators. arXiv:1906.01728, 2019.","identifier":"arXiv:1906.01728","url":"https://arxiv.org/abs/1906.01728"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"G. B. Margolis, G. Yang, K. Paigwar, T. Chen, and P. Agrawal. Rapid Locomotion via Reinforcement Learning. arXiv:2205.02824, 2022.","identifier":"arXiv:2205.02824","url":"https://arxiv.org/abs/2205.02824"},{"text":"G. B. Margolis and P. Agrawal. Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior. arXiv:2212.03238, 2022.","identifier":"arXiv:2212.03238","url":"https://arxiv.org/abs/2212.03238"},{"text":"Y. Yang, T. Zhang, E. Coumans, J. Tan, and B. Boots. Fast and Efficient Locomotion via Learned Gait Transitions. arXiv:2104.04644, 2021.","identifier":"arXiv:2104.04644","url":"https://arxiv.org/abs/2104.04644"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. arXiv:2211.07638, 2022.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. arXiv:2309.14341, 2023.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"X. Xiao, B. Liu, G. Warnell, and P. Stone. Motion Planning and Control for Mobile Robot Navigation Using Machine Learning: a Survey. arXiv:2011.13112, 2020.","identifier":"arXiv:2011.13112","url":"https://arxiv.org/abs/2011.13112"},{"text":"X. Xiao, J. Biswas, and P. Stone. Learning Inverse Kinodynamics for Accurate High-Speed Off-Road Navigation on Unstructured Terrain. arXiv:2102.12667, 2021.","identifier":"arXiv:2102.12667","url":"https://arxiv.org/abs/2102.12667"},{"text":"H. Karnan, K. S. Sikand, P. Atreya, S. Rabiee, X. Xiao, G. Warnell, P. Stone, and J. Biswas. VI-IKD: High-Speed Accurate Off-Road Navigation using Learned Visual-Inertial Inverse Kinodynamics. arXiv:2203.15983, 2022.","identifier":"arXiv:2203.15983","url":"https://arxiv.org/abs/2203.15983"},{"text":"G. Kahn, P. Abbeel, and S. Levine. BADGR: An Autonomous Self-Supervised Learning-Based Navigation System. arXiv:2002.05700, 2020.","identifier":"arXiv:2002.05700","url":"https://arxiv.org/abs/2002.05700"},{"text":"Z. Fu, A. Kumar, A. Agarwal, H. Qi, J. Malik, and D. Pathak. Coupling Vision and Proprioception for Navigation of Legged Robots. arXiv:2112.02094, 2021.","identifier":"arXiv:2112.02094","url":"https://arxiv.org/abs/2112.02094"},{"text":"J. Truong, D. Yarats, T. Li, F. Meier, S. Chernova, D. Batra, and A. Rai. Learning Navigation Skills for Legged Robots with Learned Robot Embeddings. arXiv:2011.12255, 2020.","identifier":"arXiv:2011.12255","url":"https://arxiv.org/abs/2011.12255"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. arXiv:2305.08510, 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"T. Miki, L. Wellhausen, R. Grandia, F. Jenelten, T. Homberger, and M. Hutter. Elevation Mapping for Locomotion and Navigation using GPU. arXiv:2204.12876, 2022.","identifier":"arXiv:2204.12876","url":"https://arxiv.org/abs/2204.12876"},{"text":"J. Lee, M. Bjelonic, A. Reske, L. Wellhausen, T. Miki, and M. Hutter. Learning Robust Autonomous Navigation and Locomotion for Wheeled-Legged Robots. arXiv:2405.01792, 2024.","identifier":"arXiv:2405.01792","url":"https://arxiv.org/abs/2405.01792"},{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"N. Yokoyama, S. Ha, and D. Batra. Success Weighted by Completion Time: A Dynamics-Aware Evaluation Criteria for Embodied Navigation. arXiv:2103.08022, 2021.","identifier":"arXiv:2103.08022","url":"https://arxiv.org/abs/2103.08022"},{"text":"A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada. Control Barrier Functions: Theory and Applications. arXiv:1903.11199, 2019.","identifier":"arXiv:1903.11199","url":"https://arxiv.org/abs/1903.11199"},{"text":"S. Bansal, M. Chen, S. Herbert, and C. J. Tomlin. Hamilton-Jacobi Reachability: A Brief Overview and Recent Advances. arXiv:1709.07523, 2017.","identifier":"arXiv:1709.07523","url":"https://arxiv.org/abs/1709.07523"},{"text":"L. Brunke, M. Greeff, A. W. Hall, Z. Yuan, S. Zhou, J. Panerati, and A. P. Schoellig. Safe Learning in Robotics: From Learning-Based Control to Safe Reinforcement Learning. arXiv:2108.06266, 2021.","identifier":"arXiv:2108.06266","url":"https://arxiv.org/abs/2108.06266"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2608.00008","metadata":{"title":"Probing Opaque Velocity Interfaces on Production Quadruped Platforms: A Field Study","authors":[{"name":"Andrew Foster","affiliation":"Advanced Robotics Laboratory, Boston Dynamics, Waltham, MA, USA"},{"name":"Lena Vasquez","affiliation":"Advanced Robotics Laboratory, Boston Dynamics, Waltham, MA, USA"},{"name":"Rajiv Menon","affiliation":"Advanced Robotics Laboratory, Boston Dynamics, Waltham, MA, USA"}],"publication_date":"2026-08-18","abstract":"Production quadruped platforms expose velocity command interfaces whose internals (gait managers, safety layers, terrain modes) are undocumented, yet navigation planners routinely assume that commanded velocities are tracked instantly and uniformly. This paper reports a 14-day field characterization of three commercial quadruped platforms, two 70-class research units and one 45-class work unit, across laboratory, campus-path, and light outdoor terrain using GAUGE-style planner-conditioned active calibration. Each platform received 60 probe sessions (20 per terrain mode) logged at 100 Hz, with 0, 5, and 10 kg payloads rotated across sessions. We report command fidelity, 95th-percentile phase lag, saturation-envelope error, and mode-switch transients. On the primary unit with the full probe suite, the command fidelity RMS error is 9.7 percent with a p95 phase lag of 71 ms, compared with 11.1 percent and 79 ms on Platform B and 10.6 percent and 81 ms on Platform C. Restricting probes to the default mode raises the error to 10.5 percent, and removing disturbance rejection raises it to 10.7 percent. Payload is the dominant degradation factor: from 0 to 10 kg, the fidelity error rises from 8.9 to 12.6 percent and the phase lag from 64 to 88 ms. Probes also exposed three behaviors that break naive planning assumptions: an undocumented 3.2 m/s speed limiter in an outdoor mode, a 0.7 s gait-manager hysteresis that produced 0.4 m overshoot, and an unannounced careful mode under a 10 kg payload that halves the usable envelope. We derive interface assumptions that planners for such platforms should state explicitly.","references":[{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"G. Williams, A. Aldrich, and E. Theodorou. Model Predictive Path Integral Control using Covariance Variable Importance Sampling. arXiv:1509.01149, 2015.","identifier":"arXiv:1509.01149","url":"https://arxiv.org/abs/1509.01149"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"T. Zang, H. Huang, S. Cheng, and W. Zhang. GAUGE: Planner-Conditioned Active Calibration of Opaque Quadruped Velocity Interfaces. arXiv:2607.02841, 2026.","identifier":"arXiv:2607.02841","url":"https://arxiv.org/abs/2607.02841"},{"text":"W. Yu, J. Tan, C. K. Liu, and G. Turk. Preparing for the Unknown: Learning a Universal Policy with Online System Identification. arXiv:1702.02453, 2017.","identifier":"arXiv:1702.02453","url":"https://arxiv.org/abs/1702.02453"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"A. Nagabandi, I. Clavera, S. Liu, R. S. Fearing, P. Abbeel, S. Levine, and C. Finn. Learning to Adapt in Dynamic, Real-World Environments Through Meta-Reinforcement Learning. arXiv:1803.11347, 2018.","identifier":"arXiv:1803.11347","url":"https://arxiv.org/abs/1803.11347"},{"text":"K. Chua, R. Calandra, R. McAllister, and S. Levine. Deep Reinforcement Learning in a Handful of Trials using Probabilistic Dynamics Models. arXiv:1805.12114, 2018.","identifier":"arXiv:1805.12114","url":"https://arxiv.org/abs/1805.12114"},{"text":"T. Lew, A. Sharma, J. Harrison, A. Bylard, and M. Pavone. Safe Active Dynamics Learning and Control: A Sequential Exploration-Exploitation Framework. arXiv:2008.11700, 2020.","identifier":"arXiv:2008.11700","url":"https://arxiv.org/abs/2008.11700"},{"text":"T. Koller, F. Berkenkamp, M. Turchetta, and A. Krause. Learning-Based Model Predictive Control for Safe Exploration. arXiv:1803.08287, 2018.","identifier":"arXiv:1803.08287","url":"https://arxiv.org/abs/1803.08287"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. arXiv:1703.06907, 2017.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. arXiv:1710.06537, 2017.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"},{"text":"J. Tan, T. Zhang, E. Coumans, A. Iscen, Y. Bai, D. Hafner, S. Bohez, and V. Vanhoucke. Sim-to-Real: Learning Agile Locomotion For Quadruped Robots. arXiv:1804.10332, 2018.","identifier":"arXiv:1804.10332","url":"https://arxiv.org/abs/1804.10332"},{"text":"Z. Xie, X. Da, M. van de Panne, B. Babich, and A. Garg. Dynamics Randomization Revisited: A Case Study for Quadrupedal Locomotion. arXiv:2011.02404, 2020.","identifier":"arXiv:2011.02404","url":"https://arxiv.org/abs/2011.02404"},{"text":"Y. Chebotar, A. Handa, V. Makoviychuk, M. Macklin, J. Issac, N. Ratliff, and D. Fox. Closing the Sim-to-Real Loop: Adapting Simulation Randomization with Real World Experience. arXiv:1810.05687, 2018.","identifier":"arXiv:1810.05687","url":"https://arxiv.org/abs/1810.05687"},{"text":"F. Ramos, R. C. Possas, and D. Fox. BayesSim: Adaptive Domain Randomization via Probabilistic Inference for Robotics Simulators. arXiv:1906.01728, 2019.","identifier":"arXiv:1906.01728","url":"https://arxiv.org/abs/1906.01728"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"G. B. Margolis, G. Yang, K. Paigwar, T. Chen, and P. Agrawal. Rapid Locomotion via Reinforcement Learning. arXiv:2205.02824, 2022.","identifier":"arXiv:2205.02824","url":"https://arxiv.org/abs/2205.02824"},{"text":"G. B. Margolis and P. Agrawal. Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior. arXiv:2212.03238, 2022.","identifier":"arXiv:2212.03238","url":"https://arxiv.org/abs/2212.03238"},{"text":"Y. Yang, T. Zhang, E. Coumans, J. Tan, and B. Boots. Fast and Efficient Locomotion via Learned Gait Transitions. arXiv:2104.04644, 2021.","identifier":"arXiv:2104.04644","url":"https://arxiv.org/abs/2104.04644"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. arXiv:2211.07638, 2022.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. arXiv:2309.14341, 2023.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"X. Xiao, B. Liu, G. Warnell, and P. Stone. Motion Planning and Control for Mobile Robot Navigation Using Machine Learning: a Survey. arXiv:2011.13112, 2020.","identifier":"arXiv:2011.13112","url":"https://arxiv.org/abs/2011.13112"},{"text":"X. Xiao, J. Biswas, and P. Stone. Learning Inverse Kinodynamics for Accurate High-Speed Off-Road Navigation on Unstructured Terrain. arXiv:2102.12667, 2021.","identifier":"arXiv:2102.12667","url":"https://arxiv.org/abs/2102.12667"},{"text":"H. Karnan, K. S. Sikand, P. Atreya, S. Rabiee, X. Xiao, G. Warnell, P. Stone, and J. Biswas. VI-IKD: High-Speed Accurate Off-Road Navigation using Learned Visual-Inertial Inverse Kinodynamics. arXiv:2203.15983, 2022.","identifier":"arXiv:2203.15983","url":"https://arxiv.org/abs/2203.15983"},{"text":"G. Kahn, P. Abbeel, and S. Levine. BADGR: An Autonomous Self-Supervised Learning-Based Navigation System. arXiv:2002.05700, 2020.","identifier":"arXiv:2002.05700","url":"https://arxiv.org/abs/2002.05700"},{"text":"Z. Fu, A. Kumar, A. Agarwal, H. Qi, J. Malik, and D. Pathak. Coupling Vision and Proprioception for Navigation of Legged Robots. arXiv:2112.02094, 2021.","identifier":"arXiv:2112.02094","url":"https://arxiv.org/abs/2112.02094"},{"text":"J. Truong, D. Yarats, T. Li, F. Meier, S. Chernova, D. Batra, and A. Rai. Learning Navigation Skills for Legged Robots with Learned Robot Embeddings. arXiv:2011.12255, 2020.","identifier":"arXiv:2011.12255","url":"https://arxiv.org/abs/2011.12255"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. arXiv:2305.08510, 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"T. Miki, L. Wellhausen, R. Grandia, F. Jenelten, T. Homberger, and M. Hutter. Elevation Mapping for Locomotion and Navigation using GPU. arXiv:2204.12876, 2022.","identifier":"arXiv:2204.12876","url":"https://arxiv.org/abs/2204.12876"},{"text":"J. Lee, M. Bjelonic, A. Reske, L. Wellhausen, T. Miki, and M. Hutter. Learning Robust Autonomous Navigation and Locomotion for Wheeled-Legged Robots. arXiv:2405.01792, 2024.","identifier":"arXiv:2405.01792","url":"https://arxiv.org/abs/2405.01792"},{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"N. Yokoyama, S. Ha, and D. Batra. Success Weighted by Completion Time: A Dynamics-Aware Evaluation Criteria for Embodied Navigation. arXiv:2103.08022, 2021.","identifier":"arXiv:2103.08022","url":"https://arxiv.org/abs/2103.08022"},{"text":"A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada. Control Barrier Functions: Theory and Applications. arXiv:1903.11199, 2019.","identifier":"arXiv:1903.11199","url":"https://arxiv.org/abs/1903.11199"},{"text":"S. Bansal, M. Chen, S. Herbert, and C. J. Tomlin. Hamilton-Jacobi Reachability: A Brief Overview and Recent Advances. arXiv:1709.07523, 2017.","identifier":"arXiv:1709.07523","url":"https://arxiv.org/abs/1709.07523"},{"text":"L. Brunke, M. Greeff, A. W. Hall, Z. Yuan, S. Zhou, J. Panerati, and A. P. Schoellig. Safe Learning in Robotics: From Learning-Based Control to Safe Reinforcement Learning. arXiv:2108.06266, 2021.","identifier":"arXiv:2108.06266","url":"https://arxiv.org/abs/2108.06266"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2608.00008","slug":"2608.00008","registered_at":"2026-08-21T08:45:00Z","updated_at":"2026-08-23T17:57:00Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2608.00008","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2608.00008","tex_path":"articles/2608.00008/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-08-21T11:02:00Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2608.00008"},{"kagg_id":"KAGG:2608.00007","slug":"2608.00007","title":"How Much Future Range Do You Need? An Ablation Study for Agile Quadruped Navigation","authors":[{"name":"Minju Seo","affiliation":"Legged Locomotion and Control Lab, KAIST, Daejeon, South Korea"},{"name":"Junho Kang","affiliation":"Legged Locomotion and Control Lab, KAIST, Daejeon, South Korea"},{"name":"Hyesun Bae","affiliation":"Legged Locomotion and Control Lab, KAIST, Daejeon, South Korea"}],"affiliation":null,"abstract":"Future-range predictors estimate how far a legged robot can travel along control-aligned directions within a prediction horizon H, but the horizon is usually fixed by hand even though the appropriate look-ahead depends on speed and terrain. In a systematic ablation with the full navigation stack held fixed, the horizon (0.25, 0.5, 1.0, and 2.0 s), the query density, and the number of control-aligned directions are varied for a simulated Unitree Go2 in Isaac Lab across 12 scene families, speeds from 1 to 5 m/s, and flat, mixed, and rough terrain. From these results, a lightweight rule is fitted that selects the horizon from the commanded speed and an online terrain estimate, and the rule is then run on a physical Go2. In the simulated main comparison, the selected rule reaches a 92.3 percent success rate, compared with 81.1 percent for the best fixed horizon and 80.9 percent without future range, while maintaining a mean speed of 3.30 m/s. In the horizon sweep, success peaks at H = 0.5 s and declines for shorter and longer horizons; long horizons are most harmful on occluded turns at 4 m/s, where H = 2.0 s produces 4.7 braking events per km. The rule adds 0.4 ms of decision latency, and on hardware it selected horizons between 0.5 and 1.0 s in 8/8 recorded runs. The fixed-horizon entries are not consistent across the recorded tables, and in the sweep the rule matches but does not exceed the best fixed horizon.","publication_date":"2026-08-14","registered_at":"2026-08-16T10:45:00Z","updated_at":"2026-08-18T15:57:00Z","status":"published","html_path":null,"pdf_path":"articles/2608.00007/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"T. Zang, S. Wang, H. Huang, and S. Cheng. FutureRay: Control-Aligned Future Range for Agile Quadruped Navigation. arXiv:2608.01197, 2026.","identifier":"arXiv:2608.01197","url":"https://arxiv.org/abs/2608.01197"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Tan, T. Zhang, E. Coumans, A. Iscen, Y. Bai, D. Hafner, S. Bohez, and V. Vanhoucke. Sim-to-Real: Learning Agile Locomotion For Quadruped Robots. arXiv:1804.10332, 2018.","identifier":"arXiv:1804.10332","url":"https://arxiv.org/abs/1804.10332"},{"text":"G. Ji, J. Mun, H. Kim, and J. Hwangbo. Concurrent Training of a Control Policy and a State Estimator for Dynamic and Robust Legged Locomotion. arXiv:2202.05481, 2022.","identifier":"arXiv:2202.05481","url":"https://arxiv.org/abs/2202.05481"},{"text":"Y. Kim, H. Oh, J. Lee, J. Choi, G. Ji, M. Jung, D. Youm, and J. Hwangbo. Not Only Rewards But Also Constraints: Applications on Legged Robot Locomotion. arXiv:2308.12517, 2023.","identifier":"arXiv:2308.12517","url":"https://arxiv.org/abs/2308.12517"},{"text":"G. B. Margolis, G. Yang, K. Paigwar, T. Chen, and P. Agrawal. Rapid Locomotion via Reinforcement Learning. arXiv:2205.02824, 2022.","identifier":"arXiv:2205.02824","url":"https://arxiv.org/abs/2205.02824"},{"text":"G. B. Margolis and P. Agrawal. Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior. arXiv:2212.03238, 2022.","identifier":"arXiv:2212.03238","url":"https://arxiv.org/abs/2212.03238"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. arXiv:2211.07638, 2022.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. arXiv:2309.14341, 2023.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. arXiv:2309.05665, 2023.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, L. Wellhausen, F. Farshidian, and M. Hutter. Learning a State Representation and Navigation in Cluttered and Dynamic Environments. arXiv:2103.04351, 2021.","identifier":"arXiv:2103.04351","url":"https://arxiv.org/abs/2103.04351"},{"text":"Z. Fu, A. Kumar, A. Agarwal, H. Qi, J. Malik, and D. Pathak. Coupling Vision and Proprioception for Navigation of Legged Robots. arXiv:2112.02094, 2021.","identifier":"arXiv:2112.02094","url":"https://arxiv.org/abs/2112.02094"},{"text":"J. Truong, D. Yarats, T. Li, F. Meier, S. Chernova, D. Batra, and A. Rai. Learning Navigation Skills for Legged Robots with Learned Robot Embeddings. arXiv:2011.12255, 2020.","identifier":"arXiv:2011.12255","url":"https://arxiv.org/abs/2011.12255"},{"text":"S. Kareer, N. Yokoyama, D. Batra, S. Ha, and J. Truong. ViNL: Visual Navigation and Locomotion Over Obstacles. arXiv:2210.14791, 2022.","identifier":"arXiv:2210.14791","url":"https://arxiv.org/abs/2210.14791"},{"text":"Y. Kim, C. Kim, and J. Hwangbo. Learning Forward Dynamics Model and Informed Trajectory Sampler for Safe Quadruped Navigation. arXiv:2204.08647, 2022.","identifier":"arXiv:2204.08647","url":"https://arxiv.org/abs/2204.08647"},{"text":"L. Wellhausen and M. Hutter. ArtPlanner: Robust Legged Robot Navigation in the Field. arXiv:2303.01420, 2023.","identifier":"arXiv:2303.01420","url":"https://arxiv.org/abs/2303.01420"},{"text":"J. Lee, M. Bjelonic, A. Reske, L. Wellhausen, T. Miki, and M. Hutter. Learning Robust Autonomous Navigation and Locomotion for Wheeled-Legged Robots. arXiv:2405.01792, 2024.","identifier":"arXiv:2405.01792","url":"https://arxiv.org/abs/2405.01792"},{"text":"R. Grandia, F. Jenelten, S. Yang, F. Farshidian, and M. Hutter. Perceptive Locomotion through Nonlinear Model Predictive Control. arXiv:2208.08373, 2022.","identifier":"arXiv:2208.08373","url":"https://arxiv.org/abs/2208.08373"},{"text":"D. Hoeller, N. Rudin, C. Choy, A. Anandkumar, and M. Hutter. Neural Scene Representation for Locomotion on Structured Terrain. arXiv:2206.08077, 2022.","identifier":"arXiv:2206.08077","url":"https://arxiv.org/abs/2206.08077"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. arXiv:2305.08510, 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"G. Kahn, P. Abbeel, and S. Levine. BADGR: An Autonomous Self-Supervised Learning-Based Navigation System. arXiv:2002.05700, 2020.","identifier":"arXiv:2002.05700","url":"https://arxiv.org/abs/2002.05700"},{"text":"G. Kahn, A. Villaflor, B. Ding, P. Abbeel, and S. Levine. Self-supervised Deep Reinforcement Learning with Generalized Computation Graphs for Robot Navigation. arXiv:1709.10489, 2017.","identifier":"arXiv:1709.10489","url":"https://arxiv.org/abs/1709.10489"},{"text":"G. Williams, P. Drews, B. Goldfain, J. M. Rehg, and E. A. Theodorou. Information Theoretic Model Predictive Control: Theory and Applications to Autonomous Driving. arXiv:1707.02342, 2017.","identifier":"arXiv:1707.02342","url":"https://arxiv.org/abs/1707.02342"},{"text":"J. Tordesillas, B. T. Lopez, and J. P. How. FASTER: Fast and Safe Trajectory Planner for Flights in Unknown Environments. arXiv:1903.03558, 2019.","identifier":"arXiv:1903.03558","url":"https://arxiv.org/abs/1903.03558"},{"text":"X. Zhou, Z. Wang, H. Ye, C. Xu, and F. Gao. EGO-Planner: An ESDF-free Gradient-based Local Planner for Quadrotors. arXiv:2008.08835, 2020.","identifier":"arXiv:2008.08835","url":"https://arxiv.org/abs/2008.08835"},{"text":"A. Loquercio, E. Kaufmann, R. Ranftl, M. Müller, V. Koltun, and D. Scaramuzza. Learning High-Speed Flight in the Wild. arXiv:2110.05113, 2021.","identifier":"arXiv:2110.05113","url":"https://arxiv.org/abs/2110.05113"},{"text":"S. Bansal, V. Tolani, S. Gupta, J. Malik, and C. Tomlin. Combining Optimal Control and Learning for Visual Navigation in Novel Environments. arXiv:1903.02531, 2019.","identifier":"arXiv:1903.02531","url":"https://arxiv.org/abs/1903.02531"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"M. Mittal et al. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. arXiv:2301.04195, 2023.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"V. Makoviychuk et al. Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. arXiv:1703.06907, 2017.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"},{"text":"A. Bouman et al. Autonomous Spot: Long-Range Autonomous Exploration of Extreme Environments with Legged Locomotion. arXiv:2010.09259, 2020.","identifier":"arXiv:2010.09259","url":"https://arxiv.org/abs/2010.09259"},{"text":"M. Tranzatto et al. CERBERUS in the DARPA Subterranean Challenge. arXiv:2201.07067, 2022.","identifier":"arXiv:2201.07067","url":"https://arxiv.org/abs/2201.07067"},{"text":"A. Agha et al. NeBula: Quest for Robotic Autonomy in Challenging Environments; TEAM CoSTAR at the DARPA Subterranean Challenge. arXiv:2103.11470, 2021.","identifier":"arXiv:2103.11470","url":"https://arxiv.org/abs/2103.11470"},{"text":"R. Yang, M. Zhang, N. Hansen, H. Xu, and X. Wang. Learning Vision-Guided Quadrupedal Locomotion End-to-End with Cross-Modal Transformers. arXiv:2107.03996, 2021.","identifier":"arXiv:2107.03996","url":"https://arxiv.org/abs/2107.03996"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2608.00007","metadata":{"title":"How Much Future Range Do You Need? An Ablation Study for Agile Quadruped Navigation","authors":[{"name":"Minju Seo","affiliation":"Legged Locomotion and Control Lab, KAIST, Daejeon, South Korea"},{"name":"Junho Kang","affiliation":"Legged Locomotion and Control Lab, KAIST, Daejeon, South Korea"},{"name":"Hyesun Bae","affiliation":"Legged Locomotion and Control Lab, KAIST, Daejeon, South Korea"}],"publication_date":"2026-08-14","abstract":"Future-range predictors estimate how far a legged robot can travel along control-aligned directions within a prediction horizon H, but the horizon is usually fixed by hand even though the appropriate look-ahead depends on speed and terrain. In a systematic ablation with the full navigation stack held fixed, the horizon (0.25, 0.5, 1.0, and 2.0 s), the query density, and the number of control-aligned directions are varied for a simulated Unitree Go2 in Isaac Lab across 12 scene families, speeds from 1 to 5 m/s, and flat, mixed, and rough terrain. From these results, a lightweight rule is fitted that selects the horizon from the commanded speed and an online terrain estimate, and the rule is then run on a physical Go2. In the simulated main comparison, the selected rule reaches a 92.3 percent success rate, compared with 81.1 percent for the best fixed horizon and 80.9 percent without future range, while maintaining a mean speed of 3.30 m/s. In the horizon sweep, success peaks at H = 0.5 s and declines for shorter and longer horizons; long horizons are most harmful on occluded turns at 4 m/s, where H = 2.0 s produces 4.7 braking events per km. The rule adds 0.4 ms of decision latency, and on hardware it selected horizons between 0.5 and 1.0 s in 8/8 recorded runs. The fixed-horizon entries are not consistent across the recorded tables, and in the sweep the rule matches but does not exceed the best fixed horizon.","references":[{"text":"T. Zang, S. Wang, H. Huang, and S. Cheng. FutureRay: Control-Aligned Future Range for Agile Quadruped Navigation. arXiv:2608.01197, 2026.","identifier":"arXiv:2608.01197","url":"https://arxiv.org/abs/2608.01197"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Tan, T. Zhang, E. Coumans, A. Iscen, Y. Bai, D. Hafner, S. Bohez, and V. Vanhoucke. Sim-to-Real: Learning Agile Locomotion For Quadruped Robots. arXiv:1804.10332, 2018.","identifier":"arXiv:1804.10332","url":"https://arxiv.org/abs/1804.10332"},{"text":"G. Ji, J. Mun, H. Kim, and J. Hwangbo. Concurrent Training of a Control Policy and a State Estimator for Dynamic and Robust Legged Locomotion. arXiv:2202.05481, 2022.","identifier":"arXiv:2202.05481","url":"https://arxiv.org/abs/2202.05481"},{"text":"Y. Kim, H. Oh, J. Lee, J. Choi, G. Ji, M. Jung, D. Youm, and J. Hwangbo. Not Only Rewards But Also Constraints: Applications on Legged Robot Locomotion. arXiv:2308.12517, 2023.","identifier":"arXiv:2308.12517","url":"https://arxiv.org/abs/2308.12517"},{"text":"G. B. Margolis, G. Yang, K. Paigwar, T. Chen, and P. Agrawal. Rapid Locomotion via Reinforcement Learning. arXiv:2205.02824, 2022.","identifier":"arXiv:2205.02824","url":"https://arxiv.org/abs/2205.02824"},{"text":"G. B. Margolis and P. Agrawal. Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior. arXiv:2212.03238, 2022.","identifier":"arXiv:2212.03238","url":"https://arxiv.org/abs/2212.03238"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. arXiv:2211.07638, 2022.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. arXiv:2309.14341, 2023.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. arXiv:2309.05665, 2023.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, L. Wellhausen, F. Farshidian, and M. Hutter. Learning a State Representation and Navigation in Cluttered and Dynamic Environments. arXiv:2103.04351, 2021.","identifier":"arXiv:2103.04351","url":"https://arxiv.org/abs/2103.04351"},{"text":"Z. Fu, A. Kumar, A. Agarwal, H. Qi, J. Malik, and D. Pathak. Coupling Vision and Proprioception for Navigation of Legged Robots. arXiv:2112.02094, 2021.","identifier":"arXiv:2112.02094","url":"https://arxiv.org/abs/2112.02094"},{"text":"J. Truong, D. Yarats, T. Li, F. Meier, S. Chernova, D. Batra, and A. Rai. Learning Navigation Skills for Legged Robots with Learned Robot Embeddings. arXiv:2011.12255, 2020.","identifier":"arXiv:2011.12255","url":"https://arxiv.org/abs/2011.12255"},{"text":"S. Kareer, N. Yokoyama, D. Batra, S. Ha, and J. Truong. ViNL: Visual Navigation and Locomotion Over Obstacles. arXiv:2210.14791, 2022.","identifier":"arXiv:2210.14791","url":"https://arxiv.org/abs/2210.14791"},{"text":"Y. Kim, C. Kim, and J. Hwangbo. Learning Forward Dynamics Model and Informed Trajectory Sampler for Safe Quadruped Navigation. arXiv:2204.08647, 2022.","identifier":"arXiv:2204.08647","url":"https://arxiv.org/abs/2204.08647"},{"text":"L. Wellhausen and M. Hutter. ArtPlanner: Robust Legged Robot Navigation in the Field. arXiv:2303.01420, 2023.","identifier":"arXiv:2303.01420","url":"https://arxiv.org/abs/2303.01420"},{"text":"J. Lee, M. Bjelonic, A. Reske, L. Wellhausen, T. Miki, and M. Hutter. Learning Robust Autonomous Navigation and Locomotion for Wheeled-Legged Robots. arXiv:2405.01792, 2024.","identifier":"arXiv:2405.01792","url":"https://arxiv.org/abs/2405.01792"},{"text":"R. Grandia, F. Jenelten, S. Yang, F. Farshidian, and M. Hutter. Perceptive Locomotion through Nonlinear Model Predictive Control. arXiv:2208.08373, 2022.","identifier":"arXiv:2208.08373","url":"https://arxiv.org/abs/2208.08373"},{"text":"D. Hoeller, N. Rudin, C. Choy, A. Anandkumar, and M. Hutter. Neural Scene Representation for Locomotion on Structured Terrain. arXiv:2206.08077, 2022.","identifier":"arXiv:2206.08077","url":"https://arxiv.org/abs/2206.08077"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. arXiv:2305.08510, 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"G. Kahn, P. Abbeel, and S. Levine. BADGR: An Autonomous Self-Supervised Learning-Based Navigation System. arXiv:2002.05700, 2020.","identifier":"arXiv:2002.05700","url":"https://arxiv.org/abs/2002.05700"},{"text":"G. Kahn, A. Villaflor, B. Ding, P. Abbeel, and S. Levine. Self-supervised Deep Reinforcement Learning with Generalized Computation Graphs for Robot Navigation. arXiv:1709.10489, 2017.","identifier":"arXiv:1709.10489","url":"https://arxiv.org/abs/1709.10489"},{"text":"G. Williams, P. Drews, B. Goldfain, J. M. Rehg, and E. A. Theodorou. Information Theoretic Model Predictive Control: Theory and Applications to Autonomous Driving. arXiv:1707.02342, 2017.","identifier":"arXiv:1707.02342","url":"https://arxiv.org/abs/1707.02342"},{"text":"J. Tordesillas, B. T. Lopez, and J. P. How. FASTER: Fast and Safe Trajectory Planner for Flights in Unknown Environments. arXiv:1903.03558, 2019.","identifier":"arXiv:1903.03558","url":"https://arxiv.org/abs/1903.03558"},{"text":"X. Zhou, Z. Wang, H. Ye, C. Xu, and F. Gao. EGO-Planner: An ESDF-free Gradient-based Local Planner for Quadrotors. arXiv:2008.08835, 2020.","identifier":"arXiv:2008.08835","url":"https://arxiv.org/abs/2008.08835"},{"text":"A. Loquercio, E. Kaufmann, R. Ranftl, M. Müller, V. Koltun, and D. Scaramuzza. Learning High-Speed Flight in the Wild. arXiv:2110.05113, 2021.","identifier":"arXiv:2110.05113","url":"https://arxiv.org/abs/2110.05113"},{"text":"S. Bansal, V. Tolani, S. Gupta, J. Malik, and C. Tomlin. Combining Optimal Control and Learning for Visual Navigation in Novel Environments. arXiv:1903.02531, 2019.","identifier":"arXiv:1903.02531","url":"https://arxiv.org/abs/1903.02531"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"M. Mittal et al. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. arXiv:2301.04195, 2023.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"V. Makoviychuk et al. Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. arXiv:1703.06907, 2017.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"},{"text":"A. Bouman et al. Autonomous Spot: Long-Range Autonomous Exploration of Extreme Environments with Legged Locomotion. arXiv:2010.09259, 2020.","identifier":"arXiv:2010.09259","url":"https://arxiv.org/abs/2010.09259"},{"text":"M. Tranzatto et al. CERBERUS in the DARPA Subterranean Challenge. arXiv:2201.07067, 2022.","identifier":"arXiv:2201.07067","url":"https://arxiv.org/abs/2201.07067"},{"text":"A. Agha et al. NeBula: Quest for Robotic Autonomy in Challenging Environments; TEAM CoSTAR at the DARPA Subterranean Challenge. arXiv:2103.11470, 2021.","identifier":"arXiv:2103.11470","url":"https://arxiv.org/abs/2103.11470"},{"text":"R. Yang, M. Zhang, N. Hansen, H. Xu, and X. Wang. Learning Vision-Guided Quadrupedal Locomotion End-to-End with Cross-Modal Transformers. arXiv:2107.03996, 2021.","identifier":"arXiv:2107.03996","url":"https://arxiv.org/abs/2107.03996"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2608.00007","slug":"2608.00007","registered_at":"2026-08-16T10:45:00Z","updated_at":"2026-08-18T15:57:00Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2608.00007","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2608.00007","tex_path":"articles/2608.00007/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-08-16T13:02:00Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2608.00007"},{"kagg_id":"KAGG:2608.00006","slug":"2608.00006","title":"Cross-Embodiment Transfer of Local Navigation Skills via Configuration Randomization","authors":[{"name":"Sakura Ishikawa","affiliation":"Legged Systems Group, The University of Tokyo, Tokyo, Japan"},{"name":"Kenji Watanabe","affiliation":"Legged Systems Group, The University of Tokyo, Tokyo, Japan"},{"name":"Daichi Mori","affiliation":"Legged Systems Group, The University of Tokyo, Tokyo, Japan"},{"name":"Aiko Tanaka","affiliation":"Legged Systems Group, The University of Tokyo, Tokyo, Japan"}],"affiliation":null,"abstract":"Local navigation skills learned for one quadruped transfer poorly to another, because leg length, payload, and sensor placement change what the robot perceives and which motions it can execute. This paper extends configuration randomization from low-level locomotion to the skill level. A single local navigation policy is trained in simulation with the AgniNav configuration descriptor, augmented with leg length and payload mass, as an explicit input, while the configuration parameters are sampled from the envelope spanned by a robot fleet. The descriptor modulates the policy through feature-wise affine conditioning and scales its velocity limits. Each new embodiment is adapted with a few hundred rollout episodes by a fine-tune that updates only the conditioning and output layers and is anchored to the pretrained policy by a KL penalty. In Isaac Lab, with a fleet of four quadruped morphologies (leg lengths 0.42 to 0.58 m, payloads 0 to 12 kg, sensor heights 0.35 to 0.55 m), 20 navigation scenarios, and 5 seeds, the method reaches 82.6 percent success on held-out morphologies outside the training support, compared with 73.1 percent for morphology-specific training and 77.8 percent for per-morphology fine-tuning from scratch, and reaches 90 percent of its final success after 420 fine-tune episodes on average, compared with 481 and 455. Ablations identify the configuration descriptor and payload randomization as the most consequential components. The gains come with a smaller clearance margin, and success drops to 74.1 percent when leg length is extrapolated by 22 percent.","publication_date":"2026-08-12","registered_at":"2026-08-13T12:14:00Z","updated_at":"2026-08-14T17:38:00Z","status":"published","html_path":null,"pdf_path":"articles/2608.00006/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"D. Hoeller, L. Wellhausen, F. Farshidian, and M. Hutter. Learning a State Representation and Navigation in Cluttered and Dynamic Environments. arXiv:2103.04351, 2021.","identifier":"arXiv:2103.04351","url":"https://arxiv.org/abs/2103.04351"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. arXiv:1703.06907, 2017.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. arXiv:1710.06537, 2017.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"},{"text":"OpenAI, I. Akkaya et al.. Solving Rubik's Cube with a Robot Hand. arXiv:1910.07113, 2019.","identifier":"arXiv:1910.07113","url":"https://arxiv.org/abs/1910.07113"},{"text":"G. Feng et al.. GenLoco: Generalized Locomotion Controllers for Quadrupedal Robots. arXiv:2209.05309, 2022.","identifier":"arXiv:2209.05309","url":"https://arxiv.org/abs/2209.05309"},{"text":"N. Bohlinger, G. Czechmanowski, M. Krupka, P. Kicki, K. Walas, J. Peters, and D. Tateo. One Policy to Run Them All: an End-to-end Learning Approach to Multi-Embodiment Locomotion. arXiv:2409.06366, 2024.","identifier":"arXiv:2409.06366","url":"https://arxiv.org/abs/2409.06366"},{"text":"E. Perez, F. Strub, H. de Vries, V. Dumoulin, and A. Courville. FiLM: Visual Reasoning with a General Conditioning Layer. arXiv:1709.07871, 2017.","identifier":"arXiv:1709.07871","url":"https://arxiv.org/abs/1709.07871"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. arXiv:2309.05665, 2023.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. arXiv:2309.14341, 2023.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"J. Truong, D. Yarats, T. Li, F. Meier, S. Chernova, D. Batra, and A. Rai. Learning Navigation Skills for Legged Robots with Learned Robot Embeddings. arXiv:2011.12255, 2020.","identifier":"arXiv:2011.12255","url":"https://arxiv.org/abs/2011.12255"},{"text":"W. Yu, J. Tan, C. K. Liu, and G. Turk. Preparing for the Unknown: Learning a Universal Policy with Online System Identification. arXiv:1702.02453, 2017.","identifier":"arXiv:1702.02453","url":"https://arxiv.org/abs/1702.02453"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"C. Finn, P. Abbeel, and S. Levine. Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks. arXiv:1703.03400, 2017.","identifier":"arXiv:1703.03400","url":"https://arxiv.org/abs/1703.03400"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Sridhar, D. Shah, C. Glossop, and S. Levine. NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration. arXiv:2310.07896, 2023.","identifier":"arXiv:2310.07896","url":"https://arxiv.org/abs/2310.07896"},{"text":"A. Eftekhar et al.. The One RING: a Robotic Indoor Navigation Generalist. arXiv:2412.14401, 2024.","identifier":"arXiv:2412.14401","url":"https://arxiv.org/abs/2412.14401"},{"text":"J. Zhang et al.. Embodied Navigation Foundation Model. arXiv:2509.12129, 2025.","identifier":"arXiv:2509.12129","url":"https://arxiv.org/abs/2509.12129"},{"text":"J. Zhang, Y. Du, X. Guo, S. Sun, X. Liu, Y. Sun, G. Lu, W. Sui, and J. Li. EA-Nav: Learning Safe Visual Navigation Policies with Embodiment Awareness. arXiv:2607.19880, 2026.","identifier":"arXiv:2607.19880","url":"https://arxiv.org/abs/2607.19880"},{"text":"Open X-Embodiment Collaboration et al.. Open X-Embodiment: Robotic Learning Datasets and RT-X Models. arXiv:2310.08864, 2023.","identifier":"arXiv:2310.08864","url":"https://arxiv.org/abs/2310.08864"},{"text":"T. Chen, A. Murali, and A. Gupta. Hardware Conditioned Policies for Multi-Robot Transfer Learning. arXiv:1811.09864, 2018.","identifier":"arXiv:1811.09864","url":"https://arxiv.org/abs/1811.09864"},{"text":"W. Huang, I. Mordatch, and D. Pathak. One Policy to Control Them All: Shared Modular Policies for Agent-Agnostic Control. arXiv:2007.04976, 2020.","identifier":"arXiv:2007.04976","url":"https://arxiv.org/abs/2007.04976"},{"text":"R. Doshi, H. Walke, O. Mees, S. Dasari, and S. Levine. Scaling Cross-Embodied Learning: One Policy for Manipulation, Navigation, Locomotion and Aviation. arXiv:2408.11812, 2024.","identifier":"arXiv:2408.11812","url":"https://arxiv.org/abs/2408.11812"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"J. Hu, S. Yuan, G. C. R. Bethala, A. Tzes, and Y. Fang. Learning Adaptive Safety Margins for Visual Navigation. arXiv:2607.18200, 2026.","identifier":"arXiv:2607.18200","url":"https://arxiv.org/abs/2607.18200"},{"text":"V. Makoviychuk et al.. Isaac Gym: High Performance GPU-Based Physics Simulation for Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal et al.. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. arXiv:2301.04195, 2023.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"A. Kadian et al.. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance? arXiv:1912.06321, 2019.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"},{"text":"J. Truong, M. Rudolph, N. H. Yokoyama, S. Chernova, D. Batra, and A. Rai. Rethinking Sim2Real: Lower Fidelity Simulation Leads to Higher Sim2Real Transfer in Navigation. arXiv:2207.10821, 2022.","identifier":"arXiv:2207.10821","url":"https://arxiv.org/abs/2207.10821"},{"text":"P. Anderson et al.. On Evaluation of Embodied Navigation Agents. arXiv:1807.06757, 2018.","identifier":"arXiv:1807.06757","url":"https://arxiv.org/abs/1807.06757"},{"text":"J. Schulman, F. Wolski, P. Dhariwal, A. Radford, and O. Klimov. Proximal Policy Optimization Algorithms. arXiv:1707.06347, 2017.","identifier":"arXiv:1707.06347","url":"https://arxiv.org/abs/1707.06347"},{"text":"J. Schulman, P. Moritz, S. Levine, M. Jordan, and P. Abbeel. High-Dimensional Continuous Control Using Generalized Advantage Estimation. arXiv:1506.02438, 2015.","identifier":"arXiv:1506.02438","url":"https://arxiv.org/abs/1506.02438"},{"text":"L. Pinto, M. Andrychowicz, P. Welinder, W. Zaremba, and P. Abbeel. Asymmetric Actor Critic for Image-Based Robot Learning. arXiv:1710.06542, 2017.","identifier":"arXiv:1710.06542","url":"https://arxiv.org/abs/1710.06542"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2608.00006","metadata":{"title":"Cross-Embodiment Transfer of Local Navigation Skills via Configuration Randomization","authors":[{"name":"Sakura Ishikawa","affiliation":"Legged Systems Group, The University of Tokyo, Tokyo, Japan"},{"name":"Kenji Watanabe","affiliation":"Legged Systems Group, The University of Tokyo, Tokyo, Japan"},{"name":"Daichi Mori","affiliation":"Legged Systems Group, The University of Tokyo, Tokyo, Japan"},{"name":"Aiko Tanaka","affiliation":"Legged Systems Group, The University of Tokyo, Tokyo, Japan"}],"publication_date":"2026-08-12","abstract":"Local navigation skills learned for one quadruped transfer poorly to another, because leg length, payload, and sensor placement change what the robot perceives and which motions it can execute. This paper extends configuration randomization from low-level locomotion to the skill level. A single local navigation policy is trained in simulation with the AgniNav configuration descriptor, augmented with leg length and payload mass, as an explicit input, while the configuration parameters are sampled from the envelope spanned by a robot fleet. The descriptor modulates the policy through feature-wise affine conditioning and scales its velocity limits. Each new embodiment is adapted with a few hundred rollout episodes by a fine-tune that updates only the conditioning and output layers and is anchored to the pretrained policy by a KL penalty. In Isaac Lab, with a fleet of four quadruped morphologies (leg lengths 0.42 to 0.58 m, payloads 0 to 12 kg, sensor heights 0.35 to 0.55 m), 20 navigation scenarios, and 5 seeds, the method reaches 82.6 percent success on held-out morphologies outside the training support, compared with 73.1 percent for morphology-specific training and 77.8 percent for per-morphology fine-tuning from scratch, and reaches 90 percent of its final success after 420 fine-tune episodes on average, compared with 481 and 455. Ablations identify the configuration descriptor and payload randomization as the most consequential components. The gains come with a smaller clearance margin, and success drops to 74.1 percent when leg length is extrapolated by 22 percent.","references":[{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"D. Hoeller, L. Wellhausen, F. Farshidian, and M. Hutter. Learning a State Representation and Navigation in Cluttered and Dynamic Environments. arXiv:2103.04351, 2021.","identifier":"arXiv:2103.04351","url":"https://arxiv.org/abs/2103.04351"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. arXiv:1703.06907, 2017.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. arXiv:1710.06537, 2017.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"},{"text":"OpenAI, I. Akkaya et al.. Solving Rubik's Cube with a Robot Hand. arXiv:1910.07113, 2019.","identifier":"arXiv:1910.07113","url":"https://arxiv.org/abs/1910.07113"},{"text":"G. Feng et al.. GenLoco: Generalized Locomotion Controllers for Quadrupedal Robots. arXiv:2209.05309, 2022.","identifier":"arXiv:2209.05309","url":"https://arxiv.org/abs/2209.05309"},{"text":"N. Bohlinger, G. Czechmanowski, M. Krupka, P. Kicki, K. Walas, J. Peters, and D. Tateo. One Policy to Run Them All: an End-to-end Learning Approach to Multi-Embodiment Locomotion. arXiv:2409.06366, 2024.","identifier":"arXiv:2409.06366","url":"https://arxiv.org/abs/2409.06366"},{"text":"E. Perez, F. Strub, H. de Vries, V. Dumoulin, and A. Courville. FiLM: Visual Reasoning with a General Conditioning Layer. arXiv:1709.07871, 2017.","identifier":"arXiv:1709.07871","url":"https://arxiv.org/abs/1709.07871"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. arXiv:2309.05665, 2023.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. arXiv:2309.14341, 2023.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"J. Truong, D. Yarats, T. Li, F. Meier, S. Chernova, D. Batra, and A. Rai. Learning Navigation Skills for Legged Robots with Learned Robot Embeddings. arXiv:2011.12255, 2020.","identifier":"arXiv:2011.12255","url":"https://arxiv.org/abs/2011.12255"},{"text":"W. Yu, J. Tan, C. K. Liu, and G. Turk. Preparing for the Unknown: Learning a Universal Policy with Online System Identification. arXiv:1702.02453, 2017.","identifier":"arXiv:1702.02453","url":"https://arxiv.org/abs/1702.02453"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"C. Finn, P. Abbeel, and S. Levine. Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks. arXiv:1703.03400, 2017.","identifier":"arXiv:1703.03400","url":"https://arxiv.org/abs/1703.03400"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Sridhar, D. Shah, C. Glossop, and S. Levine. NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration. arXiv:2310.07896, 2023.","identifier":"arXiv:2310.07896","url":"https://arxiv.org/abs/2310.07896"},{"text":"A. Eftekhar et al.. The One RING: a Robotic Indoor Navigation Generalist. arXiv:2412.14401, 2024.","identifier":"arXiv:2412.14401","url":"https://arxiv.org/abs/2412.14401"},{"text":"J. Zhang et al.. Embodied Navigation Foundation Model. arXiv:2509.12129, 2025.","identifier":"arXiv:2509.12129","url":"https://arxiv.org/abs/2509.12129"},{"text":"J. Zhang, Y. Du, X. Guo, S. Sun, X. Liu, Y. Sun, G. Lu, W. Sui, and J. Li. EA-Nav: Learning Safe Visual Navigation Policies with Embodiment Awareness. arXiv:2607.19880, 2026.","identifier":"arXiv:2607.19880","url":"https://arxiv.org/abs/2607.19880"},{"text":"Open X-Embodiment Collaboration et al.. Open X-Embodiment: Robotic Learning Datasets and RT-X Models. arXiv:2310.08864, 2023.","identifier":"arXiv:2310.08864","url":"https://arxiv.org/abs/2310.08864"},{"text":"T. Chen, A. Murali, and A. Gupta. Hardware Conditioned Policies for Multi-Robot Transfer Learning. arXiv:1811.09864, 2018.","identifier":"arXiv:1811.09864","url":"https://arxiv.org/abs/1811.09864"},{"text":"W. Huang, I. Mordatch, and D. Pathak. One Policy to Control Them All: Shared Modular Policies for Agent-Agnostic Control. arXiv:2007.04976, 2020.","identifier":"arXiv:2007.04976","url":"https://arxiv.org/abs/2007.04976"},{"text":"R. Doshi, H. Walke, O. Mees, S. Dasari, and S. Levine. Scaling Cross-Embodied Learning: One Policy for Manipulation, Navigation, Locomotion and Aviation. arXiv:2408.11812, 2024.","identifier":"arXiv:2408.11812","url":"https://arxiv.org/abs/2408.11812"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"J. Hu, S. Yuan, G. C. R. Bethala, A. Tzes, and Y. Fang. Learning Adaptive Safety Margins for Visual Navigation. arXiv:2607.18200, 2026.","identifier":"arXiv:2607.18200","url":"https://arxiv.org/abs/2607.18200"},{"text":"V. Makoviychuk et al.. Isaac Gym: High Performance GPU-Based Physics Simulation for Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal et al.. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. arXiv:2301.04195, 2023.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"A. Kadian et al.. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance? arXiv:1912.06321, 2019.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"},{"text":"J. Truong, M. Rudolph, N. H. Yokoyama, S. Chernova, D. Batra, and A. Rai. Rethinking Sim2Real: Lower Fidelity Simulation Leads to Higher Sim2Real Transfer in Navigation. arXiv:2207.10821, 2022.","identifier":"arXiv:2207.10821","url":"https://arxiv.org/abs/2207.10821"},{"text":"P. Anderson et al.. On Evaluation of Embodied Navigation Agents. arXiv:1807.06757, 2018.","identifier":"arXiv:1807.06757","url":"https://arxiv.org/abs/1807.06757"},{"text":"J. Schulman, F. Wolski, P. Dhariwal, A. Radford, and O. Klimov. Proximal Policy Optimization Algorithms. arXiv:1707.06347, 2017.","identifier":"arXiv:1707.06347","url":"https://arxiv.org/abs/1707.06347"},{"text":"J. Schulman, P. Moritz, S. Levine, M. Jordan, and P. Abbeel. High-Dimensional Continuous Control Using Generalized Advantage Estimation. arXiv:1506.02438, 2015.","identifier":"arXiv:1506.02438","url":"https://arxiv.org/abs/1506.02438"},{"text":"L. Pinto, M. Andrychowicz, P. Welinder, W. Zaremba, and P. Abbeel. Asymmetric Actor Critic for Image-Based Robot Learning. arXiv:1710.06542, 2017.","identifier":"arXiv:1710.06542","url":"https://arxiv.org/abs/1710.06542"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2608.00006","slug":"2608.00006","registered_at":"2026-08-13T12:14:00Z","updated_at":"2026-08-14T17:38:00Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2608.00006","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2608.00006","tex_path":"articles/2608.00006/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-08-13T14:31:00Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2608.00006"},{"kagg_id":"KAGG:2608.00005","slug":"2608.00005","title":"Range-Plane Abstractions for Legged Robot Navigation in Cluttered Interiors","authors":[{"name":"Omar Farouk","affiliation":"Perception and Control Group, Technical University of Munich, Munich, Germany"},{"name":"Maximilian Weber","affiliation":"Perception and Control Group, Technical University of Munich, Munich, Germany"},{"name":"Lin Zhao","affiliation":"Perception and Control Group, Technical University of Munich, Munich, Germany"}],"affiliation":null,"abstract":"Legged navigation in building interiors needs more than a floor map but much less than a dense volumetric forecast. Two-dimensional floor projections discard the step heights and overhead clearances that decide whether a quadruped must crawl, walk normally, or raise its body, whereas full three-dimensional future-range prediction spends most of its output on elevations that no motion mode queries. This paper introduces range-plane abstractions: a small set of control-aligned planes, namely floor, low-obstacle, and headroom, each carrying its own future-range map along the directions the robot is predicted to execute. The height bands of the planes are anchored to the body envelope of the active motion mode (crawl, standard, or raised), and the local planner selects the plane set together with the mode, so that each feasibility query reduces to a first-hit range and an extremal height within one band. In Isaac Lab apartment suites (8 floor plans, 5 seeds, 30 trials each) and on a 900 m four-floor office-building course traversed by an ANYmal, the abstraction reduces predictor output dimensionality by 61.4 percent relative to a dense reference, against 56.6 percent for full three-dimensional future-range prediction. It also lowers p95 query latency from 9.2 to 8.9 ms, uses 842k instead of 931k predictor parameters, and raises task success from 85.9 to 90.2 percent. Ablations attribute the largest share of the gain to the low-obstacle plane; recorded failures, including a late mode switch at a 12 cm step, are analyzed.","publication_date":"2026-08-09","registered_at":"2026-08-10T15:23:00Z","updated_at":"2026-08-11T19:26:00Z","status":"published","html_path":null,"pdf_path":"articles/2608.00005/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"T. Zang, S. Wang, H. Huang, and S. Cheng. FutureRay: Control-Aligned Future Range for Agile Quadruped Navigation. arXiv:2608.01197, 2026.","identifier":"arXiv:2608.01197","url":"https://arxiv.org/abs/2608.01197"},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. arXiv:2211.07638, 2022.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. arXiv:2309.05665, 2023.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. arXiv:2309.14341, 2023.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, L. Wellhausen, F. Farshidian, and M. Hutter. Learning a State Representation and Navigation in Cluttered and Dynamic Environments. arXiv:2103.04351, 2021.","identifier":"arXiv:2103.04351","url":"https://arxiv.org/abs/2103.04351"},{"text":"C. Zhang, J. Jin, J. Frey, N. Rudin, M. Mattamala, C. Cadena, and M. Hutter. Resilient Legged Local Navigation: Learning to Traverse with Compromised Perception End-to-End. arXiv:2310.03581, 2023.","identifier":"arXiv:2310.03581","url":"https://arxiv.org/abs/2310.03581"},{"text":"J. Lee, M. Bjelonic, A. Reske, L. Wellhausen, T. Miki, and M. Hutter. Learning Robust Autonomous Navigation and Locomotion for Wheeled-Legged Robots. arXiv:2405.01792, 2024.","identifier":"arXiv:2405.01792","url":"https://arxiv.org/abs/2405.01792"},{"text":"T. Miki, L. Wellhausen, R. Grandia, F. Jenelten, T. Homberger, and M. Hutter. Elevation Mapping for Locomotion and Navigation using GPU. arXiv:2204.12876, 2022.","identifier":"arXiv:2204.12876","url":"https://arxiv.org/abs/2204.12876"},{"text":"L. Wellhausen and M. Hutter. ArtPlanner: Robust Legged Robot Navigation in the Field. arXiv:2303.01420, 2023.","identifier":"arXiv:2303.01420","url":"https://arxiv.org/abs/2303.01420"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. arXiv:2305.08510, 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"R. Grandia, F. Jenelten, S. Yang, F. Farshidian, and M. Hutter. Perceptive Locomotion through Nonlinear Model Predictive Control. arXiv:2208.08373, 2022.","identifier":"arXiv:2208.08373","url":"https://arxiv.org/abs/2208.08373"},{"text":"F. Jenelten, R. Grandia, F. Farshidian, and M. Hutter. TAMOLS: Terrain-Aware Motion Optimization for Legged Systems. arXiv:2206.14049, 2022.","identifier":"arXiv:2206.14049","url":"https://arxiv.org/abs/2206.14049"},{"text":"D. Hoeller, N. Rudin, C. Choy, A. Anandkumar, and M. Hutter. Neural Scene Representation for Locomotion on Structured Terrain. arXiv:2206.08077, 2022.","identifier":"arXiv:2206.08077","url":"https://arxiv.org/abs/2206.08077"},{"text":"R. Yang, G. Yang, and X. Wang. Neural Volumetric Memory for Visual Locomotion Control. arXiv:2304.01201, 2023.","identifier":"arXiv:2304.01201","url":"https://arxiv.org/abs/2304.01201"},{"text":"Z. Fu, A. Kumar, A. Agarwal, H. Qi, J. Malik, and D. Pathak. Coupling Vision and Proprioception for Navigation of Legged Robots. arXiv:2112.02094, 2021.","identifier":"arXiv:2112.02094","url":"https://arxiv.org/abs/2112.02094"},{"text":"S. Kareer, N. Yokoyama, D. Batra, S. Ha, and J. Truong. ViNL: Visual Navigation and Locomotion Over Obstacles. arXiv:2210.14791, 2022.","identifier":"arXiv:2210.14791","url":"https://arxiv.org/abs/2210.14791"},{"text":"H. Oleynikova, Z. Taylor, M. Fehr, R. Siegwart, and J. Nieto. Voxblox: Incremental 3D Euclidean Signed Distance Fields for On-Board MAV Planning. arXiv:1611.03631, 2016.","identifier":"arXiv:1611.03631","url":"https://arxiv.org/abs/1611.03631"},{"text":"X. Zhou, Z. Wang, H. Ye, C. Xu, and F. Gao. EGO-Planner: An ESDF-free Gradient-based Local Planner for Quadrotors. arXiv:2008.08835, 2020.","identifier":"arXiv:2008.08835","url":"https://arxiv.org/abs/2008.08835"},{"text":"G. Williams, P. Drews, B. Goldfain, J. M. Rehg, and E. A. Theodorou. Information-Theoretic Model Predictive Control: Theory and Applications to Autonomous Driving. arXiv:1707.02342, 2017.","identifier":"arXiv:1707.02342","url":"https://arxiv.org/abs/1707.02342"},{"text":"J. Philion and S. Fidler. Lift, Splat, Shoot: Encoding Images From Arbitrary Camera Rigs by Implicitly Unprojecting to 3D. arXiv:2008.05625, 2020.","identifier":"arXiv:2008.05625","url":"https://arxiv.org/abs/2008.05625"},{"text":"Z. Li, W. Wang, H. Li, E. Xie, C. Sima, T. Lu, Y. Qiao, and J. Dai. BEVFormer: Learning Bird's-Eye-View Representation from Multi-Camera Images via Spatiotemporal Transformers. arXiv:2203.17270, 2022.","identifier":"arXiv:2203.17270","url":"https://arxiv.org/abs/2203.17270"},{"text":"Y. Huang, W. Zheng, Y. Zhang, J. Zhou, and J. Lu. Tri-Perspective View for Vision-Based 3D Semantic Occupancy Prediction. arXiv:2302.07817, 2023.","identifier":"arXiv:2302.07817","url":"https://arxiv.org/abs/2302.07817"},{"text":"W. Tong, C. Sima, T. Wang, L. Chen, S. Wu, H. Deng, Y. Gu, L. Lu, P. Luo, D. Lin, and H. Li. Scene as Occupancy. arXiv:2306.02851, 2023.","identifier":"arXiv:2306.02851","url":"https://arxiv.org/abs/2306.02851"},{"text":"E. R. Chan et al.. Efficient Geometry-aware 3D Generative Adversarial Networks. arXiv:2112.07945, 2021.","identifier":"arXiv:2112.07945","url":"https://arxiv.org/abs/2112.07945"},{"text":"S. Fridovich-Keil, G. Meanti, F. R. Warburg, B. Recht, and A. Kanazawa. K-Planes: Explicit Radiance Fields in Space, Time, and Appearance. arXiv:2301.10241, 2023.","identifier":"arXiv:2301.10241","url":"https://arxiv.org/abs/2301.10241"},{"text":"G. Kahn, P. Abbeel, and S. Levine. BADGR: An Autonomous Self-Supervised Learning-Based Navigation System. arXiv:2002.05700, 2020.","identifier":"arXiv:2002.05700","url":"https://arxiv.org/abs/2002.05700"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Bouman et al.. Autonomous Spot: Long-Range Autonomous Exploration of Extreme Environments with Legged Locomotion. arXiv:2010.09259, 2020.","identifier":"arXiv:2010.09259","url":"https://arxiv.org/abs/2010.09259"},{"text":"M. Tranzatto et al.. CERBERUS: Autonomous Legged and Aerial Robotic Exploration in the Tunnel and Urban Circuits of the DARPA Subterranean Challenge. arXiv:2201.07067, 2022.","identifier":"arXiv:2201.07067","url":"https://arxiv.org/abs/2201.07067"},{"text":"M. Mittal et al.. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. arXiv:2301.04195, 2023.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"V. Makoviychuk et al.. Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"J. Schulman, F. Wolski, P. Dhariwal, A. Radford, and O. Klimov. Proximal Policy Optimization Algorithms. arXiv:1707.06347, 2017.","identifier":"arXiv:1707.06347","url":"https://arxiv.org/abs/1707.06347"},{"text":"J. Truong, M. Rudolph, N. Yokoyama, S. Chernova, D. Batra, and A. Rai. Rethinking Sim2Real: Lower Fidelity Simulation Leads to Higher Sim2Real Transfer in Navigation. arXiv:2207.10821, 2022.","identifier":"arXiv:2207.10821","url":"https://arxiv.org/abs/2207.10821"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2608.00005","metadata":{"title":"Range-Plane Abstractions for Legged Robot Navigation in Cluttered Interiors","authors":[{"name":"Omar Farouk","affiliation":"Perception and Control Group, Technical University of Munich, Munich, Germany"},{"name":"Maximilian Weber","affiliation":"Perception and Control Group, Technical University of Munich, Munich, Germany"},{"name":"Lin Zhao","affiliation":"Perception and Control Group, Technical University of Munich, Munich, Germany"}],"publication_date":"2026-08-09","abstract":"Legged navigation in building interiors needs more than a floor map but much less than a dense volumetric forecast. Two-dimensional floor projections discard the step heights and overhead clearances that decide whether a quadruped must crawl, walk normally, or raise its body, whereas full three-dimensional future-range prediction spends most of its output on elevations that no motion mode queries. This paper introduces range-plane abstractions: a small set of control-aligned planes, namely floor, low-obstacle, and headroom, each carrying its own future-range map along the directions the robot is predicted to execute. The height bands of the planes are anchored to the body envelope of the active motion mode (crawl, standard, or raised), and the local planner selects the plane set together with the mode, so that each feasibility query reduces to a first-hit range and an extremal height within one band. In Isaac Lab apartment suites (8 floor plans, 5 seeds, 30 trials each) and on a 900 m four-floor office-building course traversed by an ANYmal, the abstraction reduces predictor output dimensionality by 61.4 percent relative to a dense reference, against 56.6 percent for full three-dimensional future-range prediction. It also lowers p95 query latency from 9.2 to 8.9 ms, uses 842k instead of 931k predictor parameters, and raises task success from 85.9 to 90.2 percent. Ablations attribute the largest share of the gain to the low-obstacle plane; recorded failures, including a late mode switch at a 12 cm step, are analyzed.","references":[{"text":"T. Zang, S. Wang, H. Huang, and S. Cheng. FutureRay: Control-Aligned Future Range for Agile Quadruped Navigation. arXiv:2608.01197, 2026.","identifier":"arXiv:2608.01197","url":"https://arxiv.org/abs/2608.01197"},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. arXiv:2211.07638, 2022.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. arXiv:2309.05665, 2023.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. arXiv:2309.14341, 2023.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, L. Wellhausen, F. Farshidian, and M. Hutter. Learning a State Representation and Navigation in Cluttered and Dynamic Environments. arXiv:2103.04351, 2021.","identifier":"arXiv:2103.04351","url":"https://arxiv.org/abs/2103.04351"},{"text":"C. Zhang, J. Jin, J. Frey, N. Rudin, M. Mattamala, C. Cadena, and M. Hutter. Resilient Legged Local Navigation: Learning to Traverse with Compromised Perception End-to-End. arXiv:2310.03581, 2023.","identifier":"arXiv:2310.03581","url":"https://arxiv.org/abs/2310.03581"},{"text":"J. Lee, M. Bjelonic, A. Reske, L. Wellhausen, T. Miki, and M. Hutter. Learning Robust Autonomous Navigation and Locomotion for Wheeled-Legged Robots. arXiv:2405.01792, 2024.","identifier":"arXiv:2405.01792","url":"https://arxiv.org/abs/2405.01792"},{"text":"T. Miki, L. Wellhausen, R. Grandia, F. Jenelten, T. Homberger, and M. Hutter. Elevation Mapping for Locomotion and Navigation using GPU. arXiv:2204.12876, 2022.","identifier":"arXiv:2204.12876","url":"https://arxiv.org/abs/2204.12876"},{"text":"L. Wellhausen and M. Hutter. ArtPlanner: Robust Legged Robot Navigation in the Field. arXiv:2303.01420, 2023.","identifier":"arXiv:2303.01420","url":"https://arxiv.org/abs/2303.01420"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. arXiv:2305.08510, 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"R. Grandia, F. Jenelten, S. Yang, F. Farshidian, and M. Hutter. Perceptive Locomotion through Nonlinear Model Predictive Control. arXiv:2208.08373, 2022.","identifier":"arXiv:2208.08373","url":"https://arxiv.org/abs/2208.08373"},{"text":"F. Jenelten, R. Grandia, F. Farshidian, and M. Hutter. TAMOLS: Terrain-Aware Motion Optimization for Legged Systems. arXiv:2206.14049, 2022.","identifier":"arXiv:2206.14049","url":"https://arxiv.org/abs/2206.14049"},{"text":"D. Hoeller, N. Rudin, C. Choy, A. Anandkumar, and M. Hutter. Neural Scene Representation for Locomotion on Structured Terrain. arXiv:2206.08077, 2022.","identifier":"arXiv:2206.08077","url":"https://arxiv.org/abs/2206.08077"},{"text":"R. Yang, G. Yang, and X. Wang. Neural Volumetric Memory for Visual Locomotion Control. arXiv:2304.01201, 2023.","identifier":"arXiv:2304.01201","url":"https://arxiv.org/abs/2304.01201"},{"text":"Z. Fu, A. Kumar, A. Agarwal, H. Qi, J. Malik, and D. Pathak. Coupling Vision and Proprioception for Navigation of Legged Robots. arXiv:2112.02094, 2021.","identifier":"arXiv:2112.02094","url":"https://arxiv.org/abs/2112.02094"},{"text":"S. Kareer, N. Yokoyama, D. Batra, S. Ha, and J. Truong. ViNL: Visual Navigation and Locomotion Over Obstacles. arXiv:2210.14791, 2022.","identifier":"arXiv:2210.14791","url":"https://arxiv.org/abs/2210.14791"},{"text":"H. Oleynikova, Z. Taylor, M. Fehr, R. Siegwart, and J. Nieto. Voxblox: Incremental 3D Euclidean Signed Distance Fields for On-Board MAV Planning. arXiv:1611.03631, 2016.","identifier":"arXiv:1611.03631","url":"https://arxiv.org/abs/1611.03631"},{"text":"X. Zhou, Z. Wang, H. Ye, C. Xu, and F. Gao. EGO-Planner: An ESDF-free Gradient-based Local Planner for Quadrotors. arXiv:2008.08835, 2020.","identifier":"arXiv:2008.08835","url":"https://arxiv.org/abs/2008.08835"},{"text":"G. Williams, P. Drews, B. Goldfain, J. M. Rehg, and E. A. Theodorou. Information-Theoretic Model Predictive Control: Theory and Applications to Autonomous Driving. arXiv:1707.02342, 2017.","identifier":"arXiv:1707.02342","url":"https://arxiv.org/abs/1707.02342"},{"text":"J. Philion and S. Fidler. Lift, Splat, Shoot: Encoding Images From Arbitrary Camera Rigs by Implicitly Unprojecting to 3D. arXiv:2008.05625, 2020.","identifier":"arXiv:2008.05625","url":"https://arxiv.org/abs/2008.05625"},{"text":"Z. Li, W. Wang, H. Li, E. Xie, C. Sima, T. Lu, Y. Qiao, and J. Dai. BEVFormer: Learning Bird's-Eye-View Representation from Multi-Camera Images via Spatiotemporal Transformers. arXiv:2203.17270, 2022.","identifier":"arXiv:2203.17270","url":"https://arxiv.org/abs/2203.17270"},{"text":"Y. Huang, W. Zheng, Y. Zhang, J. Zhou, and J. Lu. Tri-Perspective View for Vision-Based 3D Semantic Occupancy Prediction. arXiv:2302.07817, 2023.","identifier":"arXiv:2302.07817","url":"https://arxiv.org/abs/2302.07817"},{"text":"W. Tong, C. Sima, T. Wang, L. Chen, S. Wu, H. Deng, Y. Gu, L. Lu, P. Luo, D. Lin, and H. Li. Scene as Occupancy. arXiv:2306.02851, 2023.","identifier":"arXiv:2306.02851","url":"https://arxiv.org/abs/2306.02851"},{"text":"E. R. Chan et al.. Efficient Geometry-aware 3D Generative Adversarial Networks. arXiv:2112.07945, 2021.","identifier":"arXiv:2112.07945","url":"https://arxiv.org/abs/2112.07945"},{"text":"S. Fridovich-Keil, G. Meanti, F. R. Warburg, B. Recht, and A. Kanazawa. K-Planes: Explicit Radiance Fields in Space, Time, and Appearance. arXiv:2301.10241, 2023.","identifier":"arXiv:2301.10241","url":"https://arxiv.org/abs/2301.10241"},{"text":"G. Kahn, P. Abbeel, and S. Levine. BADGR: An Autonomous Self-Supervised Learning-Based Navigation System. arXiv:2002.05700, 2020.","identifier":"arXiv:2002.05700","url":"https://arxiv.org/abs/2002.05700"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Bouman et al.. Autonomous Spot: Long-Range Autonomous Exploration of Extreme Environments with Legged Locomotion. arXiv:2010.09259, 2020.","identifier":"arXiv:2010.09259","url":"https://arxiv.org/abs/2010.09259"},{"text":"M. Tranzatto et al.. CERBERUS: Autonomous Legged and Aerial Robotic Exploration in the Tunnel and Urban Circuits of the DARPA Subterranean Challenge. arXiv:2201.07067, 2022.","identifier":"arXiv:2201.07067","url":"https://arxiv.org/abs/2201.07067"},{"text":"M. Mittal et al.. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. arXiv:2301.04195, 2023.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"V. Makoviychuk et al.. Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"J. Schulman, F. Wolski, P. Dhariwal, A. Radford, and O. Klimov. Proximal Policy Optimization Algorithms. arXiv:1707.06347, 2017.","identifier":"arXiv:1707.06347","url":"https://arxiv.org/abs/1707.06347"},{"text":"J. Truong, M. Rudolph, N. Yokoyama, S. Chernova, D. Batra, and A. Rai. Rethinking Sim2Real: Lower Fidelity Simulation Leads to Higher Sim2Real Transfer in Navigation. arXiv:2207.10821, 2022.","identifier":"arXiv:2207.10821","url":"https://arxiv.org/abs/2207.10821"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2608.00005","slug":"2608.00005","registered_at":"2026-08-10T15:23:00Z","updated_at":"2026-08-11T19:26:00Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2608.00005","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2608.00005","tex_path":"articles/2608.00005/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-08-10T17:40:00Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2608.00005"},{"kagg_id":"KAGG:2608.00004","slug":"2608.00004","title":"Safe Velocity Interface Identification Under Input Saturation","authors":[{"name":"Siyuan Cui","affiliation":"Autonomous Systems Center, National University of Defense Technology, Changsha, China"},{"name":"Kaiwen Deng","affiliation":"Autonomous Systems Center, National University of Defense Technology, Changsha, China"},{"name":"Zhenghao Liu","affiliation":"Autonomous Systems Center, National University of Defense Technology, Changsha, China"}],"affiliation":null,"abstract":"Active calibration of opaque quadruped velocity interfaces, as in GAUGE, issues planner-conditioned probe commands and fits a command-to-motion model online. Informative probes can push the robot toward the saturation boundary of the interface, where the response becomes nonlinear and potentially unsafe on rough ground. This paper adds a probe safety layer: each probe is admitted only if it keeps the robot inside a verified reachable-and-recoverable region computed from a conservative inner model, namely a linear command-response model valid below a certified saturation bound with uncertainty from a recursive least-squares confidence set. Probes are selected to maximize planner-weighted information subject to saturation, excursion, and recovery-time constraints, and the safety margin is derived from the same confidence set as the identification estimate, so the safe region expands as confidence grows. Under bounded-disturbance and inner-linearity assumptions, the analysis shows that the certified identification residual and the safety margin evolve jointly: the margin is nonincreasing, safety holds with high probability, and both quantities decay at the same rate under persistent excitation. In 120 simulated identification sessions on a Unitree Go2 in a cluttered course with an 8-degree slope, the method reached a 10.8 percent RMS identification residual versus 11.3 percent for unconstrained GAUGE-style probes, reduced the mean maximum excursion outside the safe region from 5.4 cm to 4.7 cm, and reached navigation-grade accuracy in 52 s instead of 60 s with smaller probes. In ten hardware sessions, a runtime monitor truncated one probe after a 0.04 m/s violation of the certified saturation bound.","publication_date":"2026-08-07","registered_at":"2026-08-14T11:23:00Z","updated_at":"2026-08-16T13:38:00Z","status":"published","html_path":null,"pdf_path":"articles/2608.00004/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"T. Zang, H. Huang, S. Cheng, and W. Zhang. GAUGE: Planner-Conditioned Active Calibration of Opaque Quadruped Velocity Interfaces. arXiv:2607.02841, 2026.","identifier":"arXiv:2607.02841","url":"https://arxiv.org/abs/2607.02841"},{"text":"W. Yu, J. Tan, C. K. Liu, and G. Turk. Preparing for the Unknown: Learning a Universal Policy with Online System Identification. arXiv:1702.02453, 2017.","identifier":"arXiv:1702.02453","url":"https://arxiv.org/abs/1702.02453"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"I. M. A. Nahrendra, B. Yu, and H. Myung. DreamWaQ: Learning Robust Quadrupedal Locomotion With Implicit Terrain Imagination via Deep Reinforcement Learning. arXiv:2301.10602, 2023.","identifier":"arXiv:2301.10602","url":"https://arxiv.org/abs/2301.10602"},{"text":"G. B. Margolis and P. Agrawal. Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior. arXiv:2212.03238, 2022.","identifier":"arXiv:2212.03238","url":"https://arxiv.org/abs/2212.03238"},{"text":"M. O'Connell, G. Shi, X. Shi, K. Azizzadenesheli, A. Anandkumar, Y. Yue, and S.-J. Chung. Neural-Fly Enables Rapid Learning for Agile Flight in Strong Winds. arXiv:2205.06908, 2022.","identifier":"arXiv:2205.06908","url":"https://arxiv.org/abs/2205.06908"},{"text":"J. Tan, T. Zhang, E. Coumans, A. Iscen, Y. Bai, D. Hafner, S. Bohez, and V. Vanhoucke. Sim-to-Real: Learning Agile Locomotion For Quadruped Robots. arXiv:1804.10332, 2018.","identifier":"arXiv:1804.10332","url":"https://arxiv.org/abs/1804.10332"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. arXiv:1710.06537, 2017.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"V. Makoviychuk et al.. Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal et al.. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. arXiv:2301.04195, 2023.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"G. B. Margolis, G. Yang, K. Paigwar, T. Chen, and P. Agrawal. Rapid Locomotion via Reinforcement Learning. arXiv:2205.02824, 2022.","identifier":"arXiv:2205.02824","url":"https://arxiv.org/abs/2205.02824"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. arXiv:2305.08510, 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"T. M. Moldovan and P. Abbeel. Safe Exploration in Markov Decision Processes. arXiv:1205.4810, 2012.","identifier":"arXiv:1205.4810","url":"https://arxiv.org/abs/1205.4810"},{"text":"M. Turchetta, F. Berkenkamp, and A. Krause. Safe Exploration in Finite Markov Decision Processes with Gaussian Processes. arXiv:1606.04753, 2016.","identifier":"arXiv:1606.04753","url":"https://arxiv.org/abs/1606.04753"},{"text":"F. Berkenkamp, M. Turchetta, A. P. Schoellig, and A. Krause. Safe Model-based Reinforcement Learning with Stability Guarantees. arXiv:1705.08551, 2017.","identifier":"arXiv:1705.08551","url":"https://arxiv.org/abs/1705.08551"},{"text":"T. Koller, F. Berkenkamp, M. Turchetta, and A. Krause. Learning-based Model Predictive Control for Safe Exploration. arXiv:1803.08287, 2018.","identifier":"arXiv:1803.08287","url":"https://arxiv.org/abs/1803.08287"},{"text":"T. Lew, A. Sharma, J. Harrison, A. Bylard, and M. Pavone. Safe Active Dynamics Learning and Control: A Sequential Exploration-Exploitation Framework. arXiv:2008.11700, 2020.","identifier":"arXiv:2008.11700","url":"https://arxiv.org/abs/2008.11700"},{"text":"L. Brunke, M. Greeff, A. W. Hall, Z. Yuan, S. Zhou, J. Panerati, and A. P. Schoellig. Safe Learning in Robotics: From Learning-Based Control to Safe Reinforcement Learning. arXiv:2108.06266, 2021.","identifier":"arXiv:2108.06266","url":"https://arxiv.org/abs/2108.06266"},{"text":"A. D. Ames, X. Xu, J. W. Grizzle, and P. Tabuada. Control Barrier Function Based Quadratic Programs for Safety Critical Systems. arXiv:1609.06408, 2016.","identifier":"arXiv:1609.06408","url":"https://arxiv.org/abs/1609.06408"},{"text":"A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada. Control Barrier Functions: Theory and Applications. arXiv:1903.11199, 2019.","identifier":"arXiv:1903.11199","url":"https://arxiv.org/abs/1903.11199"},{"text":"A. J. Taylor, A. Singletary, Y. Yue, and A. D. Ames. Learning for Safety-Critical Control with Control Barrier Functions. arXiv:1912.10099, 2019.","identifier":"arXiv:1912.10099","url":"https://arxiv.org/abs/1912.10099"},{"text":"R. Cheng, G. Orosz, R. M. Murray, and J. W. Burdick. End-to-End Safe Reinforcement Learning through Barrier Functions for Safety-Critical Continuous Control Tasks. arXiv:1903.08792, 2019.","identifier":"arXiv:1903.08792","url":"https://arxiv.org/abs/1903.08792"},{"text":"J. Zeng, B. Zhang, and K. Sreenath. Safety-Critical Model Predictive Control with Discrete-Time Control Barrier Function. arXiv:2007.11718, 2020.","identifier":"arXiv:2007.11718","url":"https://arxiv.org/abs/2007.11718"},{"text":"R. Grandia, A. J. Taylor, A. D. Ames, and M. Hutter. Multi-Layered Safety for Legged Robots via Control Barrier Functions and Model Predictive Control. arXiv:2011.00032, 2020.","identifier":"arXiv:2011.00032","url":"https://arxiv.org/abs/2011.00032"},{"text":"K. P. Wabersich and M. N. Zeilinger. Linear Model Predictive Safety Certification for Learning-Based Control. arXiv:1803.08552, 2018.","identifier":"arXiv:1803.08552","url":"https://arxiv.org/abs/1803.08552"},{"text":"K. P. Wabersich and M. N. Zeilinger. A Predictive Safety Filter for Learning-Based Control of Constrained Nonlinear Dynamical Systems. arXiv:1812.05506, 2018.","identifier":"arXiv:1812.05506","url":"https://arxiv.org/abs/1812.05506"},{"text":"S. Bansal, M. Chen, S. Herbert, and C. J. Tomlin. Hamilton-Jacobi Reachability: A Brief Overview and Recent Advances. arXiv:1709.07523, 2017.","identifier":"arXiv:1709.07523","url":"https://arxiv.org/abs/1709.07523"},{"text":"J. F. Fisac, A. K. Akametalu, M. N. Zeilinger, S. Kaynama, J. Gillula, and C. J. Tomlin. A General Safety Framework for Learning-Based Control in Uncertain Robotic Systems. arXiv:1705.01292, 2017.","identifier":"arXiv:1705.01292","url":"https://arxiv.org/abs/1705.01292"},{"text":"K.-C. Hsu, H. Hu, and J. F. Fisac. The Safety Filter: A Unified View of Safety-Critical Control in Autonomous Systems. arXiv:2309.05837, 2023.","identifier":"arXiv:2309.05837","url":"https://arxiv.org/abs/2309.05837"},{"text":"M. Alshiekh, R. Bloem, R. Ehlers, B. Könighofer, S. Niekum, and U. Topcu. Safe Reinforcement Learning via Shielding. arXiv:1708.08611, 2017.","identifier":"arXiv:1708.08611","url":"https://arxiv.org/abs/1708.08611"},{"text":"T. He, C. Zhang, W. Xiao, G. He, C. Liu, and G. Shi. Agile But Safe: Learning Collision-Free High-Speed Legged Locomotion. arXiv:2401.17583, 2024.","identifier":"arXiv:2401.17583","url":"https://arxiv.org/abs/2401.17583"},{"text":"T.-Y. Yang, T. Zhang, L. Luu, S. Ha, J. Tan, and W. Yu. Safe Reinforcement Learning for Legged Locomotion. arXiv:2203.02638, 2022.","identifier":"arXiv:2203.02638","url":"https://arxiv.org/abs/2203.02638"},{"text":"S. Dean, H. Mania, N. Matni, B. Recht, and S. Tu. On the Sample Complexity of the Linear Quadratic Regulator. arXiv:1710.01688, 2017.","identifier":"arXiv:1710.01688","url":"https://arxiv.org/abs/1710.01688"},{"text":"M. Simchowitz, H. Mania, S. Tu, M. I. Jordan, and B. Recht. Learning Without Mixing: Towards a Sharp Analysis of Linear System Identification. arXiv:1802.08334, 2018.","identifier":"arXiv:1802.08334","url":"https://arxiv.org/abs/1802.08334"},{"text":"A. Wagenmaker and K. Jamieson. Active Learning for Identification of Linear Dynamical Systems. arXiv:2002.00495, 2020.","identifier":"arXiv:2002.00495","url":"https://arxiv.org/abs/2002.00495"},{"text":"H. Mania, M. I. Jordan, and B. Recht. Active Learning for Nonlinear System Identification with Guarantees. arXiv:2006.10277, 2020.","identifier":"arXiv:2006.10277","url":"https://arxiv.org/abs/2006.10277"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2608.00004","metadata":{"title":"Safe Velocity Interface Identification Under Input Saturation","authors":[{"name":"Siyuan Cui","affiliation":"Autonomous Systems Center, National University of Defense Technology, Changsha, China"},{"name":"Kaiwen Deng","affiliation":"Autonomous Systems Center, National University of Defense Technology, Changsha, China"},{"name":"Zhenghao Liu","affiliation":"Autonomous Systems Center, National University of Defense Technology, Changsha, China"}],"publication_date":"2026-08-07","abstract":"Active calibration of opaque quadruped velocity interfaces, as in GAUGE, issues planner-conditioned probe commands and fits a command-to-motion model online. Informative probes can push the robot toward the saturation boundary of the interface, where the response becomes nonlinear and potentially unsafe on rough ground. This paper adds a probe safety layer: each probe is admitted only if it keeps the robot inside a verified reachable-and-recoverable region computed from a conservative inner model, namely a linear command-response model valid below a certified saturation bound with uncertainty from a recursive least-squares confidence set. Probes are selected to maximize planner-weighted information subject to saturation, excursion, and recovery-time constraints, and the safety margin is derived from the same confidence set as the identification estimate, so the safe region expands as confidence grows. Under bounded-disturbance and inner-linearity assumptions, the analysis shows that the certified identification residual and the safety margin evolve jointly: the margin is nonincreasing, safety holds with high probability, and both quantities decay at the same rate under persistent excitation. In 120 simulated identification sessions on a Unitree Go2 in a cluttered course with an 8-degree slope, the method reached a 10.8 percent RMS identification residual versus 11.3 percent for unconstrained GAUGE-style probes, reduced the mean maximum excursion outside the safe region from 5.4 cm to 4.7 cm, and reached navigation-grade accuracy in 52 s instead of 60 s with smaller probes. In ten hardware sessions, a runtime monitor truncated one probe after a 0.04 m/s violation of the certified saturation bound.","references":[{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"T. Zang, H. Huang, S. Cheng, and W. Zhang. GAUGE: Planner-Conditioned Active Calibration of Opaque Quadruped Velocity Interfaces. arXiv:2607.02841, 2026.","identifier":"arXiv:2607.02841","url":"https://arxiv.org/abs/2607.02841"},{"text":"W. Yu, J. Tan, C. K. Liu, and G. Turk. Preparing for the Unknown: Learning a Universal Policy with Online System Identification. arXiv:1702.02453, 2017.","identifier":"arXiv:1702.02453","url":"https://arxiv.org/abs/1702.02453"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"I. M. A. Nahrendra, B. Yu, and H. Myung. DreamWaQ: Learning Robust Quadrupedal Locomotion With Implicit Terrain Imagination via Deep Reinforcement Learning. arXiv:2301.10602, 2023.","identifier":"arXiv:2301.10602","url":"https://arxiv.org/abs/2301.10602"},{"text":"G. B. Margolis and P. Agrawal. Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior. arXiv:2212.03238, 2022.","identifier":"arXiv:2212.03238","url":"https://arxiv.org/abs/2212.03238"},{"text":"M. O'Connell, G. Shi, X. Shi, K. Azizzadenesheli, A. Anandkumar, Y. Yue, and S.-J. Chung. Neural-Fly Enables Rapid Learning for Agile Flight in Strong Winds. arXiv:2205.06908, 2022.","identifier":"arXiv:2205.06908","url":"https://arxiv.org/abs/2205.06908"},{"text":"J. Tan, T. Zhang, E. Coumans, A. Iscen, Y. Bai, D. Hafner, S. Bohez, and V. Vanhoucke. Sim-to-Real: Learning Agile Locomotion For Quadruped Robots. arXiv:1804.10332, 2018.","identifier":"arXiv:1804.10332","url":"https://arxiv.org/abs/1804.10332"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. arXiv:1710.06537, 2017.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"V. Makoviychuk et al.. Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal et al.. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. arXiv:2301.04195, 2023.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"G. B. Margolis, G. Yang, K. Paigwar, T. Chen, and P. Agrawal. Rapid Locomotion via Reinforcement Learning. arXiv:2205.02824, 2022.","identifier":"arXiv:2205.02824","url":"https://arxiv.org/abs/2205.02824"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. arXiv:2305.08510, 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"T. M. Moldovan and P. Abbeel. Safe Exploration in Markov Decision Processes. arXiv:1205.4810, 2012.","identifier":"arXiv:1205.4810","url":"https://arxiv.org/abs/1205.4810"},{"text":"M. Turchetta, F. Berkenkamp, and A. Krause. Safe Exploration in Finite Markov Decision Processes with Gaussian Processes. arXiv:1606.04753, 2016.","identifier":"arXiv:1606.04753","url":"https://arxiv.org/abs/1606.04753"},{"text":"F. Berkenkamp, M. Turchetta, A. P. Schoellig, and A. Krause. Safe Model-based Reinforcement Learning with Stability Guarantees. arXiv:1705.08551, 2017.","identifier":"arXiv:1705.08551","url":"https://arxiv.org/abs/1705.08551"},{"text":"T. Koller, F. Berkenkamp, M. Turchetta, and A. Krause. Learning-based Model Predictive Control for Safe Exploration. arXiv:1803.08287, 2018.","identifier":"arXiv:1803.08287","url":"https://arxiv.org/abs/1803.08287"},{"text":"T. Lew, A. Sharma, J. Harrison, A. Bylard, and M. Pavone. Safe Active Dynamics Learning and Control: A Sequential Exploration-Exploitation Framework. arXiv:2008.11700, 2020.","identifier":"arXiv:2008.11700","url":"https://arxiv.org/abs/2008.11700"},{"text":"L. Brunke, M. Greeff, A. W. Hall, Z. Yuan, S. Zhou, J. Panerati, and A. P. Schoellig. Safe Learning in Robotics: From Learning-Based Control to Safe Reinforcement Learning. arXiv:2108.06266, 2021.","identifier":"arXiv:2108.06266","url":"https://arxiv.org/abs/2108.06266"},{"text":"A. D. Ames, X. Xu, J. W. Grizzle, and P. Tabuada. Control Barrier Function Based Quadratic Programs for Safety Critical Systems. arXiv:1609.06408, 2016.","identifier":"arXiv:1609.06408","url":"https://arxiv.org/abs/1609.06408"},{"text":"A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada. Control Barrier Functions: Theory and Applications. arXiv:1903.11199, 2019.","identifier":"arXiv:1903.11199","url":"https://arxiv.org/abs/1903.11199"},{"text":"A. J. Taylor, A. Singletary, Y. Yue, and A. D. Ames. Learning for Safety-Critical Control with Control Barrier Functions. arXiv:1912.10099, 2019.","identifier":"arXiv:1912.10099","url":"https://arxiv.org/abs/1912.10099"},{"text":"R. Cheng, G. Orosz, R. M. Murray, and J. W. Burdick. End-to-End Safe Reinforcement Learning through Barrier Functions for Safety-Critical Continuous Control Tasks. arXiv:1903.08792, 2019.","identifier":"arXiv:1903.08792","url":"https://arxiv.org/abs/1903.08792"},{"text":"J. Zeng, B. Zhang, and K. Sreenath. Safety-Critical Model Predictive Control with Discrete-Time Control Barrier Function. arXiv:2007.11718, 2020.","identifier":"arXiv:2007.11718","url":"https://arxiv.org/abs/2007.11718"},{"text":"R. Grandia, A. J. Taylor, A. D. Ames, and M. Hutter. Multi-Layered Safety for Legged Robots via Control Barrier Functions and Model Predictive Control. arXiv:2011.00032, 2020.","identifier":"arXiv:2011.00032","url":"https://arxiv.org/abs/2011.00032"},{"text":"K. P. Wabersich and M. N. Zeilinger. Linear Model Predictive Safety Certification for Learning-Based Control. arXiv:1803.08552, 2018.","identifier":"arXiv:1803.08552","url":"https://arxiv.org/abs/1803.08552"},{"text":"K. P. Wabersich and M. N. Zeilinger. A Predictive Safety Filter for Learning-Based Control of Constrained Nonlinear Dynamical Systems. arXiv:1812.05506, 2018.","identifier":"arXiv:1812.05506","url":"https://arxiv.org/abs/1812.05506"},{"text":"S. Bansal, M. Chen, S. Herbert, and C. J. Tomlin. Hamilton-Jacobi Reachability: A Brief Overview and Recent Advances. arXiv:1709.07523, 2017.","identifier":"arXiv:1709.07523","url":"https://arxiv.org/abs/1709.07523"},{"text":"J. F. Fisac, A. K. Akametalu, M. N. Zeilinger, S. Kaynama, J. Gillula, and C. J. Tomlin. A General Safety Framework for Learning-Based Control in Uncertain Robotic Systems. arXiv:1705.01292, 2017.","identifier":"arXiv:1705.01292","url":"https://arxiv.org/abs/1705.01292"},{"text":"K.-C. Hsu, H. Hu, and J. F. Fisac. The Safety Filter: A Unified View of Safety-Critical Control in Autonomous Systems. arXiv:2309.05837, 2023.","identifier":"arXiv:2309.05837","url":"https://arxiv.org/abs/2309.05837"},{"text":"M. Alshiekh, R. Bloem, R. Ehlers, B. Könighofer, S. Niekum, and U. Topcu. Safe Reinforcement Learning via Shielding. arXiv:1708.08611, 2017.","identifier":"arXiv:1708.08611","url":"https://arxiv.org/abs/1708.08611"},{"text":"T. He, C. Zhang, W. Xiao, G. He, C. Liu, and G. Shi. Agile But Safe: Learning Collision-Free High-Speed Legged Locomotion. arXiv:2401.17583, 2024.","identifier":"arXiv:2401.17583","url":"https://arxiv.org/abs/2401.17583"},{"text":"T.-Y. Yang, T. Zhang, L. Luu, S. Ha, J. Tan, and W. Yu. Safe Reinforcement Learning for Legged Locomotion. arXiv:2203.02638, 2022.","identifier":"arXiv:2203.02638","url":"https://arxiv.org/abs/2203.02638"},{"text":"S. Dean, H. Mania, N. Matni, B. Recht, and S. Tu. On the Sample Complexity of the Linear Quadratic Regulator. arXiv:1710.01688, 2017.","identifier":"arXiv:1710.01688","url":"https://arxiv.org/abs/1710.01688"},{"text":"M. Simchowitz, H. Mania, S. Tu, M. I. Jordan, and B. Recht. Learning Without Mixing: Towards a Sharp Analysis of Linear System Identification. arXiv:1802.08334, 2018.","identifier":"arXiv:1802.08334","url":"https://arxiv.org/abs/1802.08334"},{"text":"A. Wagenmaker and K. Jamieson. Active Learning for Identification of Linear Dynamical Systems. arXiv:2002.00495, 2020.","identifier":"arXiv:2002.00495","url":"https://arxiv.org/abs/2002.00495"},{"text":"H. Mania, M. I. Jordan, and B. Recht. Active Learning for Nonlinear System Identification with Guarantees. arXiv:2006.10277, 2020.","identifier":"arXiv:2006.10277","url":"https://arxiv.org/abs/2006.10277"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2608.00004","slug":"2608.00004","registered_at":"2026-08-14T11:23:00Z","updated_at":"2026-08-16T13:38:00Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2608.00004","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2608.00004","tex_path":"articles/2608.00004/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-08-14T13:40:00Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2608.00004"},{"kagg_id":"KAGG:2608.00003","slug":"2608.00003","title":"Embodiment-Conditioned Goal Navigation for Mixed Quadruped-Humanoid Groups","authors":[{"name":"Ivan Sokolov","affiliation":"Robotics Institute, Georgia Institute of Technology, Atlanta, GA, USA"},{"name":"Yating Feng","affiliation":"Robotics Institute, Georgia Institute of Technology, Atlanta, GA, USA"},{"name":"Rachel Kim","affiliation":"Robotics Institute, Georgia Institute of Technology, Atlanta, GA, USA"},{"name":"Xiaotong Zhang","affiliation":"Robotics and Autonomous Systems Lab, Shanghai Jiao Tong University, Shanghai, China"}],"affiliation":null,"abstract":"Mixed groups of quadruped and humanoid robots that navigate toward a common goal differ in footprint, speed envelope, and sensor viewpoint, so a single shared navigation policy is ill-posed unless these differences are made explicit. This paper presents an embodiment-conditioned goal-navigation policy for such groups. Building on the configuration descriptor of AgniNav, each agent receives a normalized embodiment vector encoding footprint, velocity limits, and sensor height, together with egocentric perception, its goal, and the group-relative states and descriptors of its neighbors. A single network maps these inputs to normalized motion commands that are scaled by the agent's own limits and passed through a lightweight resolver enforcing pairwise separation, velocity-bound, and deadlock-breaking invariants. The policy is trained in Isaac Lab with embodiment randomization over the full fleet envelope, while one morphology per platform family is withheld for zero-shot evaluation. Across 25 goal-navigation scenes with 5 seeds and 40 episodes each, groups of two to six agents reach the goal in 88.1 percent of episodes with 0.27 collisions per episode and a 95th-percentile goal time of 41 s, compared with 82.9 percent for per-embodiment policies and 84.0 percent for a shared policy that receives the embodiment as a constant input. On held-out morphologies, zero-shot success is 79.4 percent, versus 68.4 percent and 68.7 percent for these two baselines. In ablations, removing embodiment randomization causes the largest drop. Remaining failures concentrate in doorway clearance and in corridor throughput for six-agent groups.","publication_date":"2026-08-05","registered_at":"2026-08-08T09:23:00Z","updated_at":"2026-08-10T12:09:00Z","status":"published","html_path":null,"pdf_path":"articles/2608.00003/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Sridhar, D. Shah, C. Glossop, and S. Levine. NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration. arXiv:2310.07896, 2023.","identifier":"arXiv:2310.07896","url":"https://arxiv.org/abs/2310.07896"},{"text":"A. Eftekhar et al.. The One RING: a Robotic Indoor Navigation Generalist. arXiv:2412.14401, 2024.","identifier":"arXiv:2412.14401","url":"https://arxiv.org/abs/2412.14401"},{"text":"J. Zhang et al.. Embodied Navigation Foundation Model. arXiv:2509.12129, 2025.","identifier":"arXiv:2509.12129","url":"https://arxiv.org/abs/2509.12129"},{"text":"T. Chen, A. Murali, and A. Gupta. Hardware Conditioned Policies for Multi-Robot Transfer Learning. arXiv:1811.09864, 2018.","identifier":"arXiv:1811.09864","url":"https://arxiv.org/abs/1811.09864"},{"text":"W. Huang, I. Mordatch, and D. Pathak. One Policy to Control Them All: Shared Modular Policies for Agent-Agnostic Control. arXiv:2007.04976, 2020.","identifier":"arXiv:2007.04976","url":"https://arxiv.org/abs/2007.04976"},{"text":"J. Yang, C. Glossop, A. Bhorkar, D. Shah, Q. Vuong, C. Finn, D. Sadigh, and S. Levine. Pushing the Limits of Cross-Embodiment Learning for Manipulation and Navigation. arXiv:2402.19432, 2024.","identifier":"arXiv:2402.19432","url":"https://arxiv.org/abs/2402.19432"},{"text":"R. Doshi, H. Walke, O. Mees, S. Dasari, and S. Levine. Scaling Cross-Embodied Learning: One Policy for Manipulation, Navigation, Locomotion and Aviation. arXiv:2408.11812, 2024.","identifier":"arXiv:2408.11812","url":"https://arxiv.org/abs/2408.11812"},{"text":"Open X-Embodiment Collaboration et al.. Open X-Embodiment: Robotic Learning Datasets and RT-X Models. arXiv:2310.08864, 2023.","identifier":"arXiv:2310.08864","url":"https://arxiv.org/abs/2310.08864"},{"text":"G. Feng et al.. GenLoco: Generalized Locomotion Controllers for Quadrupedal Robots. arXiv:2209.05309, 2022.","identifier":"arXiv:2209.05309","url":"https://arxiv.org/abs/2209.05309"},{"text":"M. Shafiee, G. Bellegarda, and A. Ijspeert. ManyQuadrupeds: Learning a Single Locomotion Policy for Diverse Quadruped Robots. arXiv:2310.10486, 2023.","identifier":"arXiv:2310.10486","url":"https://arxiv.org/abs/2310.10486"},{"text":"N. Bohlinger, G. Czechmanowski, M. Krupka, P. Kicki, K. Walas, J. Peters, and D. Tateo. One Policy to Run Them All: an End-to-end Learning Approach to Multi-Embodiment Locomotion. arXiv:2409.06366, 2024.","identifier":"arXiv:2409.06366","url":"https://arxiv.org/abs/2409.06366"},{"text":"Y. F. Chen, M. Liu, M. Everett, and J. P. How. Decentralized Non-communicating Multiagent Collision Avoidance with Deep Reinforcement Learning. arXiv:1609.07845, 2016.","identifier":"arXiv:1609.07845","url":"https://arxiv.org/abs/1609.07845"},{"text":"M. Everett, Y. F. Chen, and J. P. How. Motion Planning Among Dynamic, Decision-Making Agents with Deep Reinforcement Learning. arXiv:1805.01956, 2018.","identifier":"arXiv:1805.01956","url":"https://arxiv.org/abs/1805.01956"},{"text":"P. Long, T. Fan, X. Liao, W. Liu, H. Zhang, and J. Pan. Towards Optimally Decentralized Multi-Robot Collision Avoidance via Deep Reinforcement Learning. arXiv:1709.10082, 2017.","identifier":"arXiv:1709.10082","url":"https://arxiv.org/abs/1709.10082"},{"text":"T. Fan, P. Long, W. Liu, and J. Pan. Distributed Multi-Robot Collision Avoidance via Deep Reinforcement Learning for Navigation in Complex Scenarios. arXiv:1808.03841, 2018.","identifier":"arXiv:1808.03841","url":"https://arxiv.org/abs/1808.03841"},{"text":"C. Chen, Y. Liu, S. Kreiss, and A. Alahi. Crowd-Robot Interaction: Crowd-aware Robot Navigation with Attention-based Deep Reinforcement Learning. arXiv:1809.08835, 2018.","identifier":"arXiv:1809.08835","url":"https://arxiv.org/abs/1809.08835"},{"text":"G. Sartoretti, J. Kerr, Y. Shi, G. Wagner, T. K. S. Kumar, S. Koenig, and H. Choset. PRIMAL: Pathfinding via Reinforcement and Imitation Multi-Agent Learning. arXiv:1809.03531, 2018.","identifier":"arXiv:1809.03531","url":"https://arxiv.org/abs/1809.03531"},{"text":"C. Yu, A. Velu, E. Vinitsky, J. Gao, Y. Wang, A. Bayen, and Y. Wu. The Surprising Effectiveness of PPO in Cooperative, Multi-Agent Games. arXiv:2103.01955, 2021.","identifier":"arXiv:2103.01955","url":"https://arxiv.org/abs/2103.01955"},{"text":"R. Lowe, Y. Wu, A. Tamar, J. Harb, P. Abbeel, and I. Mordatch. Multi-Agent Actor-Critic for Mixed Cooperative-Competitive Environments. arXiv:1706.02275, 2017.","identifier":"arXiv:1706.02275","url":"https://arxiv.org/abs/1706.02275"},{"text":"A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada. Control Barrier Functions: Theory and Applications. arXiv:1903.11199, 2019.","identifier":"arXiv:1903.11199","url":"https://arxiv.org/abs/1903.11199"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. arXiv:1703.06907, 2017.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. arXiv:1710.06537, 2017.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. arXiv:2211.07638, 2022.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"I. Radosavovic, T. Xiao, B. Zhang, T. Darrell, J. Malik, and K. Sreenath. Real-World Humanoid Locomotion with Reinforcement Learning. arXiv:2303.03381, 2023.","identifier":"arXiv:2303.03381","url":"https://arxiv.org/abs/2303.03381"},{"text":"M. Mittal et al.. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. arXiv:2301.04195, 2023.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"V. Makoviychuk et al.. Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"J. Schulman, F. Wolski, P. Dhariwal, A. Radford, and O. Klimov. Proximal Policy Optimization Algorithms. arXiv:1707.06347, 2017.","identifier":"arXiv:1707.06347","url":"https://arxiv.org/abs/1707.06347"},{"text":"E. Perez, F. Strub, H. de Vries, V. Dumoulin, and A. Courville. FiLM: Visual Reasoning with a General Conditioning Layer. arXiv:1709.07871, 2017.","identifier":"arXiv:1709.07871","url":"https://arxiv.org/abs/1709.07871"},{"text":"M. Zaheer, S. Kottur, S. Ravanbakhsh, B. Poczos, R. Salakhutdinov, and A. Smola. Deep Sets. arXiv:1703.06114, 2017.","identifier":"arXiv:1703.06114","url":"https://arxiv.org/abs/1703.06114"},{"text":"A. Vaswani et al.. Attention Is All You Need. arXiv:1706.03762, 2017.","identifier":"arXiv:1706.03762","url":"https://arxiv.org/abs/1706.03762"},{"text":"P. Anderson et al.. On Evaluation of Embodied Navigation Agents. arXiv:1807.06757, 2018.","identifier":"arXiv:1807.06757","url":"https://arxiv.org/abs/1807.06757"},{"text":"N. Yokoyama, S. Ha, and D. Batra. Success Weighted by Completion Time: A Dynamics-Aware Evaluation Criteria for Embodied Navigation. arXiv:2103.08022, 2021.","identifier":"arXiv:2103.08022","url":"https://arxiv.org/abs/2103.08022"},{"text":"A. Kadian et al.. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance? arXiv:1912.06321, 2019.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2608.00003","metadata":{"title":"Embodiment-Conditioned Goal Navigation for Mixed Quadruped-Humanoid Groups","authors":[{"name":"Ivan Sokolov","affiliation":"Robotics Institute, Georgia Institute of Technology, Atlanta, GA, USA"},{"name":"Yating Feng","affiliation":"Robotics Institute, Georgia Institute of Technology, Atlanta, GA, USA"},{"name":"Rachel Kim","affiliation":"Robotics Institute, Georgia Institute of Technology, Atlanta, GA, USA"},{"name":"Xiaotong Zhang","affiliation":"Robotics and Autonomous Systems Lab, Shanghai Jiao Tong University, Shanghai, China"}],"publication_date":"2026-08-05","abstract":"Mixed groups of quadruped and humanoid robots that navigate toward a common goal differ in footprint, speed envelope, and sensor viewpoint, so a single shared navigation policy is ill-posed unless these differences are made explicit. This paper presents an embodiment-conditioned goal-navigation policy for such groups. Building on the configuration descriptor of AgniNav, each agent receives a normalized embodiment vector encoding footprint, velocity limits, and sensor height, together with egocentric perception, its goal, and the group-relative states and descriptors of its neighbors. A single network maps these inputs to normalized motion commands that are scaled by the agent's own limits and passed through a lightweight resolver enforcing pairwise separation, velocity-bound, and deadlock-breaking invariants. The policy is trained in Isaac Lab with embodiment randomization over the full fleet envelope, while one morphology per platform family is withheld for zero-shot evaluation. Across 25 goal-navigation scenes with 5 seeds and 40 episodes each, groups of two to six agents reach the goal in 88.1 percent of episodes with 0.27 collisions per episode and a 95th-percentile goal time of 41 s, compared with 82.9 percent for per-embodiment policies and 84.0 percent for a shared policy that receives the embodiment as a constant input. On held-out morphologies, zero-shot success is 79.4 percent, versus 68.4 percent and 68.7 percent for these two baselines. In ablations, removing embodiment randomization causes the largest drop. Remaining failures concentrate in doorway clearance and in corridor throughput for six-agent groups.","references":[{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Sridhar, D. Shah, C. Glossop, and S. Levine. NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration. arXiv:2310.07896, 2023.","identifier":"arXiv:2310.07896","url":"https://arxiv.org/abs/2310.07896"},{"text":"A. Eftekhar et al.. The One RING: a Robotic Indoor Navigation Generalist. arXiv:2412.14401, 2024.","identifier":"arXiv:2412.14401","url":"https://arxiv.org/abs/2412.14401"},{"text":"J. Zhang et al.. Embodied Navigation Foundation Model. arXiv:2509.12129, 2025.","identifier":"arXiv:2509.12129","url":"https://arxiv.org/abs/2509.12129"},{"text":"T. Chen, A. Murali, and A. Gupta. Hardware Conditioned Policies for Multi-Robot Transfer Learning. arXiv:1811.09864, 2018.","identifier":"arXiv:1811.09864","url":"https://arxiv.org/abs/1811.09864"},{"text":"W. Huang, I. Mordatch, and D. Pathak. One Policy to Control Them All: Shared Modular Policies for Agent-Agnostic Control. arXiv:2007.04976, 2020.","identifier":"arXiv:2007.04976","url":"https://arxiv.org/abs/2007.04976"},{"text":"J. Yang, C. Glossop, A. Bhorkar, D. Shah, Q. Vuong, C. Finn, D. Sadigh, and S. Levine. Pushing the Limits of Cross-Embodiment Learning for Manipulation and Navigation. arXiv:2402.19432, 2024.","identifier":"arXiv:2402.19432","url":"https://arxiv.org/abs/2402.19432"},{"text":"R. Doshi, H. Walke, O. Mees, S. Dasari, and S. Levine. Scaling Cross-Embodied Learning: One Policy for Manipulation, Navigation, Locomotion and Aviation. arXiv:2408.11812, 2024.","identifier":"arXiv:2408.11812","url":"https://arxiv.org/abs/2408.11812"},{"text":"Open X-Embodiment Collaboration et al.. Open X-Embodiment: Robotic Learning Datasets and RT-X Models. arXiv:2310.08864, 2023.","identifier":"arXiv:2310.08864","url":"https://arxiv.org/abs/2310.08864"},{"text":"G. Feng et al.. GenLoco: Generalized Locomotion Controllers for Quadrupedal Robots. arXiv:2209.05309, 2022.","identifier":"arXiv:2209.05309","url":"https://arxiv.org/abs/2209.05309"},{"text":"M. Shafiee, G. Bellegarda, and A. Ijspeert. ManyQuadrupeds: Learning a Single Locomotion Policy for Diverse Quadruped Robots. arXiv:2310.10486, 2023.","identifier":"arXiv:2310.10486","url":"https://arxiv.org/abs/2310.10486"},{"text":"N. Bohlinger, G. Czechmanowski, M. Krupka, P. Kicki, K. Walas, J. Peters, and D. Tateo. One Policy to Run Them All: an End-to-end Learning Approach to Multi-Embodiment Locomotion. arXiv:2409.06366, 2024.","identifier":"arXiv:2409.06366","url":"https://arxiv.org/abs/2409.06366"},{"text":"Y. F. Chen, M. Liu, M. Everett, and J. P. How. Decentralized Non-communicating Multiagent Collision Avoidance with Deep Reinforcement Learning. arXiv:1609.07845, 2016.","identifier":"arXiv:1609.07845","url":"https://arxiv.org/abs/1609.07845"},{"text":"M. Everett, Y. F. Chen, and J. P. How. Motion Planning Among Dynamic, Decision-Making Agents with Deep Reinforcement Learning. arXiv:1805.01956, 2018.","identifier":"arXiv:1805.01956","url":"https://arxiv.org/abs/1805.01956"},{"text":"P. Long, T. Fan, X. Liao, W. Liu, H. Zhang, and J. Pan. Towards Optimally Decentralized Multi-Robot Collision Avoidance via Deep Reinforcement Learning. arXiv:1709.10082, 2017.","identifier":"arXiv:1709.10082","url":"https://arxiv.org/abs/1709.10082"},{"text":"T. Fan, P. Long, W. Liu, and J. Pan. Distributed Multi-Robot Collision Avoidance via Deep Reinforcement Learning for Navigation in Complex Scenarios. arXiv:1808.03841, 2018.","identifier":"arXiv:1808.03841","url":"https://arxiv.org/abs/1808.03841"},{"text":"C. Chen, Y. Liu, S. Kreiss, and A. Alahi. Crowd-Robot Interaction: Crowd-aware Robot Navigation with Attention-based Deep Reinforcement Learning. arXiv:1809.08835, 2018.","identifier":"arXiv:1809.08835","url":"https://arxiv.org/abs/1809.08835"},{"text":"G. Sartoretti, J. Kerr, Y. Shi, G. Wagner, T. K. S. Kumar, S. Koenig, and H. Choset. PRIMAL: Pathfinding via Reinforcement and Imitation Multi-Agent Learning. arXiv:1809.03531, 2018.","identifier":"arXiv:1809.03531","url":"https://arxiv.org/abs/1809.03531"},{"text":"C. Yu, A. Velu, E. Vinitsky, J. Gao, Y. Wang, A. Bayen, and Y. Wu. The Surprising Effectiveness of PPO in Cooperative, Multi-Agent Games. arXiv:2103.01955, 2021.","identifier":"arXiv:2103.01955","url":"https://arxiv.org/abs/2103.01955"},{"text":"R. Lowe, Y. Wu, A. Tamar, J. Harb, P. Abbeel, and I. Mordatch. Multi-Agent Actor-Critic for Mixed Cooperative-Competitive Environments. arXiv:1706.02275, 2017.","identifier":"arXiv:1706.02275","url":"https://arxiv.org/abs/1706.02275"},{"text":"A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada. Control Barrier Functions: Theory and Applications. arXiv:1903.11199, 2019.","identifier":"arXiv:1903.11199","url":"https://arxiv.org/abs/1903.11199"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. arXiv:1703.06907, 2017.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. arXiv:1710.06537, 2017.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. arXiv:2211.07638, 2022.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"I. Radosavovic, T. Xiao, B. Zhang, T. Darrell, J. Malik, and K. Sreenath. Real-World Humanoid Locomotion with Reinforcement Learning. arXiv:2303.03381, 2023.","identifier":"arXiv:2303.03381","url":"https://arxiv.org/abs/2303.03381"},{"text":"M. Mittal et al.. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. arXiv:2301.04195, 2023.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"V. Makoviychuk et al.. Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"J. Schulman, F. Wolski, P. Dhariwal, A. Radford, and O. Klimov. Proximal Policy Optimization Algorithms. arXiv:1707.06347, 2017.","identifier":"arXiv:1707.06347","url":"https://arxiv.org/abs/1707.06347"},{"text":"E. Perez, F. Strub, H. de Vries, V. Dumoulin, and A. Courville. FiLM: Visual Reasoning with a General Conditioning Layer. arXiv:1709.07871, 2017.","identifier":"arXiv:1709.07871","url":"https://arxiv.org/abs/1709.07871"},{"text":"M. Zaheer, S. Kottur, S. Ravanbakhsh, B. Poczos, R. Salakhutdinov, and A. Smola. Deep Sets. arXiv:1703.06114, 2017.","identifier":"arXiv:1703.06114","url":"https://arxiv.org/abs/1703.06114"},{"text":"A. Vaswani et al.. Attention Is All You Need. arXiv:1706.03762, 2017.","identifier":"arXiv:1706.03762","url":"https://arxiv.org/abs/1706.03762"},{"text":"P. Anderson et al.. On Evaluation of Embodied Navigation Agents. arXiv:1807.06757, 2018.","identifier":"arXiv:1807.06757","url":"https://arxiv.org/abs/1807.06757"},{"text":"N. Yokoyama, S. Ha, and D. Batra. Success Weighted by Completion Time: A Dynamics-Aware Evaluation Criteria for Embodied Navigation. arXiv:2103.08022, 2021.","identifier":"arXiv:2103.08022","url":"https://arxiv.org/abs/2103.08022"},{"text":"A. Kadian et al.. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance? arXiv:1912.06321, 2019.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2608.00003","slug":"2608.00003","registered_at":"2026-08-08T09:23:00Z","updated_at":"2026-08-10T12:09:00Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2608.00003","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2608.00003","tex_path":"articles/2608.00003/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-08-08T11:40:00Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2608.00003"},{"kagg_id":"KAGG:2608.00002","slug":"2608.00002","title":"Latent Motion Tokens for Future-Range Prediction in Quadruped Obstacle Avoidance","authors":[{"name":"Marcus Hill","affiliation":"Robotics Institute, Georgia Institute of Technology, Atlanta, GA, USA"},{"name":"Ivan Sokolov","affiliation":"Robotics Institute, Georgia Institute of Technology, Atlanta, GA, USA"},{"name":"Yating Feng","affiliation":"Robotics Institute, Georgia Institute of Technology, Atlanta, GA, USA"},{"name":"Rachel Kim","affiliation":"Robotics Institute, Georgia Institute of Technology, Atlanta, GA, USA"}],"affiliation":null,"abstract":"Learned range predictors for legged-robot navigation usually regress the free distance along a fixed set of rays. This output format is difficult to condition on the motion the robot intends to execute, and it forces the planner and the reactive safety layer to share a single ray resolution. This paper encodes the predicted future range field of a quadruped as a small set of latent motion tokens, one per cluster of control-aligned directions. A transformer generates the tokens from recent range history and the planned motion, and a lightweight query decoder returns the future free range for any ray, horizon, and candidate motion. Because queries are resolved against the tokens rather than a fixed output layer, the same predictor serves a 2 Hz planner issuing coarse queries and a 50 Hz reactive layer issuing dense queries. In Isaac Lab experiments with ANYmal in 20 cluttered obstacle-avoidance scenes, with 5 seeds and 30 trials per scene and seed, the token predictor attains a range prediction error of 0.87 m, a 95th-percentile query latency of 13.6 ms on the onboard Orin computer, and an avoidance success rate of 91.8 percent, compared with 0.91 m, 14.4 ms, and 79.8 percent for a FutureRay-style direct-regression baseline, while using 1.2 million instead of 1.3 million parameters. Ablations indicate that motion-plan conditioning and the number of tokens matter most, and the recorded failure cases identify occluded turns and planner stalls as remaining weaknesses.","publication_date":"2026-08-04","registered_at":"2026-08-06T17:00:00Z","updated_at":"2026-08-07T18:38:00Z","status":"published","html_path":null,"pdf_path":"articles/2608.00002/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"T. Zang, S. Wang, H. Huang, and S. Cheng. FutureRay: Control-Aligned Future Range for Agile Quadruped Navigation. arXiv:2608.01197, 2026.","identifier":"arXiv:2608.01197","url":"https://arxiv.org/abs/2608.01197"},{"text":"V. Makoviychuk et al. Isaac Gym: High Performance GPU-Based Physics Simulation for Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal et al. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. arXiv:2301.04195, 2023.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"R. Yang, M. Zhang, N. Hansen, H. Xu, and X. Wang. Learning Vision-Guided Quadrupedal Locomotion End-to-End with Cross-Modal Transformers. arXiv:2107.03996, 2021.","identifier":"arXiv:2107.03996","url":"https://arxiv.org/abs/2107.03996"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. arXiv:2211.07638, 2022.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. arXiv:2309.14341, 2023.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. arXiv:2309.05665, 2023.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, L. Wellhausen, F. Farshidian, and M. Hutter. Learning a State Representation and Navigation in Cluttered and Dynamic Environments. arXiv:2103.04351, 2021.","identifier":"arXiv:2103.04351","url":"https://arxiv.org/abs/2103.04351"},{"text":"J. Truong, D. Yarats, T. Li, F. Meier, S. Chernova, D. Batra, and A. Rai. Learning Navigation Skills for Legged Robots with Learned Robot Embeddings. arXiv:2011.12255, 2020.","identifier":"arXiv:2011.12255","url":"https://arxiv.org/abs/2011.12255"},{"text":"Y. Kim, C. Kim, and J. Hwangbo. Learning Forward Dynamics Model and Informed Trajectory Sampler for Safe Quadruped Navigation. arXiv:2204.08647, 2022.","identifier":"arXiv:2204.08647","url":"https://arxiv.org/abs/2204.08647"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. arXiv:2305.08510, 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"R. Grandia, F. Jenelten, S. Yang, F. Farshidian, and M. Hutter. Perceptive Locomotion through Nonlinear Model Predictive Control. arXiv:2208.08373, 2022.","identifier":"arXiv:2208.08373","url":"https://arxiv.org/abs/2208.08373"},{"text":"M. Pfeiffer, M. Schaeuble, J. Nieto, R. Siegwart, and C. Cadena. From Perception to Decision: A Data-driven Approach to End-to-end Motion Planning for Autonomous Ground Robots. arXiv:1609.07910, 2016.","identifier":"arXiv:1609.07910","url":"https://arxiv.org/abs/1609.07910"},{"text":"L. Tai, G. Paolo, and M. Liu. Virtual-to-real Deep Reinforcement Learning: Continuous Control of Mobile Robots for Mapless Navigation. arXiv:1703.00420, 2017.","identifier":"arXiv:1703.00420","url":"https://arxiv.org/abs/1703.00420"},{"text":"H.-T. L. Chiang, A. Faust, M. Fiser, and A. Francis. Learning Navigation Behaviors End-to-End with AutoRL. arXiv:1809.10124, 2018.","identifier":"arXiv:1809.10124","url":"https://arxiv.org/abs/1809.10124"},{"text":"G. Kahn, A. Villaflor, B. Ding, P. Abbeel, and S. Levine. Self-supervised Deep Reinforcement Learning with Generalized Computation Graphs for Robot Navigation. arXiv:1709.10489, 2017.","identifier":"arXiv:1709.10489","url":"https://arxiv.org/abs/1709.10489"},{"text":"G. Kahn, P. Abbeel, and S. Levine. BADGR: An Autonomous Self-Supervised Learning-Based Navigation System. arXiv:2002.05700, 2020.","identifier":"arXiv:2002.05700","url":"https://arxiv.org/abs/2002.05700"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Sridhar, D. Shah, C. Glossop, and S. Levine. NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration. arXiv:2310.07896, 2023.","identifier":"arXiv:2310.07896","url":"https://arxiv.org/abs/2310.07896"},{"text":"A. Loquercio, E. Kaufmann, R. Ranftl, M. Müller, V. Koltun, and D. Scaramuzza. Learning High-Speed Flight in the Wild. arXiv:2110.05113, 2021.","identifier":"arXiv:2110.05113","url":"https://arxiv.org/abs/2110.05113"},{"text":"N. Mohajerin and M. Rohani. Multi-Step Prediction of Occupancy Grid Maps with Recurrent Neural Networks. arXiv:1812.09395, 2018.","identifier":"arXiv:1812.09395","url":"https://arxiv.org/abs/1812.09395"},{"text":"M. Itkina, K. Driggs-Campbell, and M. J. Kochenderfer. Dynamic Environment Prediction in Urban Scenes using Recurrent Representation Learning. arXiv:1904.12374, 2019.","identifier":"arXiv:1904.12374","url":"https://arxiv.org/abs/1904.12374"},{"text":"X. Weng, J. Wang, S. Levine, K. Kitani, and N. Rhinehart. Inverting the Pose Forecasting Pipeline with SPF2: Sequential Pointcloud Forecasting for Sequential Pose Forecasting. arXiv:2003.08376, 2020.","identifier":"arXiv:2003.08376","url":"https://arxiv.org/abs/2003.08376"},{"text":"B. Mersch, X. Chen, J. Behley, and C. Stachniss. Self-supervised Point Cloud Prediction Using 3D Spatio-temporal Convolutional Networks. arXiv:2110.04076, 2021.","identifier":"arXiv:2110.04076","url":"https://arxiv.org/abs/2110.04076"},{"text":"M. Bansal, A. Krizhevsky, and A. Ogale. ChauffeurNet: Learning to Drive by Imitating the Best and Synthesizing the Worst. arXiv:1812.03079, 2018.","identifier":"arXiv:1812.03079","url":"https://arxiv.org/abs/1812.03079"},{"text":"A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin. Attention Is All You Need. arXiv:1706.03762, 2017.","identifier":"arXiv:1706.03762","url":"https://arxiv.org/abs/1706.03762"},{"text":"J. Lee, Y. Lee, J. Kim, A. R. Kosiorek, S. Choi, and Y. W. Teh. Set Transformer: A Framework for Attention-based Permutation-Invariant Neural Networks. arXiv:1810.00825, 2018.","identifier":"arXiv:1810.00825","url":"https://arxiv.org/abs/1810.00825"},{"text":"A. Jaegle, F. Gimeno, A. Brock, A. Zisserman, O. Vinyals, and J. Carreira. Perceiver: General Perception with Iterative Attention. arXiv:2103.03206, 2021.","identifier":"arXiv:2103.03206","url":"https://arxiv.org/abs/2103.03206"},{"text":"A. Jaegle et al. Perceiver IO: A General Architecture for Structured Inputs and Outputs. arXiv:2107.14795, 2021.","identifier":"arXiv:2107.14795","url":"https://arxiv.org/abs/2107.14795"},{"text":"N. Carion, F. Massa, G. Synnaeve, N. Usunier, A. Kirillov, and S. Zagoruyko. End-to-End Object Detection with Transformers. arXiv:2005.12872, 2020.","identifier":"arXiv:2005.12872","url":"https://arxiv.org/abs/2005.12872"},{"text":"L. Mescheder, M. Oechsle, M. Niemeyer, S. Nowozin, and A. Geiger. Occupancy Networks: Learning 3D Reconstruction in Function Space. arXiv:1812.03828, 2018.","identifier":"arXiv:1812.03828","url":"https://arxiv.org/abs/1812.03828"},{"text":"J. J. Park, P. Florence, J. Straub, R. Newcombe, and S. Lovegrove. DeepSDF: Learning Continuous Signed Distance Functions for Shape Representation. arXiv:1901.05103, 2019.","identifier":"arXiv:1901.05103","url":"https://arxiv.org/abs/1901.05103"},{"text":"B. Mildenhall, P. P. Srinivasan, M. Tancik, J. T. Barron, R. Ramamoorthi, and R. Ng. NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis. arXiv:2003.08934, 2020.","identifier":"arXiv:2003.08934","url":"https://arxiv.org/abs/2003.08934"},{"text":"M. Tancik, P. P. Srinivasan, B. Mildenhall, S. Fridovich-Keil, N. Raghavan, U. Singhal, R. Ramamoorthi, J. T. Barron, and R. Ng. Fourier Features Let Networks Learn High Frequency Functions in Low Dimensional Domains. arXiv:2006.10739, 2020.","identifier":"arXiv:2006.10739","url":"https://arxiv.org/abs/2006.10739"},{"text":"D. Hafner, J. Pasukonis, J. Ba, and T. Lillicrap. Mastering Diverse Domains through World Models. arXiv:2301.04104, 2023.","identifier":"arXiv:2301.04104","url":"https://arxiv.org/abs/2301.04104"},{"text":"A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada. Control Barrier Functions: Theory and Applications. arXiv:1903.11199, 2019.","identifier":"arXiv:1903.11199","url":"https://arxiv.org/abs/1903.11199"},{"text":"A. Kendall and Y. Gal. What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision? arXiv:1703.04977, 2017.","identifier":"arXiv:1703.04977","url":"https://arxiv.org/abs/1703.04977"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2608.00002","metadata":{"title":"Latent Motion Tokens for Future-Range Prediction in Quadruped Obstacle Avoidance","authors":[{"name":"Marcus Hill","affiliation":"Robotics Institute, Georgia Institute of Technology, Atlanta, GA, USA"},{"name":"Ivan Sokolov","affiliation":"Robotics Institute, Georgia Institute of Technology, Atlanta, GA, USA"},{"name":"Yating Feng","affiliation":"Robotics Institute, Georgia Institute of Technology, Atlanta, GA, USA"},{"name":"Rachel Kim","affiliation":"Robotics Institute, Georgia Institute of Technology, Atlanta, GA, USA"}],"publication_date":"2026-08-04","abstract":"Learned range predictors for legged-robot navigation usually regress the free distance along a fixed set of rays. This output format is difficult to condition on the motion the robot intends to execute, and it forces the planner and the reactive safety layer to share a single ray resolution. This paper encodes the predicted future range field of a quadruped as a small set of latent motion tokens, one per cluster of control-aligned directions. A transformer generates the tokens from recent range history and the planned motion, and a lightweight query decoder returns the future free range for any ray, horizon, and candidate motion. Because queries are resolved against the tokens rather than a fixed output layer, the same predictor serves a 2 Hz planner issuing coarse queries and a 50 Hz reactive layer issuing dense queries. In Isaac Lab experiments with ANYmal in 20 cluttered obstacle-avoidance scenes, with 5 seeds and 30 trials per scene and seed, the token predictor attains a range prediction error of 0.87 m, a 95th-percentile query latency of 13.6 ms on the onboard Orin computer, and an avoidance success rate of 91.8 percent, compared with 0.91 m, 14.4 ms, and 79.8 percent for a FutureRay-style direct-regression baseline, while using 1.2 million instead of 1.3 million parameters. Ablations indicate that motion-plan conditioning and the number of tokens matter most, and the recorded failure cases identify occluded turns and planner stalls as remaining weaknesses.","references":[{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"T. Zang, S. Wang, H. Huang, and S. Cheng. FutureRay: Control-Aligned Future Range for Agile Quadruped Navigation. arXiv:2608.01197, 2026.","identifier":"arXiv:2608.01197","url":"https://arxiv.org/abs/2608.01197"},{"text":"V. Makoviychuk et al. Isaac Gym: High Performance GPU-Based Physics Simulation for Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal et al. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. arXiv:2301.04195, 2023.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"R. Yang, M. Zhang, N. Hansen, H. Xu, and X. Wang. Learning Vision-Guided Quadrupedal Locomotion End-to-End with Cross-Modal Transformers. arXiv:2107.03996, 2021.","identifier":"arXiv:2107.03996","url":"https://arxiv.org/abs/2107.03996"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. arXiv:2211.07638, 2022.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. arXiv:2309.14341, 2023.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. arXiv:2309.05665, 2023.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, L. Wellhausen, F. Farshidian, and M. Hutter. Learning a State Representation and Navigation in Cluttered and Dynamic Environments. arXiv:2103.04351, 2021.","identifier":"arXiv:2103.04351","url":"https://arxiv.org/abs/2103.04351"},{"text":"J. Truong, D. Yarats, T. Li, F. Meier, S. Chernova, D. Batra, and A. Rai. Learning Navigation Skills for Legged Robots with Learned Robot Embeddings. arXiv:2011.12255, 2020.","identifier":"arXiv:2011.12255","url":"https://arxiv.org/abs/2011.12255"},{"text":"Y. Kim, C. Kim, and J. Hwangbo. Learning Forward Dynamics Model and Informed Trajectory Sampler for Safe Quadruped Navigation. arXiv:2204.08647, 2022.","identifier":"arXiv:2204.08647","url":"https://arxiv.org/abs/2204.08647"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. arXiv:2305.08510, 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"R. Grandia, F. Jenelten, S. Yang, F. Farshidian, and M. Hutter. Perceptive Locomotion through Nonlinear Model Predictive Control. arXiv:2208.08373, 2022.","identifier":"arXiv:2208.08373","url":"https://arxiv.org/abs/2208.08373"},{"text":"M. Pfeiffer, M. Schaeuble, J. Nieto, R. Siegwart, and C. Cadena. From Perception to Decision: A Data-driven Approach to End-to-end Motion Planning for Autonomous Ground Robots. arXiv:1609.07910, 2016.","identifier":"arXiv:1609.07910","url":"https://arxiv.org/abs/1609.07910"},{"text":"L. Tai, G. Paolo, and M. Liu. Virtual-to-real Deep Reinforcement Learning: Continuous Control of Mobile Robots for Mapless Navigation. arXiv:1703.00420, 2017.","identifier":"arXiv:1703.00420","url":"https://arxiv.org/abs/1703.00420"},{"text":"H.-T. L. Chiang, A. Faust, M. Fiser, and A. Francis. Learning Navigation Behaviors End-to-End with AutoRL. arXiv:1809.10124, 2018.","identifier":"arXiv:1809.10124","url":"https://arxiv.org/abs/1809.10124"},{"text":"G. Kahn, A. Villaflor, B. Ding, P. Abbeel, and S. Levine. Self-supervised Deep Reinforcement Learning with Generalized Computation Graphs for Robot Navigation. arXiv:1709.10489, 2017.","identifier":"arXiv:1709.10489","url":"https://arxiv.org/abs/1709.10489"},{"text":"G. Kahn, P. Abbeel, and S. Levine. BADGR: An Autonomous Self-Supervised Learning-Based Navigation System. arXiv:2002.05700, 2020.","identifier":"arXiv:2002.05700","url":"https://arxiv.org/abs/2002.05700"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Sridhar, D. Shah, C. Glossop, and S. Levine. NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration. arXiv:2310.07896, 2023.","identifier":"arXiv:2310.07896","url":"https://arxiv.org/abs/2310.07896"},{"text":"A. Loquercio, E. Kaufmann, R. Ranftl, M. Müller, V. Koltun, and D. Scaramuzza. Learning High-Speed Flight in the Wild. arXiv:2110.05113, 2021.","identifier":"arXiv:2110.05113","url":"https://arxiv.org/abs/2110.05113"},{"text":"N. Mohajerin and M. Rohani. Multi-Step Prediction of Occupancy Grid Maps with Recurrent Neural Networks. arXiv:1812.09395, 2018.","identifier":"arXiv:1812.09395","url":"https://arxiv.org/abs/1812.09395"},{"text":"M. Itkina, K. Driggs-Campbell, and M. J. Kochenderfer. Dynamic Environment Prediction in Urban Scenes using Recurrent Representation Learning. arXiv:1904.12374, 2019.","identifier":"arXiv:1904.12374","url":"https://arxiv.org/abs/1904.12374"},{"text":"X. Weng, J. Wang, S. Levine, K. Kitani, and N. Rhinehart. Inverting the Pose Forecasting Pipeline with SPF2: Sequential Pointcloud Forecasting for Sequential Pose Forecasting. arXiv:2003.08376, 2020.","identifier":"arXiv:2003.08376","url":"https://arxiv.org/abs/2003.08376"},{"text":"B. Mersch, X. Chen, J. Behley, and C. Stachniss. Self-supervised Point Cloud Prediction Using 3D Spatio-temporal Convolutional Networks. arXiv:2110.04076, 2021.","identifier":"arXiv:2110.04076","url":"https://arxiv.org/abs/2110.04076"},{"text":"M. Bansal, A. Krizhevsky, and A. Ogale. ChauffeurNet: Learning to Drive by Imitating the Best and Synthesizing the Worst. arXiv:1812.03079, 2018.","identifier":"arXiv:1812.03079","url":"https://arxiv.org/abs/1812.03079"},{"text":"A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin. Attention Is All You Need. arXiv:1706.03762, 2017.","identifier":"arXiv:1706.03762","url":"https://arxiv.org/abs/1706.03762"},{"text":"J. Lee, Y. Lee, J. Kim, A. R. Kosiorek, S. Choi, and Y. W. Teh. Set Transformer: A Framework for Attention-based Permutation-Invariant Neural Networks. arXiv:1810.00825, 2018.","identifier":"arXiv:1810.00825","url":"https://arxiv.org/abs/1810.00825"},{"text":"A. Jaegle, F. Gimeno, A. Brock, A. Zisserman, O. Vinyals, and J. Carreira. Perceiver: General Perception with Iterative Attention. arXiv:2103.03206, 2021.","identifier":"arXiv:2103.03206","url":"https://arxiv.org/abs/2103.03206"},{"text":"A. Jaegle et al. Perceiver IO: A General Architecture for Structured Inputs and Outputs. arXiv:2107.14795, 2021.","identifier":"arXiv:2107.14795","url":"https://arxiv.org/abs/2107.14795"},{"text":"N. Carion, F. Massa, G. Synnaeve, N. Usunier, A. Kirillov, and S. Zagoruyko. End-to-End Object Detection with Transformers. arXiv:2005.12872, 2020.","identifier":"arXiv:2005.12872","url":"https://arxiv.org/abs/2005.12872"},{"text":"L. Mescheder, M. Oechsle, M. Niemeyer, S. Nowozin, and A. Geiger. Occupancy Networks: Learning 3D Reconstruction in Function Space. arXiv:1812.03828, 2018.","identifier":"arXiv:1812.03828","url":"https://arxiv.org/abs/1812.03828"},{"text":"J. J. Park, P. Florence, J. Straub, R. Newcombe, and S. Lovegrove. DeepSDF: Learning Continuous Signed Distance Functions for Shape Representation. arXiv:1901.05103, 2019.","identifier":"arXiv:1901.05103","url":"https://arxiv.org/abs/1901.05103"},{"text":"B. Mildenhall, P. P. Srinivasan, M. Tancik, J. T. Barron, R. Ramamoorthi, and R. Ng. NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis. arXiv:2003.08934, 2020.","identifier":"arXiv:2003.08934","url":"https://arxiv.org/abs/2003.08934"},{"text":"M. Tancik, P. P. Srinivasan, B. Mildenhall, S. Fridovich-Keil, N. Raghavan, U. Singhal, R. Ramamoorthi, J. T. Barron, and R. Ng. Fourier Features Let Networks Learn High Frequency Functions in Low Dimensional Domains. arXiv:2006.10739, 2020.","identifier":"arXiv:2006.10739","url":"https://arxiv.org/abs/2006.10739"},{"text":"D. Hafner, J. Pasukonis, J. Ba, and T. Lillicrap. Mastering Diverse Domains through World Models. arXiv:2301.04104, 2023.","identifier":"arXiv:2301.04104","url":"https://arxiv.org/abs/2301.04104"},{"text":"A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada. Control Barrier Functions: Theory and Applications. arXiv:1903.11199, 2019.","identifier":"arXiv:1903.11199","url":"https://arxiv.org/abs/1903.11199"},{"text":"A. Kendall and Y. Gal. What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision? arXiv:1703.04977, 2017.","identifier":"arXiv:1703.04977","url":"https://arxiv.org/abs/1703.04977"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2608.00002","slug":"2608.00002","registered_at":"2026-08-06T17:00:00Z","updated_at":"2026-08-07T18:38:00Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2608.00002","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2608.00002","tex_path":"articles/2608.00002/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-08-06T19:17:00Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2608.00002"},{"kagg_id":"KAGG:2608.00001","slug":"2608.00001","title":"Information-Directed Active Calibration of Unknown Velocity Dynamics","authors":[{"name":"Daniel Okafor","affiliation":"Advanced Robotics Group, Toyota Research Institute, Palo Alto, CA, USA"},{"name":"Priya Raghunathan","affiliation":"Advanced Robotics Group, Toyota Research Institute, Palo Alto, CA, USA"},{"name":"Hana Sato","affiliation":"Advanced Robotics Group, Toyota Research Institute, Palo Alto, CA, USA"}],"affiliation":null,"abstract":"Navigation stacks for legged robots usually command a vendor locomotion controller through an opaque velocity interface whose gains, saturation limits, and latency differ from their nominal values and must be calibrated before a planner can trust its motion model. Calibration requires choosing probe commands, and uniform excitation spends much of its motion on directions whose response is already well determined. This paper formulates probe selection as information-directed sampling. A particle posterior is maintained over a six-parameter state-delay model of the interface, comprising three per-axis gains, a planar-speed and a yaw-rate saturation limit, and a transport delay, and each probe is chosen to maximize the expected information gain about these parameters per unit of commanded path length. The information gain is evaluated through the saturated model, and a hierarchical shrinkage prior couples the per-axis gains so that weakly excited directions borrow strength from well-excited ones. The method is compared with uniform sweep excitation, random probing under a motion budget, and the planner-conditioned probes of GAUGE over 200 simulated calibration sessions of a Unitree B2 in Isaac Lab, with validation on five real-robot sessions. Information-directed probing attains a worst-direction calibration error of 8.9 percent against 9.5 to 10.3 percent for the baselines, reaches navigation-grade accuracy in 38 s against 40 to 42 s, and uses 11.4 m of commanded motion against 11.8 to 13.0 m. Ablations indicate that cost weighting, saturation awareness, and shrinkage are each needed to outperform uniform excitation.","publication_date":"2026-08-02","registered_at":"2026-08-08T08:45:00Z","updated_at":"2026-08-10T09:26:00Z","status":"published","html_path":null,"pdf_path":"articles/2608.00001/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"D. Fox, W. Burgard, and S. Thrun. The Dynamic Window Approach to Collision Avoidance. IEEE Robotics and Automation Magazine, 4(1):23–33, 1997.","identifier":"doi:10.1109/100.580977","url":"https://doi.org/10.1109/100.580977"},{"text":"G. Williams, A. Aldrich, and E. Theodorou. Model Predictive Path Integral Control using Covariance Variable Importance Sampling. arXiv:1509.01149, 2015.","identifier":"arXiv:1509.01149","url":"https://arxiv.org/abs/1509.01149"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"J. Di Carlo, P. M. Wensing, B. Katz, G. Bledt, and S. Kim. Dynamic Locomotion in the MIT Cheetah 3 Through Convex Model-Predictive Control. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2018.","identifier":"doi:10.1109/IROS.2018.8594448","url":"https://doi.org/10.1109/IROS.2018.8594448"},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"T. Zang, H. Huang, S. Cheng, and W. Zhang. GAUGE: Planner-Conditioned Active Calibration of Opaque Quadruped Velocity Interfaces. arXiv:2607.02841, 2026.","identifier":"arXiv:2607.02841","url":"https://arxiv.org/abs/2607.02841"},{"text":"K. Chaloner and I. Verdinelli. Bayesian Experimental Design: A Review. Statistical Science, 10(3):273–304, 1995.","identifier":"doi:10.1214/ss/1177009939","url":"https://doi.org/10.1214/ss/1177009939"},{"text":"T. Rainforth, A. Foster, D. R. Ivanova, and F. Bickford Smith. Modern Bayesian Experimental Design. arXiv:2302.14545, 2023.","identifier":"arXiv:2302.14545","url":"https://arxiv.org/abs/2302.14545"},{"text":"D. Russo and B. Van Roy. Learning to Optimize via Information-Directed Sampling. arXiv:1403.5556, 2014.","identifier":"arXiv:1403.5556","url":"https://arxiv.org/abs/1403.5556"},{"text":"J. Borenstein and L. Feng. Measurement and Correction of Systematic Odometry Errors in Mobile Robots. IEEE Transactions on Robotics and Automation, 12(6):869–880, 1996.","identifier":"doi:10.1109/70.544770","url":"https://doi.org/10.1109/70.544770"},{"text":"A. Censi, A. Franchi, L. Marchionni, and G. Oriolo. Simultaneous Calibration of Odometry and Sensor Parameters for Mobile Robots. IEEE Transactions on Robotics, 29(2):475–492, 2013.","identifier":"doi:10.1109/TRO.2012.2226380","url":"https://doi.org/10.1109/TRO.2012.2226380"},{"text":"A. Foster, M. Jankowiak, E. Bingham, P. Horsfall, Y. W. Teh, T. Rainforth, and N. Goodman. Variational Bayesian Optimal Experimental Design. arXiv:1903.05480, 2019.","identifier":"arXiv:1903.05480","url":"https://arxiv.org/abs/1903.05480"},{"text":"A. Foster, D. R. Ivanova, I. Malik, and T. Rainforth. Deep Adaptive Design: Amortizing Sequential Bayesian Experimental Design. arXiv:2103.02438, 2021.","identifier":"arXiv:2103.02438","url":"https://arxiv.org/abs/2103.02438"},{"text":"N. Houlsby, F. Huszár, Z. Ghahramani, and M. Lengyel. Bayesian Active Learning for Classification and Preference Learning. arXiv:1112.5745, 2011.","identifier":"arXiv:1112.5745","url":"https://arxiv.org/abs/1112.5745"},{"text":"J. Kirschner and A. Krause. Information Directed Sampling and Bandits with Heteroscedastic Noise. arXiv:1801.09667, 2018.","identifier":"arXiv:1801.09667","url":"https://arxiv.org/abs/1801.09667"},{"text":"J. Kirschner, T. Lattimore, and A. Krause. Information Directed Sampling for Linear Partial Monitoring. arXiv:2002.11182, 2020.","identifier":"arXiv:2002.11182","url":"https://arxiv.org/abs/2002.11182"},{"text":"N. Srinivas, A. Krause, S. M. Kakade, and M. Seeger. Gaussian Process Optimization in the Bandit Setting: No Regret and Experimental Design. arXiv:0912.3995, 2009.","identifier":"arXiv:0912.3995","url":"https://arxiv.org/abs/0912.3995"},{"text":"R. K. Mehra. Optimal Input Signals for Parameter Estimation in Dynamic Systems: Survey and New Results. IEEE Transactions on Automatic Control, 19(6):753–768, 1974.","identifier":"doi:10.1109/TAC.1974.1100701","url":"https://doi.org/10.1109/TAC.1974.1100701"},{"text":"A. Wagenmaker and K. Jamieson. Active Learning for Identification of Linear Dynamical Systems. arXiv:2002.00495, 2020.","identifier":"arXiv:2002.00495","url":"https://arxiv.org/abs/2002.00495"},{"text":"H. Mania, M. I. Jordan, and B. Recht. Active Learning for Nonlinear System Identification with Guarantees. arXiv:2006.10277, 2020.","identifier":"arXiv:2006.10277","url":"https://arxiv.org/abs/2006.10277"},{"text":"M. Buisson-Fenet, F. Solowjow, and S. Trimpe. Actively Learning Gaussian Process Dynamics. arXiv:1911.09946, 2019.","identifier":"arXiv:1911.09946","url":"https://arxiv.org/abs/1911.09946"},{"text":"P. Shyam, W. Jaśkowski, and F. Gomez. Model-Based Active Exploration. arXiv:1810.12162, 2018.","identifier":"arXiv:1810.12162","url":"https://arxiv.org/abs/1810.12162"},{"text":"D. Pathak, D. Gandhi, and A. Gupta. Self-Supervised Exploration via Disagreement. arXiv:1906.04161, 2019.","identifier":"arXiv:1906.04161","url":"https://arxiv.org/abs/1906.04161"},{"text":"B. Sukhija, L. Treven, C. Sancaktar, S. Blaes, S. Coros, and A. Krause. Optimistic Active Exploration of Dynamical Systems. arXiv:2306.12371, 2023.","identifier":"arXiv:2306.12371","url":"https://arxiv.org/abs/2306.12371"},{"text":"M. Memmel, A. Wagenmaker, C. Zhu, P. Yin, D. Fox, and A. Gupta. ASID: Active Exploration for System Identification in Robotic Manipulation. arXiv:2404.12308, 2024.","identifier":"arXiv:2404.12308","url":"https://arxiv.org/abs/2404.12308"},{"text":"T. Lew, A. Sharma, J. Harrison, A. Bylard, and M. Pavone. Safe Active Dynamics Learning and Control: A Sequential Exploration-Exploitation Framework. arXiv:2008.11700, 2020.","identifier":"arXiv:2008.11700","url":"https://arxiv.org/abs/2008.11700"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"V. Makoviychuk et al.. Isaac Gym: High Performance GPU-Based Physics Simulation for Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal et al.. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. arXiv:2301.04195, 2023.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. arXiv:1703.06907, 2017.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. arXiv:1710.06537, 2017.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"},{"text":"W. Yu, J. Tan, C. K. Liu, and G. Turk. Preparing for the Unknown: Learning a Universal Policy with Online System Identification. arXiv:1702.02453, 2017.","identifier":"arXiv:1702.02453","url":"https://arxiv.org/abs/1702.02453"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"F. Ramos, R. C. Possas, and D. Fox. BayesSim: Adaptive Domain Randomization via Probabilistic Inference for Robotics Simulators. arXiv:1906.01728, 2019.","identifier":"arXiv:1906.01728","url":"https://arxiv.org/abs/1906.01728"},{"text":"Y. Chebotar, A. Handa, V. Makoviychuk, M. Macklin, J. Issac, N. Ratliff, and D. Fox. Closing the Sim-to-Real Loop: Adapting Simulation Randomization with Real World Experience. arXiv:1810.05687, 2018.","identifier":"arXiv:1810.05687","url":"https://arxiv.org/abs/1810.05687"},{"text":"G. Kahn, P. Abbeel, and S. Levine. BADGR: An Autonomous Self-Supervised Learning-Based Navigation System. arXiv:2002.05700, 2020.","identifier":"arXiv:2002.05700","url":"https://arxiv.org/abs/2002.05700"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. arXiv:2305.08510, 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada. Control Barrier Functions: Theory and Applications. arXiv:1903.11199, 2019.","identifier":"arXiv:1903.11199","url":"https://arxiv.org/abs/1903.11199"},{"text":"M. S. Arulampalam, S. Maskell, N. Gordon, and T. Clapp. A Tutorial on Particle Filters for Online Nonlinear/Non-Gaussian Bayesian Tracking. IEEE Transactions on Signal Processing, 50(2):174–188, 2002.","identifier":"doi:10.1109/78.978374","url":"https://doi.org/10.1109/78.978374"},{"text":"E. A. Wan and R. van der Merwe. The Unscented Kalman Filter for Nonlinear Estimation. Proceedings of the IEEE Adaptive Systems for Signal Processing, Communications, and Control Symposium, 153–158, 2000.","identifier":"doi:10.1109/ASSPCC.2000.882463","url":"https://doi.org/10.1109/ASSPCC.2000.882463"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2608.00001","metadata":{"title":"Information-Directed Active Calibration of Unknown Velocity Dynamics","authors":[{"name":"Daniel Okafor","affiliation":"Advanced Robotics Group, Toyota Research Institute, Palo Alto, CA, USA"},{"name":"Priya Raghunathan","affiliation":"Advanced Robotics Group, Toyota Research Institute, Palo Alto, CA, USA"},{"name":"Hana Sato","affiliation":"Advanced Robotics Group, Toyota Research Institute, Palo Alto, CA, USA"}],"publication_date":"2026-08-02","abstract":"Navigation stacks for legged robots usually command a vendor locomotion controller through an opaque velocity interface whose gains, saturation limits, and latency differ from their nominal values and must be calibrated before a planner can trust its motion model. Calibration requires choosing probe commands, and uniform excitation spends much of its motion on directions whose response is already well determined. This paper formulates probe selection as information-directed sampling. A particle posterior is maintained over a six-parameter state-delay model of the interface, comprising three per-axis gains, a planar-speed and a yaw-rate saturation limit, and a transport delay, and each probe is chosen to maximize the expected information gain about these parameters per unit of commanded path length. The information gain is evaluated through the saturated model, and a hierarchical shrinkage prior couples the per-axis gains so that weakly excited directions borrow strength from well-excited ones. The method is compared with uniform sweep excitation, random probing under a motion budget, and the planner-conditioned probes of GAUGE over 200 simulated calibration sessions of a Unitree B2 in Isaac Lab, with validation on five real-robot sessions. Information-directed probing attains a worst-direction calibration error of 8.9 percent against 9.5 to 10.3 percent for the baselines, reaches navigation-grade accuracy in 38 s against 40 to 42 s, and uses 11.4 m of commanded motion against 11.8 to 13.0 m. Ablations indicate that cost weighting, saturation awareness, and shrinkage are each needed to outperform uniform excitation.","references":[{"text":"D. Fox, W. Burgard, and S. Thrun. The Dynamic Window Approach to Collision Avoidance. IEEE Robotics and Automation Magazine, 4(1):23–33, 1997.","identifier":"doi:10.1109/100.580977","url":"https://doi.org/10.1109/100.580977"},{"text":"G. Williams, A. Aldrich, and E. Theodorou. Model Predictive Path Integral Control using Covariance Variable Importance Sampling. arXiv:1509.01149, 2015.","identifier":"arXiv:1509.01149","url":"https://arxiv.org/abs/1509.01149"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"J. Di Carlo, P. M. Wensing, B. Katz, G. Bledt, and S. Kim. Dynamic Locomotion in the MIT Cheetah 3 Through Convex Model-Predictive Control. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2018.","identifier":"doi:10.1109/IROS.2018.8594448","url":"https://doi.org/10.1109/IROS.2018.8594448"},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"T. Zang, H. Huang, S. Cheng, and W. Zhang. GAUGE: Planner-Conditioned Active Calibration of Opaque Quadruped Velocity Interfaces. arXiv:2607.02841, 2026.","identifier":"arXiv:2607.02841","url":"https://arxiv.org/abs/2607.02841"},{"text":"K. Chaloner and I. Verdinelli. Bayesian Experimental Design: A Review. Statistical Science, 10(3):273–304, 1995.","identifier":"doi:10.1214/ss/1177009939","url":"https://doi.org/10.1214/ss/1177009939"},{"text":"T. Rainforth, A. Foster, D. R. Ivanova, and F. Bickford Smith. Modern Bayesian Experimental Design. arXiv:2302.14545, 2023.","identifier":"arXiv:2302.14545","url":"https://arxiv.org/abs/2302.14545"},{"text":"D. Russo and B. Van Roy. Learning to Optimize via Information-Directed Sampling. arXiv:1403.5556, 2014.","identifier":"arXiv:1403.5556","url":"https://arxiv.org/abs/1403.5556"},{"text":"J. Borenstein and L. Feng. Measurement and Correction of Systematic Odometry Errors in Mobile Robots. IEEE Transactions on Robotics and Automation, 12(6):869–880, 1996.","identifier":"doi:10.1109/70.544770","url":"https://doi.org/10.1109/70.544770"},{"text":"A. Censi, A. Franchi, L. Marchionni, and G. Oriolo. Simultaneous Calibration of Odometry and Sensor Parameters for Mobile Robots. IEEE Transactions on Robotics, 29(2):475–492, 2013.","identifier":"doi:10.1109/TRO.2012.2226380","url":"https://doi.org/10.1109/TRO.2012.2226380"},{"text":"A. Foster, M. Jankowiak, E. Bingham, P. Horsfall, Y. W. Teh, T. Rainforth, and N. Goodman. Variational Bayesian Optimal Experimental Design. arXiv:1903.05480, 2019.","identifier":"arXiv:1903.05480","url":"https://arxiv.org/abs/1903.05480"},{"text":"A. Foster, D. R. Ivanova, I. Malik, and T. Rainforth. Deep Adaptive Design: Amortizing Sequential Bayesian Experimental Design. arXiv:2103.02438, 2021.","identifier":"arXiv:2103.02438","url":"https://arxiv.org/abs/2103.02438"},{"text":"N. Houlsby, F. Huszár, Z. Ghahramani, and M. Lengyel. Bayesian Active Learning for Classification and Preference Learning. arXiv:1112.5745, 2011.","identifier":"arXiv:1112.5745","url":"https://arxiv.org/abs/1112.5745"},{"text":"J. Kirschner and A. Krause. Information Directed Sampling and Bandits with Heteroscedastic Noise. arXiv:1801.09667, 2018.","identifier":"arXiv:1801.09667","url":"https://arxiv.org/abs/1801.09667"},{"text":"J. Kirschner, T. Lattimore, and A. Krause. Information Directed Sampling for Linear Partial Monitoring. arXiv:2002.11182, 2020.","identifier":"arXiv:2002.11182","url":"https://arxiv.org/abs/2002.11182"},{"text":"N. Srinivas, A. Krause, S. M. Kakade, and M. Seeger. Gaussian Process Optimization in the Bandit Setting: No Regret and Experimental Design. arXiv:0912.3995, 2009.","identifier":"arXiv:0912.3995","url":"https://arxiv.org/abs/0912.3995"},{"text":"R. K. Mehra. Optimal Input Signals for Parameter Estimation in Dynamic Systems: Survey and New Results. IEEE Transactions on Automatic Control, 19(6):753–768, 1974.","identifier":"doi:10.1109/TAC.1974.1100701","url":"https://doi.org/10.1109/TAC.1974.1100701"},{"text":"A. Wagenmaker and K. Jamieson. Active Learning for Identification of Linear Dynamical Systems. arXiv:2002.00495, 2020.","identifier":"arXiv:2002.00495","url":"https://arxiv.org/abs/2002.00495"},{"text":"H. Mania, M. I. Jordan, and B. Recht. Active Learning for Nonlinear System Identification with Guarantees. arXiv:2006.10277, 2020.","identifier":"arXiv:2006.10277","url":"https://arxiv.org/abs/2006.10277"},{"text":"M. Buisson-Fenet, F. Solowjow, and S. Trimpe. Actively Learning Gaussian Process Dynamics. arXiv:1911.09946, 2019.","identifier":"arXiv:1911.09946","url":"https://arxiv.org/abs/1911.09946"},{"text":"P. Shyam, W. Jaśkowski, and F. Gomez. Model-Based Active Exploration. arXiv:1810.12162, 2018.","identifier":"arXiv:1810.12162","url":"https://arxiv.org/abs/1810.12162"},{"text":"D. Pathak, D. Gandhi, and A. Gupta. Self-Supervised Exploration via Disagreement. arXiv:1906.04161, 2019.","identifier":"arXiv:1906.04161","url":"https://arxiv.org/abs/1906.04161"},{"text":"B. Sukhija, L. Treven, C. Sancaktar, S. Blaes, S. Coros, and A. Krause. Optimistic Active Exploration of Dynamical Systems. arXiv:2306.12371, 2023.","identifier":"arXiv:2306.12371","url":"https://arxiv.org/abs/2306.12371"},{"text":"M. Memmel, A. Wagenmaker, C. Zhu, P. Yin, D. Fox, and A. Gupta. ASID: Active Exploration for System Identification in Robotic Manipulation. arXiv:2404.12308, 2024.","identifier":"arXiv:2404.12308","url":"https://arxiv.org/abs/2404.12308"},{"text":"T. Lew, A. Sharma, J. Harrison, A. Bylard, and M. Pavone. Safe Active Dynamics Learning and Control: A Sequential Exploration-Exploitation Framework. arXiv:2008.11700, 2020.","identifier":"arXiv:2008.11700","url":"https://arxiv.org/abs/2008.11700"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"V. Makoviychuk et al.. Isaac Gym: High Performance GPU-Based Physics Simulation for Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal et al.. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. arXiv:2301.04195, 2023.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. arXiv:1703.06907, 2017.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. arXiv:1710.06537, 2017.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"},{"text":"W. Yu, J. Tan, C. K. Liu, and G. Turk. Preparing for the Unknown: Learning a Universal Policy with Online System Identification. arXiv:1702.02453, 2017.","identifier":"arXiv:1702.02453","url":"https://arxiv.org/abs/1702.02453"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"F. Ramos, R. C. Possas, and D. Fox. BayesSim: Adaptive Domain Randomization via Probabilistic Inference for Robotics Simulators. arXiv:1906.01728, 2019.","identifier":"arXiv:1906.01728","url":"https://arxiv.org/abs/1906.01728"},{"text":"Y. Chebotar, A. Handa, V. Makoviychuk, M. Macklin, J. Issac, N. Ratliff, and D. Fox. Closing the Sim-to-Real Loop: Adapting Simulation Randomization with Real World Experience. arXiv:1810.05687, 2018.","identifier":"arXiv:1810.05687","url":"https://arxiv.org/abs/1810.05687"},{"text":"G. Kahn, P. Abbeel, and S. Levine. BADGR: An Autonomous Self-Supervised Learning-Based Navigation System. arXiv:2002.05700, 2020.","identifier":"arXiv:2002.05700","url":"https://arxiv.org/abs/2002.05700"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. arXiv:2305.08510, 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada. Control Barrier Functions: Theory and Applications. arXiv:1903.11199, 2019.","identifier":"arXiv:1903.11199","url":"https://arxiv.org/abs/1903.11199"},{"text":"M. S. Arulampalam, S. Maskell, N. Gordon, and T. Clapp. A Tutorial on Particle Filters for Online Nonlinear/Non-Gaussian Bayesian Tracking. IEEE Transactions on Signal Processing, 50(2):174–188, 2002.","identifier":"doi:10.1109/78.978374","url":"https://doi.org/10.1109/78.978374"},{"text":"E. A. Wan and R. van der Merwe. The Unscented Kalman Filter for Nonlinear Estimation. Proceedings of the IEEE Adaptive Systems for Signal Processing, Communications, and Control Symposium, 153–158, 2000.","identifier":"doi:10.1109/ASSPCC.2000.882463","url":"https://doi.org/10.1109/ASSPCC.2000.882463"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2608.00001","slug":"2608.00001","registered_at":"2026-08-08T08:45:00Z","updated_at":"2026-08-10T09:26:00Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2608.00001","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2608.00001","tex_path":"articles/2608.00001/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-08-08T11:02:00Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2608.00001"},{"kagg_id":"KAGG:2607.00009","slug":"2607.00009","title":"Predictive Contact-Aware Path Following for Agile Quadrupeds in Narrow Passages","authors":[{"name":"Felix Wagner","affiliation":"Legged Robotics Lab, Stanford University, Stanford, CA, USA"},{"name":"Ananya Iyer","affiliation":"Legged Robotics Lab, Stanford University, Stanford, CA, USA"},{"name":"Meilin Jiang","affiliation":"Legged Robotics Lab, Stanford University, Stanford, CA, USA"},{"name":"Carlos Ramirez","affiliation":"Legged Robotics Lab, Stanford University, Stanford, CA, USA"}],"affiliation":null,"abstract":"In narrow passages, the contact sequence of a quadruped's legs and the lag between commanded and realized body velocity decide whether the robot clears the gap, yet most path followers treat the body as a point with ideal velocity response. This paper presents a predictive contact-aware path follower that couples a calibrated model of the opaque velocity interface, obtained with GAUGE, with the configuration-based motion prediction of AgniNav. For each candidate command sequence, the follower predicts the realized body trajectory, the gait phase, and the swept leg envelope over a 0.5-1.0 s horizon; it schedules body poses and contact phases so that the legs nearest the passage boundary enter in a favorable phase, and it rejects passage entries whose predicted trajectory cannot clear the gap within the clearance and time budgets. The approach is evaluated in an Isaac Lab narrow-gate suite with doorway-scale gaps of 55-90 cm, rope barriers, and pipe runs, and in Unitree Go2 trials through a 72 cm doorway-scale gap and two rope barriers at setpoints of 1.0-3.0 m/s. Compared with point-body path following under an ideal velocity assumption, the method raises first-attempt success from 75.5 to 84.6 percent and shortens time to pass from 3.2 to 2.8 s, at the cost of a smaller mean clearance margin of 13.2 cm versus 15.1 cm. Removing contact-phase scheduling, future-range rejection, or the calibrated interface model reduces success to between 72.4 and 74.4 percent.","publication_date":"2026-07-30","registered_at":"2026-08-04T16:52:00Z","updated_at":"2026-08-04T19:52:00Z","status":"published","html_path":null,"pdf_path":"articles/2607.00009/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"S. Macenski, F. Martín, R. White, and J. Ginés Clavero. The Marathon 2: A Navigation System. arXiv:2003.00368, 2020.","identifier":"arXiv:2003.00368","url":"https://arxiv.org/abs/2003.00368"},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"T. Zang, H. Huang, S. Cheng, and W. Zhang. GAUGE: Planner-Conditioned Active Calibration of Opaque Quadruped Velocity Interfaces. arXiv:2607.02841, 2026.","identifier":"arXiv:2607.02841","url":"https://arxiv.org/abs/2607.02841"},{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"G. B. Margolis and P. Agrawal. Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior. arXiv:2212.03238, 2022.","identifier":"arXiv:2212.03238","url":"https://arxiv.org/abs/2212.03238"},{"text":"G. B. Margolis, G. Yang, K. Paigwar, T. Chen, and P. Agrawal. Rapid Locomotion via Reinforcement Learning. arXiv:2205.02824, 2022.","identifier":"arXiv:2205.02824","url":"https://arxiv.org/abs/2205.02824"},{"text":"J. Tan, T. Zhang, E. Coumans, A. Iscen, Y. Bai, D. Hafner, S. Bohez, and V. Vanhoucke. Sim-to-Real: Learning Agile Locomotion For Quadruped Robots. arXiv:1804.10332, 2018.","identifier":"arXiv:1804.10332","url":"https://arxiv.org/abs/1804.10332"},{"text":"F. Farshidian, E. Jelavic, A. Satapathy, M. Giftthaler, and J. Buchli. Real-Time Motion Planning of Legged Robots: A Model Predictive Control Approach. arXiv:1710.04029, 2017.","identifier":"arXiv:1710.04029","url":"https://arxiv.org/abs/1710.04029"},{"text":"D. Kim, J. Di Carlo, B. Katz, G. Bledt, and S. Kim. Highly Dynamic Quadruped Locomotion via Whole-Body Impulse Control and Model Predictive Control. arXiv:1909.06586, 2019.","identifier":"arXiv:1909.06586","url":"https://arxiv.org/abs/1909.06586"},{"text":"R. Grandia, F. Jenelten, S. Yang, F. Farshidian, and M. Hutter. Perceptive Locomotion through Nonlinear Model Predictive Control. arXiv:2208.08373, 2022.","identifier":"arXiv:2208.08373","url":"https://arxiv.org/abs/2208.08373"},{"text":"F. Jenelten, J. He, F. Farshidian, and M. Hutter. DTC: Deep Tracking Control. arXiv:2309.15462, 2023.","identifier":"arXiv:2309.15462","url":"https://arxiv.org/abs/2309.15462"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. arXiv:2309.05665, 2023.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. arXiv:2309.14341, 2023.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. arXiv:2211.07638, 2022.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"Z. Fu, A. Kumar, A. Agarwal, H. Qi, J. Malik, and D. Pathak. Coupling Vision and Proprioception for Navigation of Legged Robots. arXiv:2112.02094, 2021.","identifier":"arXiv:2112.02094","url":"https://arxiv.org/abs/2112.02094"},{"text":"J. Truong, D. Yarats, T. Li, F. Meier, S. Chernova, D. Batra, and A. Rai. Learning Navigation Skills for Legged Robots with Learned Robot Embeddings. arXiv:2011.12255, 2020.","identifier":"arXiv:2011.12255","url":"https://arxiv.org/abs/2011.12255"},{"text":"J. Frey, D. Hoeller, S. Khattak, and M. Hutter. Locomotion Policy Guided Traversability Learning using Volumetric Representations of Complex Environments. arXiv:2203.15854, 2022.","identifier":"arXiv:2203.15854","url":"https://arxiv.org/abs/2203.15854"},{"text":"T. Miki, L. Wellhausen, R. Grandia, F. Jenelten, T. Homberger, and M. Hutter. Elevation Mapping for Locomotion and Navigation using GPU. arXiv:2204.12876, 2022.","identifier":"arXiv:2204.12876","url":"https://arxiv.org/abs/2204.12876"},{"text":"G. Williams, A. Aldrich, and E. Theodorou. Model Predictive Path Integral Control using Covariance Variable Importance Sampling. arXiv:1509.01149, 2015.","identifier":"arXiv:1509.01149","url":"https://arxiv.org/abs/1509.01149"},{"text":"G. Kahn, P. Abbeel, and S. Levine. BADGR: An Autonomous Self-Supervised Learning-Based Navigation System. arXiv:2002.05700, 2020.","identifier":"arXiv:2002.05700","url":"https://arxiv.org/abs/2002.05700"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Sridhar, D. Shah, C. Glossop, and S. Levine. NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration. arXiv:2310.07896, 2023.","identifier":"arXiv:2310.07896","url":"https://arxiv.org/abs/2310.07896"},{"text":"A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada. Control Barrier Functions: Theory and Applications. arXiv:1903.11199, 2019.","identifier":"arXiv:1903.11199","url":"https://arxiv.org/abs/1903.11199"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. arXiv:1703.06907, 2017.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. arXiv:1710.06537, 2017.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"},{"text":"V. Makoviychuk, L. Wawrzyniak, Y. Guo, M. Lu, K. Storey, M. Macklin, D. Hoeller, N. Rudin, A. Allshire, A. Handa, and G. State. Isaac Gym: High Performance GPU-Based Physics Simulation for Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal et al.. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. arXiv:2301.04195, 2023.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"A. Kadian, J. Truong, A. Gokaslan, A. Clegg, E. Wijmans, S. Lee, M. Savva, S. Chernova, and D. Batra. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance?. arXiv:1912.06321, 2019.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"},{"text":"N. Yokoyama, S. Ha, and D. Batra. Success Weighted by Completion Time: A Dynamics-Aware Evaluation Criteria for Embodied Navigation. arXiv:2103.08022, 2021.","identifier":"arXiv:2103.08022","url":"https://arxiv.org/abs/2103.08022"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2607.00009","metadata":{"title":"Predictive Contact-Aware Path Following for Agile Quadrupeds in Narrow Passages","authors":[{"name":"Felix Wagner","affiliation":"Legged Robotics Lab, Stanford University, Stanford, CA, USA"},{"name":"Ananya Iyer","affiliation":"Legged Robotics Lab, Stanford University, Stanford, CA, USA"},{"name":"Meilin Jiang","affiliation":"Legged Robotics Lab, Stanford University, Stanford, CA, USA"},{"name":"Carlos Ramirez","affiliation":"Legged Robotics Lab, Stanford University, Stanford, CA, USA"}],"publication_date":"2026-07-30","abstract":"In narrow passages, the contact sequence of a quadruped's legs and the lag between commanded and realized body velocity decide whether the robot clears the gap, yet most path followers treat the body as a point with ideal velocity response. This paper presents a predictive contact-aware path follower that couples a calibrated model of the opaque velocity interface, obtained with GAUGE, with the configuration-based motion prediction of AgniNav. For each candidate command sequence, the follower predicts the realized body trajectory, the gait phase, and the swept leg envelope over a 0.5-1.0 s horizon; it schedules body poses and contact phases so that the legs nearest the passage boundary enter in a favorable phase, and it rejects passage entries whose predicted trajectory cannot clear the gap within the clearance and time budgets. The approach is evaluated in an Isaac Lab narrow-gate suite with doorway-scale gaps of 55-90 cm, rope barriers, and pipe runs, and in Unitree Go2 trials through a 72 cm doorway-scale gap and two rope barriers at setpoints of 1.0-3.0 m/s. Compared with point-body path following under an ideal velocity assumption, the method raises first-attempt success from 75.5 to 84.6 percent and shortens time to pass from 3.2 to 2.8 s, at the cost of a smaller mean clearance margin of 13.2 cm versus 15.1 cm. Removing contact-phase scheduling, future-range rejection, or the calibrated interface model reduces success to between 72.4 and 74.4 percent.","references":[{"text":"S. Macenski, F. Martín, R. White, and J. Ginés Clavero. The Marathon 2: A Navigation System. arXiv:2003.00368, 2020.","identifier":"arXiv:2003.00368","url":"https://arxiv.org/abs/2003.00368"},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"T. Zang, H. Huang, S. Cheng, and W. Zhang. GAUGE: Planner-Conditioned Active Calibration of Opaque Quadruped Velocity Interfaces. arXiv:2607.02841, 2026.","identifier":"arXiv:2607.02841","url":"https://arxiv.org/abs/2607.02841"},{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"G. B. Margolis and P. Agrawal. Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior. arXiv:2212.03238, 2022.","identifier":"arXiv:2212.03238","url":"https://arxiv.org/abs/2212.03238"},{"text":"G. B. Margolis, G. Yang, K. Paigwar, T. Chen, and P. Agrawal. Rapid Locomotion via Reinforcement Learning. arXiv:2205.02824, 2022.","identifier":"arXiv:2205.02824","url":"https://arxiv.org/abs/2205.02824"},{"text":"J. Tan, T. Zhang, E. Coumans, A. Iscen, Y. Bai, D. Hafner, S. Bohez, and V. Vanhoucke. Sim-to-Real: Learning Agile Locomotion For Quadruped Robots. arXiv:1804.10332, 2018.","identifier":"arXiv:1804.10332","url":"https://arxiv.org/abs/1804.10332"},{"text":"F. Farshidian, E. Jelavic, A. Satapathy, M. Giftthaler, and J. Buchli. Real-Time Motion Planning of Legged Robots: A Model Predictive Control Approach. arXiv:1710.04029, 2017.","identifier":"arXiv:1710.04029","url":"https://arxiv.org/abs/1710.04029"},{"text":"D. Kim, J. Di Carlo, B. Katz, G. Bledt, and S. Kim. Highly Dynamic Quadruped Locomotion via Whole-Body Impulse Control and Model Predictive Control. arXiv:1909.06586, 2019.","identifier":"arXiv:1909.06586","url":"https://arxiv.org/abs/1909.06586"},{"text":"R. Grandia, F. Jenelten, S. Yang, F. Farshidian, and M. Hutter. Perceptive Locomotion through Nonlinear Model Predictive Control. arXiv:2208.08373, 2022.","identifier":"arXiv:2208.08373","url":"https://arxiv.org/abs/2208.08373"},{"text":"F. Jenelten, J. He, F. Farshidian, and M. Hutter. DTC: Deep Tracking Control. arXiv:2309.15462, 2023.","identifier":"arXiv:2309.15462","url":"https://arxiv.org/abs/2309.15462"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. arXiv:2309.05665, 2023.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. arXiv:2309.14341, 2023.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. arXiv:2211.07638, 2022.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"Z. Fu, A. Kumar, A. Agarwal, H. Qi, J. Malik, and D. Pathak. Coupling Vision and Proprioception for Navigation of Legged Robots. arXiv:2112.02094, 2021.","identifier":"arXiv:2112.02094","url":"https://arxiv.org/abs/2112.02094"},{"text":"J. Truong, D. Yarats, T. Li, F. Meier, S. Chernova, D. Batra, and A. Rai. Learning Navigation Skills for Legged Robots with Learned Robot Embeddings. arXiv:2011.12255, 2020.","identifier":"arXiv:2011.12255","url":"https://arxiv.org/abs/2011.12255"},{"text":"J. Frey, D. Hoeller, S. Khattak, and M. Hutter. Locomotion Policy Guided Traversability Learning using Volumetric Representations of Complex Environments. arXiv:2203.15854, 2022.","identifier":"arXiv:2203.15854","url":"https://arxiv.org/abs/2203.15854"},{"text":"T. Miki, L. Wellhausen, R. Grandia, F. Jenelten, T. Homberger, and M. Hutter. Elevation Mapping for Locomotion and Navigation using GPU. arXiv:2204.12876, 2022.","identifier":"arXiv:2204.12876","url":"https://arxiv.org/abs/2204.12876"},{"text":"G. Williams, A. Aldrich, and E. Theodorou. Model Predictive Path Integral Control using Covariance Variable Importance Sampling. arXiv:1509.01149, 2015.","identifier":"arXiv:1509.01149","url":"https://arxiv.org/abs/1509.01149"},{"text":"G. Kahn, P. Abbeel, and S. Levine. BADGR: An Autonomous Self-Supervised Learning-Based Navigation System. arXiv:2002.05700, 2020.","identifier":"arXiv:2002.05700","url":"https://arxiv.org/abs/2002.05700"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Sridhar, D. Shah, C. Glossop, and S. Levine. NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration. arXiv:2310.07896, 2023.","identifier":"arXiv:2310.07896","url":"https://arxiv.org/abs/2310.07896"},{"text":"A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada. Control Barrier Functions: Theory and Applications. arXiv:1903.11199, 2019.","identifier":"arXiv:1903.11199","url":"https://arxiv.org/abs/1903.11199"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. arXiv:1703.06907, 2017.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. arXiv:1710.06537, 2017.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"},{"text":"V. Makoviychuk, L. Wawrzyniak, Y. Guo, M. Lu, K. Storey, M. Macklin, D. Hoeller, N. Rudin, A. Allshire, A. Handa, and G. State. Isaac Gym: High Performance GPU-Based Physics Simulation for Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal et al.. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. arXiv:2301.04195, 2023.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"A. Kadian, J. Truong, A. Gokaslan, A. Clegg, E. Wijmans, S. Lee, M. Savva, S. Chernova, and D. Batra. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance?. arXiv:1912.06321, 2019.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"},{"text":"N. Yokoyama, S. Ha, and D. Batra. Success Weighted by Completion Time: A Dynamics-Aware Evaluation Criteria for Embodied Navigation. arXiv:2103.08022, 2021.","identifier":"arXiv:2103.08022","url":"https://arxiv.org/abs/2103.08022"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2607.00009","slug":"2607.00009","registered_at":"2026-08-04T16:52:00Z","updated_at":"2026-08-04T19:52:00Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2607.00009","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2607.00009","tex_path":"articles/2607.00009/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-08-04T19:09:00Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2607.00009"},{"kagg_id":"KAGG:2607.00008","slug":"2607.00008","title":"Configuration Abstractions for Heterogeneous Legged Robot Fleets","authors":[{"name":"Wenjie Zhao","affiliation":"Intelligent Mobility Lab, University of Science and Technology of China, Hefei, China"},{"name":"Mingxuan Li","affiliation":"Intelligent Mobility Lab, University of Science and Technology of China, Hefei, China"},{"name":"Boyang Sun","affiliation":"Intelligent Mobility Lab, University of Science and Technology of China, Hefei, China"}],"affiliation":null,"abstract":"Heterogeneous legged robot fleets combine platforms that differ in body length, payload, locomotion mode, and sensor mounting, and these differences normally force each embodiment onto its own local planner. This fragmentation duplicates code, complicates validation, and allows planning behavior to drift between robots that share a mission. This paper extends the configuration-driven planning idea of AgniNav from a single robot to the fleet level. The proposed abstraction has three parts: a compact, typed configuration schema that separates collision geometry, kinematic and payload limits, locomotion class, and sensor rig; a configuration registry service that validates descriptors, issues content-addressed handles, and caches derived planning quantities; and a planner interface that accepts configuration handles instead of compiled-in robot parameters. One bit-identical planning binary was evaluated on a simulated fleet of four morphologies, namely Go2-class and B2-class quadrupeds, a 1.4 m humanoid, and a four-wheel-drive rover, each with a distinct sensor rig, across 15 mixed-fleet scenarios with 5 seeds each. Compared with four separate per-embodiment code paths, the shared planner raised shared module reuse from 83.2 to 87.2 percent, lowered configuration load latency from 1.9 to 1.8 ms and 95th-percentile replanning latency from 38 to 34 ms, and raised plan quality parity from 81.2 to 96.4 percent, whereas a shared planner driven by ad hoc configuration flags reached 88.2 percent. Ablations attribute the largest share of the gain to the sensor-rig abstraction. The remaining failures involve differential-drive behavior on steep slopes and registry cold starts under concurrent load.","publication_date":"2026-07-27","registered_at":"2026-08-01T11:07:00Z","updated_at":"2026-08-01T17:38:00Z","status":"published","html_path":null,"pdf_path":"articles/2607.00008/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Sridhar, D. Shah, C. Glossop, and S. Levine. NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration. arXiv:2310.07896, 2023.","identifier":"arXiv:2310.07896","url":"https://arxiv.org/abs/2310.07896"},{"text":"A. Eftekhar et al. The One RING: a Robotic Indoor Navigation Generalist. arXiv:2412.14401, 2024.","identifier":"arXiv:2412.14401","url":"https://arxiv.org/abs/2412.14401"},{"text":"J. Zhang et al. Embodied Navigation Foundation Model. arXiv:2509.12129, 2025.","identifier":"arXiv:2509.12129","url":"https://arxiv.org/abs/2509.12129"},{"text":"J. Zhang, Y. Du, X. Guo, S. Sun, X. Liu, Y. Sun, G. Lu, W. Sui, and J. Li. EA-Nav: Learning Safe Visual Navigation Policies with Embodiment Awareness. arXiv:2607.19880, 2026.","identifier":"arXiv:2607.19880","url":"https://arxiv.org/abs/2607.19880"},{"text":"J. Yang, C. Glossop, A. Bhorkar, D. Shah, Q. Vuong, C. Finn, D. Sadigh, and S. Levine. Pushing the Limits of Cross-Embodiment Learning for Manipulation and Navigation. arXiv:2402.19432, 2024.","identifier":"arXiv:2402.19432","url":"https://arxiv.org/abs/2402.19432"},{"text":"R. Doshi, H. Walke, O. Mees, S. Dasari, and S. Levine. Scaling Cross-Embodied Learning: One Policy for Manipulation, Navigation, Locomotion and Aviation. arXiv:2408.11812, 2024.","identifier":"arXiv:2408.11812","url":"https://arxiv.org/abs/2408.11812"},{"text":"Open X-Embodiment Collaboration. Open X-Embodiment: Robotic Learning Datasets and RT-X Models. arXiv:2310.08864, 2023.","identifier":"arXiv:2310.08864","url":"https://arxiv.org/abs/2310.08864"},{"text":"Octo Model Team. Octo: An Open-Source Generalist Robot Policy. arXiv:2405.12213, 2024.","identifier":"arXiv:2405.12213","url":"https://arxiv.org/abs/2405.12213"},{"text":"G. Kahn, P. Abbeel, and S. Levine. BADGR: An Autonomous Self-Supervised Learning-Based Navigation System. arXiv:2002.05700, 2020.","identifier":"arXiv:2002.05700","url":"https://arxiv.org/abs/2002.05700"},{"text":"G. Williams, A. Aldrich, and E. Theodorou. Model Predictive Path Integral Control using Covariance Variable Importance Sampling. arXiv:1509.01149, 2015.","identifier":"arXiv:1509.01149","url":"https://arxiv.org/abs/1509.01149"},{"text":"X. Zhou, Z. Wang, H. Ye, C. Xu, and F. Gao. EGO-Planner: An ESDF-free Gradient-based Local Planner for Quadrotors. arXiv:2008.08835, 2020.","identifier":"arXiv:2008.08835","url":"https://arxiv.org/abs/2008.08835"},{"text":"S. Macenski, F. Martín, R. White, and J. Ginés Clavero. The Marathon 2: A Navigation System. arXiv:2003.00368, 2020.","identifier":"arXiv:2003.00368","url":"https://arxiv.org/abs/2003.00368"},{"text":"T. Miki, L. Wellhausen, R. Grandia, F. Jenelten, T. Homberger, and M. Hutter. Elevation Mapping for Locomotion and Navigation using GPU. arXiv:2204.12876, 2022.","identifier":"arXiv:2204.12876","url":"https://arxiv.org/abs/2204.12876"},{"text":"M. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. arXiv:2305.08510, 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"D. D. Fan, K. Otsu, Y. Kubo, A. Dixit, J. Burdick, and A. Agha-Mohammadi. STEP: Stochastic Traversability Evaluation and Planning for Risk-Aware Off-road Navigation. arXiv:2103.02828, 2021.","identifier":"arXiv:2103.02828","url":"https://arxiv.org/abs/2103.02828"},{"text":"R. Grandia, F. Jenelten, S. Yang, F. Farshidian, and M. Hutter. Perceptive Locomotion through Nonlinear Model Predictive Control. arXiv:2208.08373, 2022.","identifier":"arXiv:2208.08373","url":"https://arxiv.org/abs/2208.08373"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"J. Hu, S. Yuan, G. C. R. Bethala, A. Tzes, and Y. Fang. Learning Adaptive Safety Margins for Visual Navigation. arXiv:2607.18200, 2026.","identifier":"arXiv:2607.18200","url":"https://arxiv.org/abs/2607.18200"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. arXiv:2211.07638, 2022.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"V. Makoviychuk et al. Isaac Gym: High Performance GPU-Based Physics Simulation for Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"G. Feng, H. Zhang, Z. Li, X. B. Peng, B. Basireddy, L. Yue, Z. Song, L. Yang, Y. Liu, K. Sreenath, and S. Levine. GenLoco: Generalized Locomotion Controllers for Quadrupedal Robots. arXiv:2209.05309, 2022.","identifier":"arXiv:2209.05309","url":"https://arxiv.org/abs/2209.05309"},{"text":"M. Shafiee, G. Bellegarda, and A. Ijspeert. ManyQuadrupeds: Learning a Single Locomotion Policy for Diverse Quadruped Robots. arXiv:2310.10486, 2023.","identifier":"arXiv:2310.10486","url":"https://arxiv.org/abs/2310.10486"},{"text":"N. Bohlinger, G. Czechmanowski, M. Krupka, P. Kicki, K. Walas, J. Peters, and D. Tateo. One Policy to Run Them All: an End-to-end Learning Approach to Multi-Embodiment Locomotion. arXiv:2409.06366, 2024.","identifier":"arXiv:2409.06366","url":"https://arxiv.org/abs/2409.06366"},{"text":"I. Radosavovic, T. Xiao, B. Zhang, T. Darrell, J. Malik, and K. Sreenath. Real-World Humanoid Locomotion with Reinforcement Learning. arXiv:2303.03381, 2023.","identifier":"arXiv:2303.03381","url":"https://arxiv.org/abs/2303.03381"},{"text":"S. Macenski, T. Foote, B. Gerkey, C. Lalancette, and W. Woodall. Robot Operating System 2: Design, Architecture, and Uses in the Wild. arXiv:2211.07752, 2022.","identifier":"arXiv:2211.07752","url":"https://arxiv.org/abs/2211.07752"},{"text":"D. Coleman, I. Sucan, S. Chitta, and N. Correll. Reducing the Barrier to Entry of Complex Robotic Software: a MoveIt! Case Study. arXiv:1404.3785, 2014.","identifier":"arXiv:1404.3785","url":"https://arxiv.org/abs/1404.3785"},{"text":"M. Colledanchise and P. Ögren. Behavior Trees in Robotics and AI: An Introduction. arXiv:1709.00084, 2017.","identifier":"arXiv:1709.00084","url":"https://arxiv.org/abs/1709.00084"},{"text":"M. Savva et al. Habitat: A Platform for Embodied AI Research. arXiv:1904.01201, 2019.","identifier":"arXiv:1904.01201","url":"https://arxiv.org/abs/1904.01201"},{"text":"F. Xia, A. Zamir, Z.-Y. He, A. Sax, J. Malik, and S. Savarese. Gibson Env: Real-World Perception for Embodied Agents. arXiv:1808.10654, 2018.","identifier":"arXiv:1808.10654","url":"https://arxiv.org/abs/1808.10654"},{"text":"A. Kadian et al. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance? arXiv:1912.06321, 2019.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"},{"text":"P. Anderson et al. On Evaluation of Embodied Navigation Agents. arXiv:1807.06757, 2018.","identifier":"arXiv:1807.06757","url":"https://arxiv.org/abs/1807.06757"},{"text":"N. Yokoyama, S. Ha, and D. Batra. Success Weighted by Completion Time: A Dynamics-Aware Evaluation Criteria for Embodied Navigation. arXiv:2103.08022, 2021.","identifier":"arXiv:2103.08022","url":"https://arxiv.org/abs/2103.08022"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2607.00008","metadata":{"title":"Configuration Abstractions for Heterogeneous Legged Robot Fleets","authors":[{"name":"Wenjie Zhao","affiliation":"Intelligent Mobility Lab, University of Science and Technology of China, Hefei, China"},{"name":"Mingxuan Li","affiliation":"Intelligent Mobility Lab, University of Science and Technology of China, Hefei, China"},{"name":"Boyang Sun","affiliation":"Intelligent Mobility Lab, University of Science and Technology of China, Hefei, China"}],"publication_date":"2026-07-27","abstract":"Heterogeneous legged robot fleets combine platforms that differ in body length, payload, locomotion mode, and sensor mounting, and these differences normally force each embodiment onto its own local planner. This fragmentation duplicates code, complicates validation, and allows planning behavior to drift between robots that share a mission. This paper extends the configuration-driven planning idea of AgniNav from a single robot to the fleet level. The proposed abstraction has three parts: a compact, typed configuration schema that separates collision geometry, kinematic and payload limits, locomotion class, and sensor rig; a configuration registry service that validates descriptors, issues content-addressed handles, and caches derived planning quantities; and a planner interface that accepts configuration handles instead of compiled-in robot parameters. One bit-identical planning binary was evaluated on a simulated fleet of four morphologies, namely Go2-class and B2-class quadrupeds, a 1.4 m humanoid, and a four-wheel-drive rover, each with a distinct sensor rig, across 15 mixed-fleet scenarios with 5 seeds each. Compared with four separate per-embodiment code paths, the shared planner raised shared module reuse from 83.2 to 87.2 percent, lowered configuration load latency from 1.9 to 1.8 ms and 95th-percentile replanning latency from 38 to 34 ms, and raised plan quality parity from 81.2 to 96.4 percent, whereas a shared planner driven by ad hoc configuration flags reached 88.2 percent. Ablations attribute the largest share of the gain to the sensor-rig abstraction. The remaining failures involve differential-drive behavior on steep slopes and registry cold starts under concurrent load.","references":[{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Sridhar, D. Shah, C. Glossop, and S. Levine. NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration. arXiv:2310.07896, 2023.","identifier":"arXiv:2310.07896","url":"https://arxiv.org/abs/2310.07896"},{"text":"A. Eftekhar et al. The One RING: a Robotic Indoor Navigation Generalist. arXiv:2412.14401, 2024.","identifier":"arXiv:2412.14401","url":"https://arxiv.org/abs/2412.14401"},{"text":"J. Zhang et al. Embodied Navigation Foundation Model. arXiv:2509.12129, 2025.","identifier":"arXiv:2509.12129","url":"https://arxiv.org/abs/2509.12129"},{"text":"J. Zhang, Y. Du, X. Guo, S. Sun, X. Liu, Y. Sun, G. Lu, W. Sui, and J. Li. EA-Nav: Learning Safe Visual Navigation Policies with Embodiment Awareness. arXiv:2607.19880, 2026.","identifier":"arXiv:2607.19880","url":"https://arxiv.org/abs/2607.19880"},{"text":"J. Yang, C. Glossop, A. Bhorkar, D. Shah, Q. Vuong, C. Finn, D. Sadigh, and S. Levine. Pushing the Limits of Cross-Embodiment Learning for Manipulation and Navigation. arXiv:2402.19432, 2024.","identifier":"arXiv:2402.19432","url":"https://arxiv.org/abs/2402.19432"},{"text":"R. Doshi, H. Walke, O. Mees, S. Dasari, and S. Levine. Scaling Cross-Embodied Learning: One Policy for Manipulation, Navigation, Locomotion and Aviation. arXiv:2408.11812, 2024.","identifier":"arXiv:2408.11812","url":"https://arxiv.org/abs/2408.11812"},{"text":"Open X-Embodiment Collaboration. Open X-Embodiment: Robotic Learning Datasets and RT-X Models. arXiv:2310.08864, 2023.","identifier":"arXiv:2310.08864","url":"https://arxiv.org/abs/2310.08864"},{"text":"Octo Model Team. Octo: An Open-Source Generalist Robot Policy. arXiv:2405.12213, 2024.","identifier":"arXiv:2405.12213","url":"https://arxiv.org/abs/2405.12213"},{"text":"G. Kahn, P. Abbeel, and S. Levine. BADGR: An Autonomous Self-Supervised Learning-Based Navigation System. arXiv:2002.05700, 2020.","identifier":"arXiv:2002.05700","url":"https://arxiv.org/abs/2002.05700"},{"text":"G. Williams, A. Aldrich, and E. Theodorou. Model Predictive Path Integral Control using Covariance Variable Importance Sampling. arXiv:1509.01149, 2015.","identifier":"arXiv:1509.01149","url":"https://arxiv.org/abs/1509.01149"},{"text":"X. Zhou, Z. Wang, H. Ye, C. Xu, and F. Gao. EGO-Planner: An ESDF-free Gradient-based Local Planner for Quadrotors. arXiv:2008.08835, 2020.","identifier":"arXiv:2008.08835","url":"https://arxiv.org/abs/2008.08835"},{"text":"S. Macenski, F. Martín, R. White, and J. Ginés Clavero. The Marathon 2: A Navigation System. arXiv:2003.00368, 2020.","identifier":"arXiv:2003.00368","url":"https://arxiv.org/abs/2003.00368"},{"text":"T. Miki, L. Wellhausen, R. Grandia, F. Jenelten, T. Homberger, and M. Hutter. Elevation Mapping for Locomotion and Navigation using GPU. arXiv:2204.12876, 2022.","identifier":"arXiv:2204.12876","url":"https://arxiv.org/abs/2204.12876"},{"text":"M. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. arXiv:2305.08510, 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"D. D. Fan, K. Otsu, Y. Kubo, A. Dixit, J. Burdick, and A. Agha-Mohammadi. STEP: Stochastic Traversability Evaluation and Planning for Risk-Aware Off-road Navigation. arXiv:2103.02828, 2021.","identifier":"arXiv:2103.02828","url":"https://arxiv.org/abs/2103.02828"},{"text":"R. Grandia, F. Jenelten, S. Yang, F. Farshidian, and M. Hutter. Perceptive Locomotion through Nonlinear Model Predictive Control. arXiv:2208.08373, 2022.","identifier":"arXiv:2208.08373","url":"https://arxiv.org/abs/2208.08373"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"J. Hu, S. Yuan, G. C. R. Bethala, A. Tzes, and Y. Fang. Learning Adaptive Safety Margins for Visual Navigation. arXiv:2607.18200, 2026.","identifier":"arXiv:2607.18200","url":"https://arxiv.org/abs/2607.18200"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. arXiv:2211.07638, 2022.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"V. Makoviychuk et al. Isaac Gym: High Performance GPU-Based Physics Simulation for Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"G. Feng, H. Zhang, Z. Li, X. B. Peng, B. Basireddy, L. Yue, Z. Song, L. Yang, Y. Liu, K. Sreenath, and S. Levine. GenLoco: Generalized Locomotion Controllers for Quadrupedal Robots. arXiv:2209.05309, 2022.","identifier":"arXiv:2209.05309","url":"https://arxiv.org/abs/2209.05309"},{"text":"M. Shafiee, G. Bellegarda, and A. Ijspeert. ManyQuadrupeds: Learning a Single Locomotion Policy for Diverse Quadruped Robots. arXiv:2310.10486, 2023.","identifier":"arXiv:2310.10486","url":"https://arxiv.org/abs/2310.10486"},{"text":"N. Bohlinger, G. Czechmanowski, M. Krupka, P. Kicki, K. Walas, J. Peters, and D. Tateo. One Policy to Run Them All: an End-to-end Learning Approach to Multi-Embodiment Locomotion. arXiv:2409.06366, 2024.","identifier":"arXiv:2409.06366","url":"https://arxiv.org/abs/2409.06366"},{"text":"I. Radosavovic, T. Xiao, B. Zhang, T. Darrell, J. Malik, and K. Sreenath. Real-World Humanoid Locomotion with Reinforcement Learning. arXiv:2303.03381, 2023.","identifier":"arXiv:2303.03381","url":"https://arxiv.org/abs/2303.03381"},{"text":"S. Macenski, T. Foote, B. Gerkey, C. Lalancette, and W. Woodall. Robot Operating System 2: Design, Architecture, and Uses in the Wild. arXiv:2211.07752, 2022.","identifier":"arXiv:2211.07752","url":"https://arxiv.org/abs/2211.07752"},{"text":"D. Coleman, I. Sucan, S. Chitta, and N. Correll. Reducing the Barrier to Entry of Complex Robotic Software: a MoveIt! Case Study. arXiv:1404.3785, 2014.","identifier":"arXiv:1404.3785","url":"https://arxiv.org/abs/1404.3785"},{"text":"M. Colledanchise and P. Ögren. Behavior Trees in Robotics and AI: An Introduction. arXiv:1709.00084, 2017.","identifier":"arXiv:1709.00084","url":"https://arxiv.org/abs/1709.00084"},{"text":"M. Savva et al. Habitat: A Platform for Embodied AI Research. arXiv:1904.01201, 2019.","identifier":"arXiv:1904.01201","url":"https://arxiv.org/abs/1904.01201"},{"text":"F. Xia, A. Zamir, Z.-Y. He, A. Sax, J. Malik, and S. Savarese. Gibson Env: Real-World Perception for Embodied Agents. arXiv:1808.10654, 2018.","identifier":"arXiv:1808.10654","url":"https://arxiv.org/abs/1808.10654"},{"text":"A. Kadian et al. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance? arXiv:1912.06321, 2019.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"},{"text":"P. Anderson et al. On Evaluation of Embodied Navigation Agents. arXiv:1807.06757, 2018.","identifier":"arXiv:1807.06757","url":"https://arxiv.org/abs/1807.06757"},{"text":"N. Yokoyama, S. Ha, and D. Batra. Success Weighted by Completion Time: A Dynamics-Aware Evaluation Criteria for Embodied Navigation. arXiv:2103.08022, 2021.","identifier":"arXiv:2103.08022","url":"https://arxiv.org/abs/2103.08022"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2607.00008","slug":"2607.00008","registered_at":"2026-08-01T11:07:00Z","updated_at":"2026-08-01T17:38:00Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2607.00008","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2607.00008","tex_path":"articles/2607.00008/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-08-01T13:24:00Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2607.00008"},{"kagg_id":"KAGG:2607.00007","slug":"2607.00007","title":"Multi-Modal Range Fusion for Agile Quadruped Navigation in Dense Vegetation","authors":[{"name":"Farah Idris","affiliation":"Robotics and Control Laboratory, Nanyang Technological University, Singapore"},{"name":"Hao Chen","affiliation":"Robotics and Control Laboratory, Nanyang Technological University, Singapore"},{"name":"Kai Wen Lim","affiliation":"Robotics and Control Laboratory, Nanyang Technological University, Singapore"}],"affiliation":null,"abstract":"Quadruped robots are expected to move quickly through forests, parks, and overgrown terrain, yet LiDAR range sensing degrades precisely there: foliage produces premature returns, thin stems and branches fall between beams, and occlusion hides the rigid structure that constrains motion. This paper presents a range-fusion method that combines LiDAR and stereo-vision range estimates with weights conditioned on where the robot is predicted to move. A configuration-driven local planner, in the sense of AgniNav, supplies the robot's collision envelope and candidate trajectories; a first-order model of velocity tracking converts them into a motion-predicted occupancy field over the sensing grid. Per-cell fusion weights combine this field with modality-specific noise, motion-induced latency error, and a vegetation degradation model whose density parameter is estimated online from LiDAR transmittance. The method is evaluated on a Unitree Go2 carrying a stereo rig and a 32-beam LiDAR, in an Isaac Lab synthetic forest suite with four vegetation density classes and on a 600 m park course through shrub and sapling cover. Compared with LiDAR-only sensing, it reduces range estimate error from 0.99 m to 0.94 m, raises mean speed from 2.07 m/s to 2.31 m/s, and raises obstacle-free success from 84.6 to 90.7 percent, while mean obstacle clearance decreases from 18.1 cm to 16.8 cm. Uniform-weight fusion reaches only 84.9 percent success, and removing any single component lowers success. A failure taxonomy covers dense sapling cover and head-height branches.","publication_date":"2026-07-24","registered_at":"2026-07-25T09:45:00Z","updated_at":"2026-07-25T17:02:00Z","status":"published","html_path":null,"pdf_path":"articles/2607.00007/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. Science Robotics, 2019. arXiv:1901.08652.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. Science Robotics, 2020. arXiv:2010.11251.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. Science Robotics, 2022. arXiv:2201.08117.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"M. Tranzatto et al.. CERBERUS: Autonomous Legged and Aerial Robotic Exploration in the Tunnel and Urban Circuits of the DARPA Subterranean Challenge. Field Robotics, 2022. arXiv:2201.07067.","identifier":"arXiv:2201.07067","url":"https://arxiv.org/abs/2201.07067"},{"text":"A. Bouman et al.. Autonomous Spot: Long-Range Autonomous Exploration of Extreme Environments with Legged Locomotion. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2020. arXiv:2010.09259.","identifier":"arXiv:2010.09259","url":"https://arxiv.org/abs/2010.09259"},{"text":"H. Hirschmüller. Stereo Processing by Semiglobal Matching and Mutual Information. IEEE Transactions on Pattern Analysis and Machine Intelligence, 30(2):328-341, 2008. doi:10.1109/TPAMI.2007.1166.","identifier":"doi:10.1109/TPAMI.2007.1166","url":"https://doi.org/10.1109/TPAMI.2007.1166"},{"text":"J.-R. Chang and Y.-S. Chen. Pyramid Stereo Matching Network. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2018. arXiv:1803.08669.","identifier":"arXiv:1803.08669","url":"https://arxiv.org/abs/1803.08669"},{"text":"L. Lipson, Z. Teed, and J. Deng. RAFT-Stereo: Multilevel Recurrent Field Transforms for Stereo Matching. International Conference on 3D Vision (3DV), 2021. arXiv:2109.07547.","identifier":"arXiv:2109.07547","url":"https://arxiv.org/abs/2109.07547"},{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. Conference on Robot Learning (CoRL), 2021. arXiv:2109.11978.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. Robotics: Science and Systems (RSS), 2021. arXiv:2107.04034.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. Conference on Robot Learning (CoRL), 2022. arXiv:2211.07638.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"R. Yang, M. Zhang, N. Hansen, H. Xu, and X. Wang. Learning Vision-Guided Quadrupedal Locomotion End-to-End with Cross-Modal Transformers. International Conference on Learning Representations (ICLR), 2022. arXiv:2107.03996.","identifier":"arXiv:2107.03996","url":"https://arxiv.org/abs/2107.03996"},{"text":"C. S. Imai et al.. Vision-Guided Quadrupedal Locomotion in the Wild with Multi-Modal Delay Randomization. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2022. arXiv:2109.14549.","identifier":"arXiv:2109.14549","url":"https://arxiv.org/abs/2109.14549"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. Conference on Robot Learning (CoRL), 2023. arXiv:2309.05665.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. arXiv:2309.14341, 2023.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. Science Robotics, 2024. arXiv:2306.14874.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2022. arXiv:2209.12827.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, L. Wellhausen, F. Farshidian, and M. Hutter. Learning a State Representation and Navigation in Cluttered and Dynamic Environments. IEEE Robotics and Automation Letters, 2021. arXiv:2103.04351.","identifier":"arXiv:2103.04351","url":"https://arxiv.org/abs/2103.04351"},{"text":"Z. Fu, A. Kumar, A. Agarwal, H. Qi, J. Malik, and D. Pathak. Coupling Vision and Proprioception for Navigation of Legged Robots. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022. arXiv:2112.02094.","identifier":"arXiv:2112.02094","url":"https://arxiv.org/abs/2112.02094"},{"text":"T. Miki, L. Wellhausen, R. Grandia, F. Jenelten, T. Homberger, and M. Hutter. Elevation Mapping for Locomotion and Navigation using GPU. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2022. arXiv:2204.12876.","identifier":"arXiv:2204.12876","url":"https://arxiv.org/abs/2204.12876"},{"text":"G. Kahn, P. Abbeel, and S. Levine. BADGR: An Autonomous Self-Supervised Learning-Based Navigation System. IEEE Robotics and Automation Letters, 2021. arXiv:2002.05700.","identifier":"arXiv:2002.05700","url":"https://arxiv.org/abs/2002.05700"},{"text":"M. V. Gasparino et al.. WayFAST: Navigation with Predictive Traversability in the Field. IEEE Robotics and Automation Letters, 2022. arXiv:2203.12071.","identifier":"arXiv:2203.12071","url":"https://arxiv.org/abs/2203.12071"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. Robotics: Science and Systems (RSS), 2023. arXiv:2305.08510.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"M. Mattamala et al.. Wild Visual Navigation: Fast Traversability Learning via Pre-Trained Models and Online Self-Supervision. arXiv:2404.07110, 2024.","identifier":"arXiv:2404.07110","url":"https://arxiv.org/abs/2404.07110"},{"text":"A. Loquercio, E. Kaufmann, R. Ranftl, M. Müller, V. Koltun, and D. Scaramuzza. Learning High-Speed Flight in the Wild. Science Robotics, 2021. arXiv:2110.05113.","identifier":"arXiv:2110.05113","url":"https://arxiv.org/abs/2110.05113"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. IEEE International Conference on Robotics and Automation (ICRA), 2023. arXiv:2210.03370.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Fox, W. Burgard, and S. Thrun. The Dynamic Window Approach to Collision Avoidance. IEEE Robotics and Automation Magazine, 4(1):23-33, 1997. doi:10.1109/100.580977.","identifier":"doi:10.1109/100.580977","url":"https://doi.org/10.1109/100.580977"},{"text":"B. Zhou, F. Gao, L. Wang, C. Liu, and S. Shen. Robust and Efficient Quadrotor Trajectory Generation for Fast Autonomous Flight. IEEE Robotics and Automation Letters, 2019. arXiv:1907.01531.","identifier":"arXiv:1907.01531","url":"https://arxiv.org/abs/1907.01531"},{"text":"X. Zhou, Z. Wang, H. Ye, C. Xu, and F. Gao. EGO-Planner: An ESDF-free Gradient-based Local Planner for Quadrotors. IEEE Robotics and Automation Letters, 2021. arXiv:2008.08835.","identifier":"arXiv:2008.08835","url":"https://arxiv.org/abs/2008.08835"},{"text":"X. Xiao, B. Liu, G. Warnell, and P. Stone. Motion Planning and Control for Mobile Robot Navigation Using Machine Learning: a Survey. Autonomous Robots, 2022. arXiv:2011.13112.","identifier":"arXiv:2011.13112","url":"https://arxiv.org/abs/2011.13112"},{"text":"A. Elfes. Using Occupancy Grids for Mobile Robot Perception and Navigation. Computer, 22(6):46-57, 1989. doi:10.1109/2.30720.","identifier":"doi:10.1109/2.30720","url":"https://doi.org/10.1109/2.30720"},{"text":"H. Oleynikova, Z. Taylor, M. Fehr, R. Siegwart, and J. Nieto. Voxblox: Incremental 3D Euclidean Signed Distance Fields for On-Board MAV Planning. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2017. arXiv:1611.03631.","identifier":"arXiv:1611.03631","url":"https://arxiv.org/abs/1611.03631"},{"text":"T. Shan, B. Englot, D. Meyers, W. Wang, C. Ratti, and D. Rus. LIO-SAM: Tightly-coupled Lidar Inertial Odometry via Smoothing and Mapping. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2020. arXiv:2007.00258.","identifier":"arXiv:2007.00258","url":"https://arxiv.org/abs/2007.00258"},{"text":"W. Xu, Y. Cai, D. He, J. Lin, and F. Zhang. FAST-LIO2: Fast Direct LiDAR-Inertial Odometry. IEEE Transactions on Robotics, 2022. arXiv:2107.06829.","identifier":"arXiv:2107.06829","url":"https://arxiv.org/abs/2107.06829"},{"text":"F. Ma and S. Karaman. Sparse-to-Dense: Depth Prediction from Sparse Depth Samples and a Single Image. IEEE International Conference on Robotics and Automation (ICRA), 2018. arXiv:1709.07492.","identifier":"arXiv:1709.07492","url":"https://arxiv.org/abs/1709.07492"},{"text":"J. Uhrig, N. Schneider, L. Schneider, U. Franke, T. Brox, and A. Geiger. Sparsity Invariant CNNs. International Conference on 3D Vision (3DV), 2017. arXiv:1708.06500.","identifier":"arXiv:1708.06500","url":"https://arxiv.org/abs/1708.06500"},{"text":"M. Bijelic et al.. Seeing Through Fog Without Seeing Fog: Deep Multimodal Sensor Fusion in Unseen Adverse Weather. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020. arXiv:1902.08913.","identifier":"arXiv:1902.08913","url":"https://arxiv.org/abs/1902.08913"},{"text":"A. Kendall and Y. Gal. What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision? Advances in Neural Information Processing Systems (NeurIPS), 2017. arXiv:1703.04977.","identifier":"arXiv:1703.04977","url":"https://arxiv.org/abs/1703.04977"},{"text":"V. Makoviychuk et al.. Isaac Gym: High Performance GPU-Based Physics Simulation for Robot Learning. NeurIPS Datasets and Benchmarks Track, 2021. arXiv:2108.10470.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal et al.. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. IEEE Robotics and Automation Letters, 2023. arXiv:2301.04195.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2017. arXiv:1703.06907.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2607.00007","metadata":{"title":"Multi-Modal Range Fusion for Agile Quadruped Navigation in Dense Vegetation","authors":[{"name":"Farah Idris","affiliation":"Robotics and Control Laboratory, Nanyang Technological University, Singapore"},{"name":"Hao Chen","affiliation":"Robotics and Control Laboratory, Nanyang Technological University, Singapore"},{"name":"Kai Wen Lim","affiliation":"Robotics and Control Laboratory, Nanyang Technological University, Singapore"}],"publication_date":"2026-07-24","abstract":"Quadruped robots are expected to move quickly through forests, parks, and overgrown terrain, yet LiDAR range sensing degrades precisely there: foliage produces premature returns, thin stems and branches fall between beams, and occlusion hides the rigid structure that constrains motion. This paper presents a range-fusion method that combines LiDAR and stereo-vision range estimates with weights conditioned on where the robot is predicted to move. A configuration-driven local planner, in the sense of AgniNav, supplies the robot's collision envelope and candidate trajectories; a first-order model of velocity tracking converts them into a motion-predicted occupancy field over the sensing grid. Per-cell fusion weights combine this field with modality-specific noise, motion-induced latency error, and a vegetation degradation model whose density parameter is estimated online from LiDAR transmittance. The method is evaluated on a Unitree Go2 carrying a stereo rig and a 32-beam LiDAR, in an Isaac Lab synthetic forest suite with four vegetation density classes and on a 600 m park course through shrub and sapling cover. Compared with LiDAR-only sensing, it reduces range estimate error from 0.99 m to 0.94 m, raises mean speed from 2.07 m/s to 2.31 m/s, and raises obstacle-free success from 84.6 to 90.7 percent, while mean obstacle clearance decreases from 18.1 cm to 16.8 cm. Uniform-weight fusion reaches only 84.9 percent success, and removing any single component lowers success. A failure taxonomy covers dense sapling cover and head-height branches.","references":[{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. Science Robotics, 2019. arXiv:1901.08652.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. Science Robotics, 2020. arXiv:2010.11251.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. Science Robotics, 2022. arXiv:2201.08117.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"M. Tranzatto et al.. CERBERUS: Autonomous Legged and Aerial Robotic Exploration in the Tunnel and Urban Circuits of the DARPA Subterranean Challenge. Field Robotics, 2022. arXiv:2201.07067.","identifier":"arXiv:2201.07067","url":"https://arxiv.org/abs/2201.07067"},{"text":"A. Bouman et al.. Autonomous Spot: Long-Range Autonomous Exploration of Extreme Environments with Legged Locomotion. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2020. arXiv:2010.09259.","identifier":"arXiv:2010.09259","url":"https://arxiv.org/abs/2010.09259"},{"text":"H. Hirschmüller. Stereo Processing by Semiglobal Matching and Mutual Information. IEEE Transactions on Pattern Analysis and Machine Intelligence, 30(2):328-341, 2008. doi:10.1109/TPAMI.2007.1166.","identifier":"doi:10.1109/TPAMI.2007.1166","url":"https://doi.org/10.1109/TPAMI.2007.1166"},{"text":"J.-R. Chang and Y.-S. Chen. Pyramid Stereo Matching Network. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2018. arXiv:1803.08669.","identifier":"arXiv:1803.08669","url":"https://arxiv.org/abs/1803.08669"},{"text":"L. Lipson, Z. Teed, and J. Deng. RAFT-Stereo: Multilevel Recurrent Field Transforms for Stereo Matching. International Conference on 3D Vision (3DV), 2021. arXiv:2109.07547.","identifier":"arXiv:2109.07547","url":"https://arxiv.org/abs/2109.07547"},{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. Conference on Robot Learning (CoRL), 2021. arXiv:2109.11978.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. Robotics: Science and Systems (RSS), 2021. arXiv:2107.04034.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. Conference on Robot Learning (CoRL), 2022. arXiv:2211.07638.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"R. Yang, M. Zhang, N. Hansen, H. Xu, and X. Wang. Learning Vision-Guided Quadrupedal Locomotion End-to-End with Cross-Modal Transformers. International Conference on Learning Representations (ICLR), 2022. arXiv:2107.03996.","identifier":"arXiv:2107.03996","url":"https://arxiv.org/abs/2107.03996"},{"text":"C. S. Imai et al.. Vision-Guided Quadrupedal Locomotion in the Wild with Multi-Modal Delay Randomization. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2022. arXiv:2109.14549.","identifier":"arXiv:2109.14549","url":"https://arxiv.org/abs/2109.14549"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. Conference on Robot Learning (CoRL), 2023. arXiv:2309.05665.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. arXiv:2309.14341, 2023.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. Science Robotics, 2024. arXiv:2306.14874.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2022. arXiv:2209.12827.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, L. Wellhausen, F. Farshidian, and M. Hutter. Learning a State Representation and Navigation in Cluttered and Dynamic Environments. IEEE Robotics and Automation Letters, 2021. arXiv:2103.04351.","identifier":"arXiv:2103.04351","url":"https://arxiv.org/abs/2103.04351"},{"text":"Z. Fu, A. Kumar, A. Agarwal, H. Qi, J. Malik, and D. Pathak. Coupling Vision and Proprioception for Navigation of Legged Robots. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022. arXiv:2112.02094.","identifier":"arXiv:2112.02094","url":"https://arxiv.org/abs/2112.02094"},{"text":"T. Miki, L. Wellhausen, R. Grandia, F. Jenelten, T. Homberger, and M. Hutter. Elevation Mapping for Locomotion and Navigation using GPU. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2022. arXiv:2204.12876.","identifier":"arXiv:2204.12876","url":"https://arxiv.org/abs/2204.12876"},{"text":"G. Kahn, P. Abbeel, and S. Levine. BADGR: An Autonomous Self-Supervised Learning-Based Navigation System. IEEE Robotics and Automation Letters, 2021. arXiv:2002.05700.","identifier":"arXiv:2002.05700","url":"https://arxiv.org/abs/2002.05700"},{"text":"M. V. Gasparino et al.. WayFAST: Navigation with Predictive Traversability in the Field. IEEE Robotics and Automation Letters, 2022. arXiv:2203.12071.","identifier":"arXiv:2203.12071","url":"https://arxiv.org/abs/2203.12071"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. Robotics: Science and Systems (RSS), 2023. arXiv:2305.08510.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"M. Mattamala et al.. Wild Visual Navigation: Fast Traversability Learning via Pre-Trained Models and Online Self-Supervision. arXiv:2404.07110, 2024.","identifier":"arXiv:2404.07110","url":"https://arxiv.org/abs/2404.07110"},{"text":"A. Loquercio, E. Kaufmann, R. Ranftl, M. Müller, V. Koltun, and D. Scaramuzza. Learning High-Speed Flight in the Wild. Science Robotics, 2021. arXiv:2110.05113.","identifier":"arXiv:2110.05113","url":"https://arxiv.org/abs/2110.05113"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. IEEE International Conference on Robotics and Automation (ICRA), 2023. arXiv:2210.03370.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Fox, W. Burgard, and S. Thrun. The Dynamic Window Approach to Collision Avoidance. IEEE Robotics and Automation Magazine, 4(1):23-33, 1997. doi:10.1109/100.580977.","identifier":"doi:10.1109/100.580977","url":"https://doi.org/10.1109/100.580977"},{"text":"B. Zhou, F. Gao, L. Wang, C. Liu, and S. Shen. Robust and Efficient Quadrotor Trajectory Generation for Fast Autonomous Flight. IEEE Robotics and Automation Letters, 2019. arXiv:1907.01531.","identifier":"arXiv:1907.01531","url":"https://arxiv.org/abs/1907.01531"},{"text":"X. Zhou, Z. Wang, H. Ye, C. Xu, and F. Gao. EGO-Planner: An ESDF-free Gradient-based Local Planner for Quadrotors. IEEE Robotics and Automation Letters, 2021. arXiv:2008.08835.","identifier":"arXiv:2008.08835","url":"https://arxiv.org/abs/2008.08835"},{"text":"X. Xiao, B. Liu, G. Warnell, and P. Stone. Motion Planning and Control for Mobile Robot Navigation Using Machine Learning: a Survey. Autonomous Robots, 2022. arXiv:2011.13112.","identifier":"arXiv:2011.13112","url":"https://arxiv.org/abs/2011.13112"},{"text":"A. Elfes. Using Occupancy Grids for Mobile Robot Perception and Navigation. Computer, 22(6):46-57, 1989. doi:10.1109/2.30720.","identifier":"doi:10.1109/2.30720","url":"https://doi.org/10.1109/2.30720"},{"text":"H. Oleynikova, Z. Taylor, M. Fehr, R. Siegwart, and J. Nieto. Voxblox: Incremental 3D Euclidean Signed Distance Fields for On-Board MAV Planning. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2017. arXiv:1611.03631.","identifier":"arXiv:1611.03631","url":"https://arxiv.org/abs/1611.03631"},{"text":"T. Shan, B. Englot, D. Meyers, W. Wang, C. Ratti, and D. Rus. LIO-SAM: Tightly-coupled Lidar Inertial Odometry via Smoothing and Mapping. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2020. arXiv:2007.00258.","identifier":"arXiv:2007.00258","url":"https://arxiv.org/abs/2007.00258"},{"text":"W. Xu, Y. Cai, D. He, J. Lin, and F. Zhang. FAST-LIO2: Fast Direct LiDAR-Inertial Odometry. IEEE Transactions on Robotics, 2022. arXiv:2107.06829.","identifier":"arXiv:2107.06829","url":"https://arxiv.org/abs/2107.06829"},{"text":"F. Ma and S. Karaman. Sparse-to-Dense: Depth Prediction from Sparse Depth Samples and a Single Image. IEEE International Conference on Robotics and Automation (ICRA), 2018. arXiv:1709.07492.","identifier":"arXiv:1709.07492","url":"https://arxiv.org/abs/1709.07492"},{"text":"J. Uhrig, N. Schneider, L. Schneider, U. Franke, T. Brox, and A. Geiger. Sparsity Invariant CNNs. International Conference on 3D Vision (3DV), 2017. arXiv:1708.06500.","identifier":"arXiv:1708.06500","url":"https://arxiv.org/abs/1708.06500"},{"text":"M. Bijelic et al.. Seeing Through Fog Without Seeing Fog: Deep Multimodal Sensor Fusion in Unseen Adverse Weather. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020. arXiv:1902.08913.","identifier":"arXiv:1902.08913","url":"https://arxiv.org/abs/1902.08913"},{"text":"A. Kendall and Y. Gal. What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision? Advances in Neural Information Processing Systems (NeurIPS), 2017. arXiv:1703.04977.","identifier":"arXiv:1703.04977","url":"https://arxiv.org/abs/1703.04977"},{"text":"V. Makoviychuk et al.. Isaac Gym: High Performance GPU-Based Physics Simulation for Robot Learning. NeurIPS Datasets and Benchmarks Track, 2021. arXiv:2108.10470.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal et al.. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. IEEE Robotics and Automation Letters, 2023. arXiv:2301.04195.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2017. arXiv:1703.06907.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2607.00007","slug":"2607.00007","registered_at":"2026-07-25T09:45:00Z","updated_at":"2026-07-25T17:02:00Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2607.00007","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2607.00007","tex_path":"articles/2607.00007/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-07-25T12:02:00Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2607.00007"},{"kagg_id":"KAGG:2607.00006","slug":"2607.00006","title":"Quantifying the Sim-to-Real Gap of Quadruped Velocity Controllers on Deformable Terrain","authors":[{"name":"Nikos Stavridis","affiliation":"Robot Dynamics Group, ETH Zurich, Zurich, Switzerland"},{"name":"Amelie Laurent","affiliation":"Robot Dynamics Group, ETH Zurich, Zurich, Switzerland"},{"name":"Diego Navarro","affiliation":"Robot Dynamics Group, ETH Zurich, Zurich, Switzerland"}],"affiliation":null,"abstract":"Sim-to-real studies of legged navigation usually attribute transfer failures to perception mismatch, yet on deformable terrain the velocity interface between planner and locomotion controller also changes character: response lag grows, effective gains shift, and saturation appears earlier. This paper quantifies that interface gap for a quadruped crossing terrain boundaries. We track a calibrated interface model online, using planner-conditioned probing in the style of GAUGE at 2 Hz, and couple it to a configuration-driven motion predictor in the style of AgniNav that scores, at 10 Hz, which fraction of the perceived free range is realizable under the current interface state. A sim-to-real pre-alignment step in Isaac Lab with matched terrain assets initializes the tracker for each terrain class. On an ANYmal traversing a course of grass, loose gravel, a sand trough, packed dirt, and grass (30 m per segment, 15 crossings per direction), the approach reduced interface drift from 4.9 to 4.6 percent per km relative to a static grass-calibrated model, raised range realizability from 70.4 to 74.8 percent, raised navigation success from 80.8 to 86.3 percent, and lowered 95th-percentile tracking error from 17.9 to 15.9 cm. Ablations indicate that online drift tracking, the realizability score, and pre-alignment each contribute. Recorded failures, including a 3.2-fold lag increase in sand and a transient realizability drop to 41 percent at the gravel-to-sand boundary, delimit where the diagnostics remain insufficient.","publication_date":"2026-07-22","registered_at":"2026-07-23T16:45:00Z","updated_at":"2026-07-23T19:45:00Z","status":"published","html_path":null,"pdf_path":"articles/2607.00006/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"M. Hutter, C. Gehring, D. Jud, A. Lauber, C. D. Bellicoso, V. Tsounis, J. Hwangbo, K. Bodie, P. Fankhauser, M. Bloesch, R. Diethelm, S. Bachmann, A. Melzer, and M. Hoepflinger. ANYmal - A Highly Mobile and Dynamic Quadrupedal Robot. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2016.","identifier":"doi:10.1109/IROS.2016.7758092","url":"https://doi.org/10.1109/IROS.2016.7758092"},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"L. Wellhausen, A. Dosovitskiy, R. Ranftl, K. Walas, C. Cadena, and M. Hutter. Where Should I Walk? Predicting Terrain Properties from Images via Self-Supervised Learning. IEEE Robotics and Automation Letters, 4(2):1509-1516, 2019.","identifier":"doi:10.1109/LRA.2019.2895390","url":"https://doi.org/10.1109/LRA.2019.2895390"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. arXiv:2305.08510, 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"A. Kadian et al.. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance? arXiv:1912.06321, 2019.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"},{"text":"T. Zang, H. Huang, S. Cheng, and W. Zhang. GAUGE: Planner-Conditioned Active Calibration of Opaque Quadruped Velocity Interfaces. arXiv:2607.02841, 2026.","identifier":"arXiv:2607.02841","url":"https://arxiv.org/abs/2607.02841"},{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"V. Makoviychuk et al.. Isaac Gym: High Performance GPU-Based Physics Simulation for Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal, C. Yu, Q. Yu, J. Liu, N. Rudin, D. Hoeller, J. L. Yuan, R. Singh, Y. Guo, H. Mazhar, A. Mandlekar, B. Babich, G. State, M. Hutter, and A. Garg. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. arXiv:2301.04195, 2023.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"J. Tan, T. Zhang, E. Coumans, A. Iscen, Y. Bai, D. Hafner, S. Bohez, and V. Vanhoucke. Sim-to-Real: Learning Agile Locomotion for Quadruped Robots. arXiv:1804.10332, 2018.","identifier":"arXiv:1804.10332","url":"https://arxiv.org/abs/1804.10332"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. arXiv:1710.06537, 2017.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. arXiv:1703.06907, 2017.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"},{"text":"W. Yu, J. Tan, C. K. Liu, and G. Turk. Preparing for the Unknown: Learning a Universal Policy with Online System Identification. arXiv:1702.02453, 2017.","identifier":"arXiv:1702.02453","url":"https://arxiv.org/abs/1702.02453"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"G. Ji, J. Mun, H. Kim, and J. Hwangbo. Concurrent Training of a Control Policy and a State Estimator for Dynamic and Robust Legged Locomotion. arXiv:2202.05481, 2022.","identifier":"arXiv:2202.05481","url":"https://arxiv.org/abs/2202.05481"},{"text":"A. Nagabandi, I. Clavera, S. Liu, R. S. Fearing, P. Abbeel, S. Levine, and C. Finn. Learning to Adapt in Dynamic, Real-World Environments Through Meta-Reinforcement Learning. arXiv:1803.11347, 2018.","identifier":"arXiv:1803.11347","url":"https://arxiv.org/abs/1803.11347"},{"text":"W. Zhao, J. P. Queralta, and T. Westerlund. Sim-to-Real Transfer in Deep Reinforcement Learning for Robotics: a Survey. arXiv:2009.13303, 2020.","identifier":"arXiv:2009.13303","url":"https://arxiv.org/abs/2009.13303"},{"text":"S. Choi, G. Ji, J. Park, H. Kim, J. Mun, J. H. Lee, and J. Hwangbo. Learning Quadrupedal Locomotion on Deformable Terrain. Science Robotics, 8(74):eade2256, 2023.","identifier":"doi:10.1126/scirobotics.ade2256","url":"https://doi.org/10.1126/scirobotics.ade2256"},{"text":"G. B. Margolis, G. Yang, K. Paigwar, T. Chen, and P. Agrawal. Rapid Locomotion via Reinforcement Learning. arXiv:2205.02824, 2022.","identifier":"arXiv:2205.02824","url":"https://arxiv.org/abs/2205.02824"},{"text":"G. B. Margolis and P. Agrawal. Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior. arXiv:2212.03238, 2022.","identifier":"arXiv:2212.03238","url":"https://arxiv.org/abs/2212.03238"},{"text":"M. V. Gasparino, A. N. Sivakumar, Y. Liu, A. E. B. Velasquez, V. A. H. Higuti, J. Rogers, H. Tran, and G. Chowdhary. WayFAST: Navigation with Predictive Traversability in the Field. arXiv:2203.12071, 2022.","identifier":"arXiv:2203.12071","url":"https://arxiv.org/abs/2203.12071"},{"text":"G. Kahn, P. Abbeel, and S. Levine. BADGR: An Autonomous Self-Supervised Learning-Based Navigation System. arXiv:2002.05700, 2020.","identifier":"arXiv:2002.05700","url":"https://arxiv.org/abs/2002.05700"},{"text":"P. Fankhauser, M. Bloesch, and M. Hutter. Probabilistic Terrain Mapping for Mobile Robots with Uncertain Localization. IEEE Robotics and Automation Letters, 3(4):3019-3026, 2018.","identifier":"doi:10.1109/LRA.2018.2849506","url":"https://doi.org/10.1109/LRA.2018.2849506"},{"text":"T. Miki, L. Wellhausen, R. Grandia, F. Jenelten, T. Homberger, and M. Hutter. Elevation Mapping for Locomotion and Navigation using GPU. arXiv:2204.12876, 2022.","identifier":"arXiv:2204.12876","url":"https://arxiv.org/abs/2204.12876"},{"text":"M. Wermelinger, P. Fankhauser, R. Diethelm, P. Krüsi, R. Siegwart, and M. Hutter. Navigation Planning for Legged Robots in Challenging Terrain. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2016.","identifier":"doi:10.1109/IROS.2016.7759199","url":"https://doi.org/10.1109/IROS.2016.7759199"},{"text":"L. Ljung. System Identification: Theory for the User, 2nd ed. Prentice Hall, 1999.","identifier":"ISBN:0136566952","url":"https://openlibrary.org/isbn/0136566952"},{"text":"K. J. Åström and B. Wittenmark. Adaptive Control, 2nd ed. Addison-Wesley, 1995.","identifier":"ISBN:0201558661","url":"https://openlibrary.org/isbn/0201558661"},{"text":"E. S. Page. Continuous Inspection Schemes. Biometrika, 41(1/2):100-115, 1954.","identifier":"doi:10.1093/biomet/41.1-2.100","url":"https://doi.org/10.1093/biomet/41.1-2.100"},{"text":"X. Xiao, J. Biswas, and P. Stone. Learning Inverse Kinodynamics for Accurate High-Speed Off-Road Navigation on Unstructured Terrain. arXiv:2102.12667, 2021.","identifier":"arXiv:2102.12667","url":"https://arxiv.org/abs/2102.12667"},{"text":"H. Karnan, K. S. Sikand, P. Atreya, S. Rabiee, X. Xiao, G. Warnell, P. Stone, and J. Biswas. VI-IKD: High-Speed Accurate Off-Road Navigation using Learned Visual-Inertial Inverse Kinodynamics. arXiv:2203.15983, 2022.","identifier":"arXiv:2203.15983","url":"https://arxiv.org/abs/2203.15983"},{"text":"D. Fox, W. Burgard, and S. Thrun. The Dynamic Window Approach to Collision Avoidance. IEEE Robotics and Automation Magazine, 4(1):23-33, 1997.","identifier":"doi:10.1109/100.580977","url":"https://doi.org/10.1109/100.580977"},{"text":"G. Williams, N. Wagener, B. Goldfain, P. Drews, J. M. Rehg, B. Boots, and E. A. Theodorou. Information Theoretic MPC for Model-Based Reinforcement Learning. IEEE International Conference on Robotics and Automation (ICRA), 2017.","identifier":"doi:10.1109/ICRA.2017.7989202","url":"https://doi.org/10.1109/ICRA.2017.7989202"},{"text":"D. Hoeller, L. Wellhausen, F. Farshidian, and M. Hutter. Learning a State Representation and Navigation in Cluttered and Dynamic Environments. arXiv:2103.04351, 2021.","identifier":"arXiv:2103.04351","url":"https://arxiv.org/abs/2103.04351"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"J. Truong, M. Rudolph, N. H. Yokoyama, S. Chernova, D. Batra, and A. Rai. Rethinking Sim2Real: Lower Fidelity Simulation Leads to Higher Sim2Real Transfer in Navigation. Proceedings of Machine Learning Research, 205:859-870, 2023.","identifier":"PMLR:v205/truong23a","url":"https://proceedings.mlr.press/v205/truong23a.html"},{"text":"M. Bloesch, M. Hutter, M. A. Hoepflinger, S. Leutenegger, C. Gehring, C. D. Remy, and R. Siegwart. State Estimation for Legged Robots: Consistent Fusion of Leg Kinematics and IMU. Robotics: Science and Systems VIII, 2012.","identifier":"doi:10.15607/RSS.2012.VIII.003","url":"https://doi.org/10.15607/RSS.2012.VIII.003"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2607.00006","metadata":{"title":"Quantifying the Sim-to-Real Gap of Quadruped Velocity Controllers on Deformable Terrain","authors":[{"name":"Nikos Stavridis","affiliation":"Robot Dynamics Group, ETH Zurich, Zurich, Switzerland"},{"name":"Amelie Laurent","affiliation":"Robot Dynamics Group, ETH Zurich, Zurich, Switzerland"},{"name":"Diego Navarro","affiliation":"Robot Dynamics Group, ETH Zurich, Zurich, Switzerland"}],"publication_date":"2026-07-22","abstract":"Sim-to-real studies of legged navigation usually attribute transfer failures to perception mismatch, yet on deformable terrain the velocity interface between planner and locomotion controller also changes character: response lag grows, effective gains shift, and saturation appears earlier. This paper quantifies that interface gap for a quadruped crossing terrain boundaries. We track a calibrated interface model online, using planner-conditioned probing in the style of GAUGE at 2 Hz, and couple it to a configuration-driven motion predictor in the style of AgniNav that scores, at 10 Hz, which fraction of the perceived free range is realizable under the current interface state. A sim-to-real pre-alignment step in Isaac Lab with matched terrain assets initializes the tracker for each terrain class. On an ANYmal traversing a course of grass, loose gravel, a sand trough, packed dirt, and grass (30 m per segment, 15 crossings per direction), the approach reduced interface drift from 4.9 to 4.6 percent per km relative to a static grass-calibrated model, raised range realizability from 70.4 to 74.8 percent, raised navigation success from 80.8 to 86.3 percent, and lowered 95th-percentile tracking error from 17.9 to 15.9 cm. Ablations indicate that online drift tracking, the realizability score, and pre-alignment each contribute. Recorded failures, including a 3.2-fold lag increase in sand and a transient realizability drop to 41 percent at the gravel-to-sand boundary, delimit where the diagnostics remain insufficient.","references":[{"text":"M. Hutter, C. Gehring, D. Jud, A. Lauber, C. D. Bellicoso, V. Tsounis, J. Hwangbo, K. Bodie, P. Fankhauser, M. Bloesch, R. Diethelm, S. Bachmann, A. Melzer, and M. Hoepflinger. ANYmal - A Highly Mobile and Dynamic Quadrupedal Robot. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2016.","identifier":"doi:10.1109/IROS.2016.7758092","url":"https://doi.org/10.1109/IROS.2016.7758092"},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"L. Wellhausen, A. Dosovitskiy, R. Ranftl, K. Walas, C. Cadena, and M. Hutter. Where Should I Walk? Predicting Terrain Properties from Images via Self-Supervised Learning. IEEE Robotics and Automation Letters, 4(2):1509-1516, 2019.","identifier":"doi:10.1109/LRA.2019.2895390","url":"https://doi.org/10.1109/LRA.2019.2895390"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. arXiv:2305.08510, 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"A. Kadian et al.. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance? arXiv:1912.06321, 2019.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"},{"text":"T. Zang, H. Huang, S. Cheng, and W. Zhang. GAUGE: Planner-Conditioned Active Calibration of Opaque Quadruped Velocity Interfaces. arXiv:2607.02841, 2026.","identifier":"arXiv:2607.02841","url":"https://arxiv.org/abs/2607.02841"},{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"V. Makoviychuk et al.. Isaac Gym: High Performance GPU-Based Physics Simulation for Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal, C. Yu, Q. Yu, J. Liu, N. Rudin, D. Hoeller, J. L. Yuan, R. Singh, Y. Guo, H. Mazhar, A. Mandlekar, B. Babich, G. State, M. Hutter, and A. Garg. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. arXiv:2301.04195, 2023.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"J. Tan, T. Zhang, E. Coumans, A. Iscen, Y. Bai, D. Hafner, S. Bohez, and V. Vanhoucke. Sim-to-Real: Learning Agile Locomotion for Quadruped Robots. arXiv:1804.10332, 2018.","identifier":"arXiv:1804.10332","url":"https://arxiv.org/abs/1804.10332"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. arXiv:1710.06537, 2017.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. arXiv:1703.06907, 2017.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"},{"text":"W. Yu, J. Tan, C. K. Liu, and G. Turk. Preparing for the Unknown: Learning a Universal Policy with Online System Identification. arXiv:1702.02453, 2017.","identifier":"arXiv:1702.02453","url":"https://arxiv.org/abs/1702.02453"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"G. Ji, J. Mun, H. Kim, and J. Hwangbo. Concurrent Training of a Control Policy and a State Estimator for Dynamic and Robust Legged Locomotion. arXiv:2202.05481, 2022.","identifier":"arXiv:2202.05481","url":"https://arxiv.org/abs/2202.05481"},{"text":"A. Nagabandi, I. Clavera, S. Liu, R. S. Fearing, P. Abbeel, S. Levine, and C. Finn. Learning to Adapt in Dynamic, Real-World Environments Through Meta-Reinforcement Learning. arXiv:1803.11347, 2018.","identifier":"arXiv:1803.11347","url":"https://arxiv.org/abs/1803.11347"},{"text":"W. Zhao, J. P. Queralta, and T. Westerlund. Sim-to-Real Transfer in Deep Reinforcement Learning for Robotics: a Survey. arXiv:2009.13303, 2020.","identifier":"arXiv:2009.13303","url":"https://arxiv.org/abs/2009.13303"},{"text":"S. Choi, G. Ji, J. Park, H. Kim, J. Mun, J. H. Lee, and J. Hwangbo. Learning Quadrupedal Locomotion on Deformable Terrain. Science Robotics, 8(74):eade2256, 2023.","identifier":"doi:10.1126/scirobotics.ade2256","url":"https://doi.org/10.1126/scirobotics.ade2256"},{"text":"G. B. Margolis, G. Yang, K. Paigwar, T. Chen, and P. Agrawal. Rapid Locomotion via Reinforcement Learning. arXiv:2205.02824, 2022.","identifier":"arXiv:2205.02824","url":"https://arxiv.org/abs/2205.02824"},{"text":"G. B. Margolis and P. Agrawal. Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior. arXiv:2212.03238, 2022.","identifier":"arXiv:2212.03238","url":"https://arxiv.org/abs/2212.03238"},{"text":"M. V. Gasparino, A. N. Sivakumar, Y. Liu, A. E. B. Velasquez, V. A. H. Higuti, J. Rogers, H. Tran, and G. Chowdhary. WayFAST: Navigation with Predictive Traversability in the Field. arXiv:2203.12071, 2022.","identifier":"arXiv:2203.12071","url":"https://arxiv.org/abs/2203.12071"},{"text":"G. Kahn, P. Abbeel, and S. Levine. BADGR: An Autonomous Self-Supervised Learning-Based Navigation System. arXiv:2002.05700, 2020.","identifier":"arXiv:2002.05700","url":"https://arxiv.org/abs/2002.05700"},{"text":"P. Fankhauser, M. Bloesch, and M. Hutter. Probabilistic Terrain Mapping for Mobile Robots with Uncertain Localization. IEEE Robotics and Automation Letters, 3(4):3019-3026, 2018.","identifier":"doi:10.1109/LRA.2018.2849506","url":"https://doi.org/10.1109/LRA.2018.2849506"},{"text":"T. Miki, L. Wellhausen, R. Grandia, F. Jenelten, T. Homberger, and M. Hutter. Elevation Mapping for Locomotion and Navigation using GPU. arXiv:2204.12876, 2022.","identifier":"arXiv:2204.12876","url":"https://arxiv.org/abs/2204.12876"},{"text":"M. Wermelinger, P. Fankhauser, R. Diethelm, P. Krüsi, R. Siegwart, and M. Hutter. Navigation Planning for Legged Robots in Challenging Terrain. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2016.","identifier":"doi:10.1109/IROS.2016.7759199","url":"https://doi.org/10.1109/IROS.2016.7759199"},{"text":"L. Ljung. System Identification: Theory for the User, 2nd ed. Prentice Hall, 1999.","identifier":"ISBN:0136566952","url":"https://openlibrary.org/isbn/0136566952"},{"text":"K. J. Åström and B. Wittenmark. Adaptive Control, 2nd ed. Addison-Wesley, 1995.","identifier":"ISBN:0201558661","url":"https://openlibrary.org/isbn/0201558661"},{"text":"E. S. Page. Continuous Inspection Schemes. Biometrika, 41(1/2):100-115, 1954.","identifier":"doi:10.1093/biomet/41.1-2.100","url":"https://doi.org/10.1093/biomet/41.1-2.100"},{"text":"X. Xiao, J. Biswas, and P. Stone. Learning Inverse Kinodynamics for Accurate High-Speed Off-Road Navigation on Unstructured Terrain. arXiv:2102.12667, 2021.","identifier":"arXiv:2102.12667","url":"https://arxiv.org/abs/2102.12667"},{"text":"H. Karnan, K. S. Sikand, P. Atreya, S. Rabiee, X. Xiao, G. Warnell, P. Stone, and J. Biswas. VI-IKD: High-Speed Accurate Off-Road Navigation using Learned Visual-Inertial Inverse Kinodynamics. arXiv:2203.15983, 2022.","identifier":"arXiv:2203.15983","url":"https://arxiv.org/abs/2203.15983"},{"text":"D. Fox, W. Burgard, and S. Thrun. The Dynamic Window Approach to Collision Avoidance. IEEE Robotics and Automation Magazine, 4(1):23-33, 1997.","identifier":"doi:10.1109/100.580977","url":"https://doi.org/10.1109/100.580977"},{"text":"G. Williams, N. Wagener, B. Goldfain, P. Drews, J. M. Rehg, B. Boots, and E. A. Theodorou. Information Theoretic MPC for Model-Based Reinforcement Learning. IEEE International Conference on Robotics and Automation (ICRA), 2017.","identifier":"doi:10.1109/ICRA.2017.7989202","url":"https://doi.org/10.1109/ICRA.2017.7989202"},{"text":"D. Hoeller, L. Wellhausen, F. Farshidian, and M. Hutter. Learning a State Representation and Navigation in Cluttered and Dynamic Environments. arXiv:2103.04351, 2021.","identifier":"arXiv:2103.04351","url":"https://arxiv.org/abs/2103.04351"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"J. Truong, M. Rudolph, N. H. Yokoyama, S. Chernova, D. Batra, and A. Rai. Rethinking Sim2Real: Lower Fidelity Simulation Leads to Higher Sim2Real Transfer in Navigation. Proceedings of Machine Learning Research, 205:859-870, 2023.","identifier":"PMLR:v205/truong23a","url":"https://proceedings.mlr.press/v205/truong23a.html"},{"text":"M. Bloesch, M. Hutter, M. A. Hoepflinger, S. Leutenegger, C. Gehring, C. D. Remy, and R. Siegwart. State Estimation for Legged Robots: Consistent Fusion of Leg Kinematics and IMU. Robotics: Science and Systems VIII, 2012.","identifier":"doi:10.15607/RSS.2012.VIII.003","url":"https://doi.org/10.15607/RSS.2012.VIII.003"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2607.00006","slug":"2607.00006","registered_at":"2026-07-23T16:45:00Z","updated_at":"2026-07-23T19:45:00Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2607.00006","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2607.00006","tex_path":"articles/2607.00006/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-07-23T19:02:00Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2607.00006"},{"kagg_id":"KAGG:2607.00005","slug":"2607.00005","title":"Cross-Embodiment Distillation of Local Navigation Planners from Humanoid to Quadruped Teams","authors":[{"name":"Yifan Chen","affiliation":"Robotics and Autonomous Systems Lab, Shanghai Jiao Tong University, Shanghai, China"},{"name":"Zhiyuan Wang","affiliation":"Robotics and Autonomous Systems Lab, Shanghai Jiao Tong University, Shanghai, China"},{"name":"Haoran Liu","affiliation":"Robotics and Autonomous Systems Lab, Shanghai Jiao Tong University, Shanghai, China"},{"name":"Xiaotong Zhang","affiliation":"Robotics and Autonomous Systems Lab, Shanghai Jiao Tong University, Shanghai, China"}],"affiliation":null,"abstract":"Configuration-driven local planners such as AgniNav can serve several robot embodiments from a single planning stack, but the full planner is too expensive for the onboard computers of commodity quadrupeds. This preprint presents a cross-embodiment distillation pipeline in which a configuration-driven teacher planner generates local navigation trajectories in a shared, body-normalized configuration space, and a compact onboard student policy is distilled from these trajectories with an explicit embodiment descriptor as input. Training combines on-policy data aggregation, a finite-difference consistency regularizer that aligns the sensitivity of the student to descriptor changes with that of the teacher, and a held-out embodiment split used for model selection. The pipeline is evaluated in Isaac Lab with a Unitree Go2 and a 12-DoF H1-class humanoid over 20 indoor-outdoor transition scenes, 5 seeds, and 50 episodes per scene and seed, and onboard on a Go2 whose Jetson Orin computer runs the student. The distilled student reaches 90.4 percent navigation success, compared with 81.5 percent for a per-embodiment oracle teacher and 82.3 percent for a student without embodiment conditioning, with a teacher parity score of 93.1 percent and an onboard decision latency of 8.2 ms. Its mean path clearance margin of 18.7 cm is, however, the smallest among the compared methods. Ablations show drops of 11.3 to 15.3 percentage points when the descriptor input, the consistency regularizer, or the held-out split is removed. Recorded failures include reduced clearance on a steep entry ramp and degraded success under descriptor dropout.","publication_date":"2026-07-21","registered_at":"2026-07-25T10:45:00Z","updated_at":"2026-07-25T14:44:00Z","status":"published","html_path":null,"pdf_path":"articles/2607.00005/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"J. Zhang, Y. Du, X. Guo, S. Sun, X. Liu, Y. Sun, G. Lu, W. Sui, and J. Li. EA-Nav: Learning Safe Visual Navigation Policies with Embodiment Awareness. arXiv:2607.19880, 2026.","identifier":"arXiv:2607.19880","url":"https://arxiv.org/abs/2607.19880"},{"text":"X. Xiao, B. Liu, G. Warnell, and P. Stone. Motion Planning and Control for Mobile Robot Navigation Using Machine Learning: a Survey. arXiv:2011.13112, 2020.","identifier":"arXiv:2011.13112","url":"https://arxiv.org/abs/2011.13112"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Sridhar, D. Shah, C. Glossop, and S. Levine. NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration. arXiv:2310.07896, 2023.","identifier":"arXiv:2310.07896","url":"https://arxiv.org/abs/2310.07896"},{"text":"A. Eftekhar et al.. The One RING: a Robotic Indoor Navigation Generalist. arXiv:2412.14401, 2024.","identifier":"arXiv:2412.14401","url":"https://arxiv.org/abs/2412.14401"},{"text":"J. Zhang et al.. Embodied Navigation Foundation Model. arXiv:2509.12129, 2025.","identifier":"arXiv:2509.12129","url":"https://arxiv.org/abs/2509.12129"},{"text":"J. Yang, C. Glossop, A. Bhorkar, D. Shah, Q. Vuong, C. Finn, D. Sadigh, and S. Levine. Pushing the Limits of Cross-Embodiment Learning for Manipulation and Navigation. arXiv:2402.19432, 2024.","identifier":"arXiv:2402.19432","url":"https://arxiv.org/abs/2402.19432"},{"text":"R. Doshi, H. Walke, O. Mees, S. Dasari, and S. Levine. Scaling Cross-Embodied Learning: One Policy for Manipulation, Navigation, Locomotion and Aviation. arXiv:2408.11812, 2024.","identifier":"arXiv:2408.11812","url":"https://arxiv.org/abs/2408.11812"},{"text":"Open X-Embodiment Collaboration. Open X-Embodiment: Robotic Learning Datasets and RT-X Models. arXiv:2310.08864, 2023.","identifier":"arXiv:2310.08864","url":"https://arxiv.org/abs/2310.08864"},{"text":"T. Chen, A. Murali, and A. Gupta. Hardware Conditioned Policies for Multi-Robot Transfer Learning. arXiv:1811.09864, 2018.","identifier":"arXiv:1811.09864","url":"https://arxiv.org/abs/1811.09864"},{"text":"W. Huang, I. Mordatch, and D. Pathak. One Policy to Control Them All: Shared Modular Policies for Agent-Agnostic Control. arXiv:2007.04976, 2020.","identifier":"arXiv:2007.04976","url":"https://arxiv.org/abs/2007.04976"},{"text":"A. Gupta, L. Fan, S. Ganguli, and L. Fei-Fei. MetaMorph: Learning Universal Controllers with Transformers. arXiv:2203.11931, 2022.","identifier":"arXiv:2203.11931","url":"https://arxiv.org/abs/2203.11931"},{"text":"G. Feng et al.. GenLoco: Generalized Locomotion Controllers for Quadrupedal Robots. arXiv:2209.05309, 2022.","identifier":"arXiv:2209.05309","url":"https://arxiv.org/abs/2209.05309"},{"text":"M. Shafiee, G. Bellegarda, and A. Ijspeert. ManyQuadrupeds: Learning a Single Locomotion Policy for Diverse Quadruped Robots. arXiv:2310.10486, 2023.","identifier":"arXiv:2310.10486","url":"https://arxiv.org/abs/2310.10486"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"G. Hinton, O. Vinyals, and J. Dean. Distilling the Knowledge in a Neural Network. arXiv:1503.02531, 2015.","identifier":"arXiv:1503.02531","url":"https://arxiv.org/abs/1503.02531"},{"text":"A. A. Rusu et al.. Policy Distillation. arXiv:1511.06295, 2015.","identifier":"arXiv:1511.06295","url":"https://arxiv.org/abs/1511.06295"},{"text":"S. Ross, G. J. Gordon, and J. A. Bagnell. A Reduction of Imitation Learning and Structured Prediction to No-Regret Online Learning. arXiv:1011.0686, 2010.","identifier":"arXiv:1011.0686","url":"https://arxiv.org/abs/1011.0686"},{"text":"Y. Pan, C.-A. Cheng, K. Saigol, K. Lee, X. Yan, E. Theodorou, and B. Boots. Agile Autonomous Driving using End-to-End Deep Imitation Learning. arXiv:1709.07174, 2017.","identifier":"arXiv:1709.07174","url":"https://arxiv.org/abs/1709.07174"},{"text":"D. Chen, B. Zhou, V. Koltun, and P. Krähenbühl. Learning by Cheating. arXiv:1912.12294, 2019.","identifier":"arXiv:1912.12294","url":"https://arxiv.org/abs/1912.12294"},{"text":"E. Kaufmann, A. Loquercio, R. Ranftl, M. Müller, V. Koltun, and D. Scaramuzza. Deep Drone Acrobatics. arXiv:2006.05768, 2020.","identifier":"arXiv:2006.05768","url":"https://arxiv.org/abs/2006.05768"},{"text":"A. Loquercio, E. Kaufmann, R. Ranftl, M. Müller, V. Koltun, and D. Scaramuzza. Learning High-Speed Flight in the Wild. arXiv:2110.05113, 2021.","identifier":"arXiv:2110.05113","url":"https://arxiv.org/abs/2110.05113"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"W. M. Czarnecki, S. Osindero, M. Jaderberg, G. Swirszcz, and R. Pascanu. Sobolev Training for Neural Networks. arXiv:1706.04859, 2017.","identifier":"arXiv:1706.04859","url":"https://arxiv.org/abs/1706.04859"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, L. Wellhausen, F. Farshidian, and M. Hutter. Learning a State Representation and Navigation in Cluttered and Dynamic Environments. arXiv:2103.04351, 2021.","identifier":"arXiv:2103.04351","url":"https://arxiv.org/abs/2103.04351"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. arXiv:2309.14341, 2023.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"Z. Zhuang et al.. Robot Parkour Learning. arXiv:2309.05665, 2023.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"F. Jenelten, J. He, F. Farshidian, and M. Hutter. DTC: Deep Tracking Control. arXiv:2309.15462, 2023.","identifier":"arXiv:2309.15462","url":"https://arxiv.org/abs/2309.15462"},{"text":"I. Radosavovic, T. Xiao, B. Zhang, T. Darrell, J. Malik, and K. Sreenath. Real-World Humanoid Locomotion with Reinforcement Learning. arXiv:2303.03381, 2023.","identifier":"arXiv:2303.03381","url":"https://arxiv.org/abs/2303.03381"},{"text":"X. Gu, Y.-J. Wang, and J. Chen. Humanoid-Gym: Reinforcement Learning for Humanoid Robot with Zero-Shot Sim2Real Transfer. arXiv:2404.05695, 2024.","identifier":"arXiv:2404.05695","url":"https://arxiv.org/abs/2404.05695"},{"text":"V. Makoviychuk et al.. Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal et al.. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. arXiv:2301.04195, 2023.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. arXiv:1703.06907, 2017.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. arXiv:1710.06537, 2017.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"},{"text":"A. Kadian et al.. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance? arXiv:1912.06321, 2019.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"},{"text":"G. Williams, A. Aldrich, and E. Theodorou. Model Predictive Path Integral Control using Covariance Variable Importance Sampling. arXiv:1509.01149, 2015.","identifier":"arXiv:1509.01149","url":"https://arxiv.org/abs/1509.01149"},{"text":"A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada. Control Barrier Functions: Theory and Applications. arXiv:1903.11199, 2019.","identifier":"arXiv:1903.11199","url":"https://arxiv.org/abs/1903.11199"},{"text":"E. Perez, F. Strub, H. de Vries, V. Dumoulin, and A. Courville. FiLM: Visual Reasoning with a General Conditioning Layer. arXiv:1709.07871, 2017.","identifier":"arXiv:1709.07871","url":"https://arxiv.org/abs/1709.07871"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2607.00005","metadata":{"title":"Cross-Embodiment Distillation of Local Navigation Planners from Humanoid to Quadruped Teams","authors":[{"name":"Yifan Chen","affiliation":"Robotics and Autonomous Systems Lab, Shanghai Jiao Tong University, Shanghai, China"},{"name":"Zhiyuan Wang","affiliation":"Robotics and Autonomous Systems Lab, Shanghai Jiao Tong University, Shanghai, China"},{"name":"Haoran Liu","affiliation":"Robotics and Autonomous Systems Lab, Shanghai Jiao Tong University, Shanghai, China"},{"name":"Xiaotong Zhang","affiliation":"Robotics and Autonomous Systems Lab, Shanghai Jiao Tong University, Shanghai, China"}],"publication_date":"2026-07-21","abstract":"Configuration-driven local planners such as AgniNav can serve several robot embodiments from a single planning stack, but the full planner is too expensive for the onboard computers of commodity quadrupeds. This preprint presents a cross-embodiment distillation pipeline in which a configuration-driven teacher planner generates local navigation trajectories in a shared, body-normalized configuration space, and a compact onboard student policy is distilled from these trajectories with an explicit embodiment descriptor as input. Training combines on-policy data aggregation, a finite-difference consistency regularizer that aligns the sensitivity of the student to descriptor changes with that of the teacher, and a held-out embodiment split used for model selection. The pipeline is evaluated in Isaac Lab with a Unitree Go2 and a 12-DoF H1-class humanoid over 20 indoor-outdoor transition scenes, 5 seeds, and 50 episodes per scene and seed, and onboard on a Go2 whose Jetson Orin computer runs the student. The distilled student reaches 90.4 percent navigation success, compared with 81.5 percent for a per-embodiment oracle teacher and 82.3 percent for a student without embodiment conditioning, with a teacher parity score of 93.1 percent and an onboard decision latency of 8.2 ms. Its mean path clearance margin of 18.7 cm is, however, the smallest among the compared methods. Ablations show drops of 11.3 to 15.3 percentage points when the descriptor input, the consistency regularizer, or the held-out split is removed. Recorded failures include reduced clearance on a steep entry ramp and degraded success under descriptor dropout.","references":[{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"J. Zhang, Y. Du, X. Guo, S. Sun, X. Liu, Y. Sun, G. Lu, W. Sui, and J. Li. EA-Nav: Learning Safe Visual Navigation Policies with Embodiment Awareness. arXiv:2607.19880, 2026.","identifier":"arXiv:2607.19880","url":"https://arxiv.org/abs/2607.19880"},{"text":"X. Xiao, B. Liu, G. Warnell, and P. Stone. Motion Planning and Control for Mobile Robot Navigation Using Machine Learning: a Survey. arXiv:2011.13112, 2020.","identifier":"arXiv:2011.13112","url":"https://arxiv.org/abs/2011.13112"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Sridhar, D. Shah, C. Glossop, and S. Levine. NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration. arXiv:2310.07896, 2023.","identifier":"arXiv:2310.07896","url":"https://arxiv.org/abs/2310.07896"},{"text":"A. Eftekhar et al.. The One RING: a Robotic Indoor Navigation Generalist. arXiv:2412.14401, 2024.","identifier":"arXiv:2412.14401","url":"https://arxiv.org/abs/2412.14401"},{"text":"J. Zhang et al.. Embodied Navigation Foundation Model. arXiv:2509.12129, 2025.","identifier":"arXiv:2509.12129","url":"https://arxiv.org/abs/2509.12129"},{"text":"J. Yang, C. Glossop, A. Bhorkar, D. Shah, Q. Vuong, C. Finn, D. Sadigh, and S. Levine. Pushing the Limits of Cross-Embodiment Learning for Manipulation and Navigation. arXiv:2402.19432, 2024.","identifier":"arXiv:2402.19432","url":"https://arxiv.org/abs/2402.19432"},{"text":"R. Doshi, H. Walke, O. Mees, S. Dasari, and S. Levine. Scaling Cross-Embodied Learning: One Policy for Manipulation, Navigation, Locomotion and Aviation. arXiv:2408.11812, 2024.","identifier":"arXiv:2408.11812","url":"https://arxiv.org/abs/2408.11812"},{"text":"Open X-Embodiment Collaboration. Open X-Embodiment: Robotic Learning Datasets and RT-X Models. arXiv:2310.08864, 2023.","identifier":"arXiv:2310.08864","url":"https://arxiv.org/abs/2310.08864"},{"text":"T. Chen, A. Murali, and A. Gupta. Hardware Conditioned Policies for Multi-Robot Transfer Learning. arXiv:1811.09864, 2018.","identifier":"arXiv:1811.09864","url":"https://arxiv.org/abs/1811.09864"},{"text":"W. Huang, I. Mordatch, and D. Pathak. One Policy to Control Them All: Shared Modular Policies for Agent-Agnostic Control. arXiv:2007.04976, 2020.","identifier":"arXiv:2007.04976","url":"https://arxiv.org/abs/2007.04976"},{"text":"A. Gupta, L. Fan, S. Ganguli, and L. Fei-Fei. MetaMorph: Learning Universal Controllers with Transformers. arXiv:2203.11931, 2022.","identifier":"arXiv:2203.11931","url":"https://arxiv.org/abs/2203.11931"},{"text":"G. Feng et al.. GenLoco: Generalized Locomotion Controllers for Quadrupedal Robots. arXiv:2209.05309, 2022.","identifier":"arXiv:2209.05309","url":"https://arxiv.org/abs/2209.05309"},{"text":"M. Shafiee, G. Bellegarda, and A. Ijspeert. ManyQuadrupeds: Learning a Single Locomotion Policy for Diverse Quadruped Robots. arXiv:2310.10486, 2023.","identifier":"arXiv:2310.10486","url":"https://arxiv.org/abs/2310.10486"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"G. Hinton, O. Vinyals, and J. Dean. Distilling the Knowledge in a Neural Network. arXiv:1503.02531, 2015.","identifier":"arXiv:1503.02531","url":"https://arxiv.org/abs/1503.02531"},{"text":"A. A. Rusu et al.. Policy Distillation. arXiv:1511.06295, 2015.","identifier":"arXiv:1511.06295","url":"https://arxiv.org/abs/1511.06295"},{"text":"S. Ross, G. J. Gordon, and J. A. Bagnell. A Reduction of Imitation Learning and Structured Prediction to No-Regret Online Learning. arXiv:1011.0686, 2010.","identifier":"arXiv:1011.0686","url":"https://arxiv.org/abs/1011.0686"},{"text":"Y. Pan, C.-A. Cheng, K. Saigol, K. Lee, X. Yan, E. Theodorou, and B. Boots. Agile Autonomous Driving using End-to-End Deep Imitation Learning. arXiv:1709.07174, 2017.","identifier":"arXiv:1709.07174","url":"https://arxiv.org/abs/1709.07174"},{"text":"D. Chen, B. Zhou, V. Koltun, and P. Krähenbühl. Learning by Cheating. arXiv:1912.12294, 2019.","identifier":"arXiv:1912.12294","url":"https://arxiv.org/abs/1912.12294"},{"text":"E. Kaufmann, A. Loquercio, R. Ranftl, M. Müller, V. Koltun, and D. Scaramuzza. Deep Drone Acrobatics. arXiv:2006.05768, 2020.","identifier":"arXiv:2006.05768","url":"https://arxiv.org/abs/2006.05768"},{"text":"A. Loquercio, E. Kaufmann, R. Ranftl, M. Müller, V. Koltun, and D. Scaramuzza. Learning High-Speed Flight in the Wild. arXiv:2110.05113, 2021.","identifier":"arXiv:2110.05113","url":"https://arxiv.org/abs/2110.05113"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"W. M. Czarnecki, S. Osindero, M. Jaderberg, G. Swirszcz, and R. Pascanu. Sobolev Training for Neural Networks. arXiv:1706.04859, 2017.","identifier":"arXiv:1706.04859","url":"https://arxiv.org/abs/1706.04859"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, L. Wellhausen, F. Farshidian, and M. Hutter. Learning a State Representation and Navigation in Cluttered and Dynamic Environments. arXiv:2103.04351, 2021.","identifier":"arXiv:2103.04351","url":"https://arxiv.org/abs/2103.04351"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. arXiv:2309.14341, 2023.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"Z. Zhuang et al.. Robot Parkour Learning. arXiv:2309.05665, 2023.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"F. Jenelten, J. He, F. Farshidian, and M. Hutter. DTC: Deep Tracking Control. arXiv:2309.15462, 2023.","identifier":"arXiv:2309.15462","url":"https://arxiv.org/abs/2309.15462"},{"text":"I. Radosavovic, T. Xiao, B. Zhang, T. Darrell, J. Malik, and K. Sreenath. Real-World Humanoid Locomotion with Reinforcement Learning. arXiv:2303.03381, 2023.","identifier":"arXiv:2303.03381","url":"https://arxiv.org/abs/2303.03381"},{"text":"X. Gu, Y.-J. Wang, and J. Chen. Humanoid-Gym: Reinforcement Learning for Humanoid Robot with Zero-Shot Sim2Real Transfer. arXiv:2404.05695, 2024.","identifier":"arXiv:2404.05695","url":"https://arxiv.org/abs/2404.05695"},{"text":"V. Makoviychuk et al.. Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning. arXiv:2108.10470, 2021.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal et al.. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. arXiv:2301.04195, 2023.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. arXiv:1703.06907, 2017.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. arXiv:1710.06537, 2017.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"},{"text":"A. Kadian et al.. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance? arXiv:1912.06321, 2019.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"},{"text":"G. Williams, A. Aldrich, and E. Theodorou. Model Predictive Path Integral Control using Covariance Variable Importance Sampling. arXiv:1509.01149, 2015.","identifier":"arXiv:1509.01149","url":"https://arxiv.org/abs/1509.01149"},{"text":"A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada. Control Barrier Functions: Theory and Applications. arXiv:1903.11199, 2019.","identifier":"arXiv:1903.11199","url":"https://arxiv.org/abs/1903.11199"},{"text":"E. Perez, F. Strub, H. de Vries, V. Dumoulin, and A. Courville. FiLM: Visual Reasoning with a General Conditioning Layer. arXiv:1709.07871, 2017.","identifier":"arXiv:1709.07871","url":"https://arxiv.org/abs/1709.07871"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2607.00005","slug":"2607.00005","registered_at":"2026-07-25T10:45:00Z","updated_at":"2026-07-25T14:44:00Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2607.00005","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2607.00005","tex_path":"articles/2607.00005/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-07-25T13:02:00Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2607.00005"},{"kagg_id":"KAGG:2607.00004","slug":"2607.00004","title":"A Benchmark Protocol for Evaluating Opaque Velocity Interfaces in Legged Robots","authors":[{"name":"Grace Lin","affiliation":"Autonomous Robotics Center, Carnegie Mellon University, Pittsburgh, PA, USA"},{"name":"Wei Xu","affiliation":"Autonomous Robotics Center, Carnegie Mellon University, Pittsburgh, PA, USA"},{"name":"Arjun Mehta","affiliation":"Autonomous Robotics Center, Carnegie Mellon University, Pittsburgh, PA, USA"},{"name":"Ethan Cole","affiliation":"Autonomous Robotics Center, Carnegie Mellon University, Pittsburgh, PA, USA"}],"affiliation":null,"abstract":"Velocity commands are the de facto interface between navigation software and the locomotion controllers of legged robots, yet these controllers behave as opaque, stateful, and platform-specific systems whose command-to-motion response is rarely documented and never characterized in a common way. This paper proposes a benchmark protocol for evaluating such opaque velocity interfaces. The protocol consists of a fixed suite of 42 probe command sequences drawn from five families (step, sweep, spiral, saturation, and disturbance rejection), five standardized metrics (command fidelity, phase lag, steady-state bias, saturation envelope, and recovery time), and a reporting template that records firmware, gait mode, payload, and terrain alongside the results. GAUGE-style active calibration serves as the reference characterization method. The protocol is demonstrated on two ANYmal B units with identical firmware and different service histories and on one Unitree Go2, on a lab floor and a stair course, with three repetitions per sequence and platform. The characterization produced by the full suite attains a command fidelity error of 8.4 percent RMS and a 95th-percentile phase lag of 62 ms, compared with 9.0 percent and 66 ms for the reference-unit characterization. A probe-budget sweep shows that fidelity error falls from 17.8 to 8.4 percent as the suite grows from 6 to 42 sequences, with diminishing returns beyond 27. The recorded anomalies, including an unmodeled 0.4 s gait-mode transient and payload-dependent early saturation, illustrate behavior that single-unit characterizations miss. The probe suite and exact probe parameters are released with the paper.","publication_date":"2026-07-19","registered_at":"2026-07-21T08:23:00Z","updated_at":"2026-07-23T10:38:00Z","status":"published","html_path":null,"pdf_path":"articles/2607.00004/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"J. Di Carlo, P. M. Wensing, B. Katz, G. Bledt, and S. Kim. Dynamic Locomotion in the MIT Cheetah 3 Through Convex Model-Predictive Control. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2018.","identifier":"","url":""},{"text":"D. Kim, J. Di Carlo, B. Katz, G. Bledt, and S. Kim. Highly Dynamic Quadruped Locomotion via Whole-Body Impulse Control and Model Predictive Control. arXiv:1909.06586, 2019.","identifier":"arXiv:1909.06586","url":"https://arxiv.org/abs/1909.06586"},{"text":"C. D. Bellicoso, F. Jenelten, C. Gehring, and M. Hutter. Dynamic Locomotion Through Online Nonlinear Motion Optimization for Quadrupedal Robots. IEEE Robotics and Automation Letters, 3(3), 2018.","identifier":"","url":""},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"P. Fankhauser, M. Bloesch, and M. Hutter. Probabilistic Terrain Mapping for Mobile Robots with Uncertain Localization. IEEE Robotics and Automation Letters, 3(4), 2018.","identifier":"","url":""},{"text":"L. Wellhausen and M. Hutter. ArtPlanner: Robust Legged Robot Navigation in the Field. arXiv:2303.01420, 2023.","identifier":"arXiv:2303.01420","url":"https://arxiv.org/abs/2303.01420"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"G. Williams, N. Wagener, B. Goldfain, P. Drews, J. M. Rehg, B. Boots, and E. A. Theodorou. Information Theoretic MPC for Model-Based Reinforcement Learning. IEEE International Conference on Robotics and Automation (ICRA), 2017.","identifier":"","url":""},{"text":"T. Zang, H. Huang, S. Cheng, and W. Zhang. GAUGE: Planner-Conditioned Active Calibration of Opaque Quadruped Velocity Interfaces. arXiv:2607.02841, 2026.","identifier":"arXiv:2607.02841","url":"https://arxiv.org/abs/2607.02841"},{"text":"M. Hutter, C. Gehring, D. Jud, A. Lauber, C. D. Bellicoso, V. Tsounis, J. Hwangbo, K. Bodie, P. Fankhauser, M. Bloesch, R. Diethelm, S. Bachmann, A. Melzer, and M. Hoepflinger. ANYmal – A Highly Mobile and Dynamic Quadrupedal Robot. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2016.","identifier":"","url":""},{"text":"J. Tan, T. Zhang, E. Coumans, A. Iscen, Y. Bai, D. Hafner, S. Bohez, and V. Vanhoucke. Sim-to-Real: Learning Agile Locomotion For Quadruped Robots. arXiv:1804.10332, 2018.","identifier":"arXiv:1804.10332","url":"https://arxiv.org/abs/1804.10332"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"G. B. Margolis, G. Yang, K. Paigwar, T. Chen, and P. Agrawal. Rapid Locomotion via Reinforcement Learning. arXiv:2205.02824, 2022.","identifier":"arXiv:2205.02824","url":"https://arxiv.org/abs/2205.02824"},{"text":"L. Smith, I. Kostrikov, and S. Levine. A Walk in the Park: Learning to Walk in 20 Minutes With Model-Free Reinforcement Learning. arXiv:2208.07860, 2022.","identifier":"arXiv:2208.07860","url":"https://arxiv.org/abs/2208.07860"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. arXiv:1710.06537, 2017.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"},{"text":"G. Ji, J. Mun, H. Kim, and J. Hwangbo. Concurrent Training of a Control Policy and a State Estimator for Dynamic and Robust Legged Locomotion. arXiv:2202.05481, 2022.","identifier":"arXiv:2202.05481","url":"https://arxiv.org/abs/2202.05481"},{"text":"I. M. A. Nahrendra, B. Yu, and H. Myung. DreamWaQ: Learning Robust Quadrupedal Locomotion With Implicit Terrain Imagination via Deep Reinforcement Learning. arXiv:2301.10602, 2023.","identifier":"arXiv:2301.10602","url":"https://arxiv.org/abs/2301.10602"},{"text":"G. B. Margolis and P. Agrawal. Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior. arXiv:2212.03238, 2022.","identifier":"arXiv:2212.03238","url":"https://arxiv.org/abs/2212.03238"},{"text":"S. Gangapurwala, M. Geisert, R. Orsolino, M. Fallon, and I. Havoutis. RLOC: Terrain-Aware Legged Locomotion using Reinforcement Learning and Optimal Control. arXiv:2012.03094, 2020.","identifier":"arXiv:2012.03094","url":"https://arxiv.org/abs/2012.03094"},{"text":"Z. Fu, X. Cheng, and D. Pathak. Deep Whole-Body Control: Learning a Unified Policy for Manipulation and Locomotion. arXiv:2210.10044, 2022.","identifier":"arXiv:2210.10044","url":"https://arxiv.org/abs/2210.10044"},{"text":"R. Yang, M. Zhang, N. Hansen, H. Xu, and X. Wang. Learning Vision-Guided Quadrupedal Locomotion End-to-End with Cross-Modal Transformers. arXiv:2107.03996, 2021.","identifier":"arXiv:2107.03996","url":"https://arxiv.org/abs/2107.03996"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. arXiv:2211.07638, 2022.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. arXiv:2309.14341, 2023.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. arXiv:2309.05665, 2023.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"M. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. arXiv:2305.08510, 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"L. Ljung. System Identification: Theory for the User, 2nd ed.. Prentice Hall, 1999.","identifier":"","url":""},{"text":"R. Pintelon and J. Schoukens. System Identification: A Frequency Domain Approach, 2nd ed.. Wiley-IEEE Press, 2012.","identifier":"","url":""},{"text":"M. Bloesch, M. Hutter, M. A. Hoepflinger, S. Leutenegger, C. Gehring, C. D. Remy, and R. Siegwart. State Estimation for Legged Robots: Consistent Fusion of Leg Kinematics and IMU. Robotics: Science and Systems (RSS), 2012.","identifier":"","url":""},{"text":"R. Hartley, M. Ghaffari, R. M. Eustice, and J. W. Grizzle. Contact-Aided Invariant Extended Kalman Filtering for Legged Robot State Estimation. arXiv:1805.10410, 2018.","identifier":"arXiv:1805.10410","url":"https://arxiv.org/abs/1805.10410"},{"text":"D. Wisth, M. Camurri, and M. Fallon. VILENS: Visual, Inertial, Lidar, and Leg Odometry for All-Terrain Legged Robots. arXiv:2107.07243, 2021.","identifier":"arXiv:2107.07243","url":"https://arxiv.org/abs/2107.07243"},{"text":"B. Calli, A. Walsman, A. Singh, S. Srinivasa, P. Abbeel, and A. M. Dollar. Benchmarking in Manipulation Research: The YCB Object and Model Set and Benchmarking Protocols. arXiv:1502.03143, 2015.","identifier":"arXiv:1502.03143","url":"https://arxiv.org/abs/1502.03143"},{"text":"P. Anderson et al.. On Evaluation of Embodied Navigation Agents. arXiv:1807.06757, 2018.","identifier":"arXiv:1807.06757","url":"https://arxiv.org/abs/1807.06757"},{"text":"N. Yokoyama, S. Ha, and D. Batra. Success Weighted by Completion Time: A Dynamics-Aware Evaluation Criteria for Embodied Navigation. arXiv:2103.08022, 2021.","identifier":"arXiv:2103.08022","url":"https://arxiv.org/abs/2103.08022"},{"text":"A. Kadian et al.. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance? arXiv:1912.06321, 2019.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2607.00004","metadata":{"title":"A Benchmark Protocol for Evaluating Opaque Velocity Interfaces in Legged Robots","authors":[{"name":"Grace Lin","affiliation":"Autonomous Robotics Center, Carnegie Mellon University, Pittsburgh, PA, USA"},{"name":"Wei Xu","affiliation":"Autonomous Robotics Center, Carnegie Mellon University, Pittsburgh, PA, USA"},{"name":"Arjun Mehta","affiliation":"Autonomous Robotics Center, Carnegie Mellon University, Pittsburgh, PA, USA"},{"name":"Ethan Cole","affiliation":"Autonomous Robotics Center, Carnegie Mellon University, Pittsburgh, PA, USA"}],"publication_date":"2026-07-19","abstract":"Velocity commands are the de facto interface between navigation software and the locomotion controllers of legged robots, yet these controllers behave as opaque, stateful, and platform-specific systems whose command-to-motion response is rarely documented and never characterized in a common way. This paper proposes a benchmark protocol for evaluating such opaque velocity interfaces. The protocol consists of a fixed suite of 42 probe command sequences drawn from five families (step, sweep, spiral, saturation, and disturbance rejection), five standardized metrics (command fidelity, phase lag, steady-state bias, saturation envelope, and recovery time), and a reporting template that records firmware, gait mode, payload, and terrain alongside the results. GAUGE-style active calibration serves as the reference characterization method. The protocol is demonstrated on two ANYmal B units with identical firmware and different service histories and on one Unitree Go2, on a lab floor and a stair course, with three repetitions per sequence and platform. The characterization produced by the full suite attains a command fidelity error of 8.4 percent RMS and a 95th-percentile phase lag of 62 ms, compared with 9.0 percent and 66 ms for the reference-unit characterization. A probe-budget sweep shows that fidelity error falls from 17.8 to 8.4 percent as the suite grows from 6 to 42 sequences, with diminishing returns beyond 27. The recorded anomalies, including an unmodeled 0.4 s gait-mode transient and payload-dependent early saturation, illustrate behavior that single-unit characterizations miss. The probe suite and exact probe parameters are released with the paper.","references":[{"text":"J. Di Carlo, P. M. Wensing, B. Katz, G. Bledt, and S. Kim. Dynamic Locomotion in the MIT Cheetah 3 Through Convex Model-Predictive Control. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2018.","identifier":"","url":""},{"text":"D. Kim, J. Di Carlo, B. Katz, G. Bledt, and S. Kim. Highly Dynamic Quadruped Locomotion via Whole-Body Impulse Control and Model Predictive Control. arXiv:1909.06586, 2019.","identifier":"arXiv:1909.06586","url":"https://arxiv.org/abs/1909.06586"},{"text":"C. D. Bellicoso, F. Jenelten, C. Gehring, and M. Hutter. Dynamic Locomotion Through Online Nonlinear Motion Optimization for Quadrupedal Robots. IEEE Robotics and Automation Letters, 3(3), 2018.","identifier":"","url":""},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. arXiv:1901.08652, 2019.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. arXiv:2010.11251, 2020.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. arXiv:2201.08117, 2022.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. arXiv:2109.11978, 2021.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"P. Fankhauser, M. Bloesch, and M. Hutter. Probabilistic Terrain Mapping for Mobile Robots with Uncertain Localization. IEEE Robotics and Automation Letters, 3(4), 2018.","identifier":"","url":""},{"text":"L. Wellhausen and M. Hutter. ArtPlanner: Robust Legged Robot Navigation in the Field. arXiv:2303.01420, 2023.","identifier":"arXiv:2303.01420","url":"https://arxiv.org/abs/2303.01420"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"G. Williams, N. Wagener, B. Goldfain, P. Drews, J. M. Rehg, B. Boots, and E. A. Theodorou. Information Theoretic MPC for Model-Based Reinforcement Learning. IEEE International Conference on Robotics and Automation (ICRA), 2017.","identifier":"","url":""},{"text":"T. Zang, H. Huang, S. Cheng, and W. Zhang. GAUGE: Planner-Conditioned Active Calibration of Opaque Quadruped Velocity Interfaces. arXiv:2607.02841, 2026.","identifier":"arXiv:2607.02841","url":"https://arxiv.org/abs/2607.02841"},{"text":"M. Hutter, C. Gehring, D. Jud, A. Lauber, C. D. Bellicoso, V. Tsounis, J. Hwangbo, K. Bodie, P. Fankhauser, M. Bloesch, R. Diethelm, S. Bachmann, A. Melzer, and M. Hoepflinger. ANYmal – A Highly Mobile and Dynamic Quadrupedal Robot. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2016.","identifier":"","url":""},{"text":"J. Tan, T. Zhang, E. Coumans, A. Iscen, Y. Bai, D. Hafner, S. Bohez, and V. Vanhoucke. Sim-to-Real: Learning Agile Locomotion For Quadruped Robots. arXiv:1804.10332, 2018.","identifier":"arXiv:1804.10332","url":"https://arxiv.org/abs/1804.10332"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. arXiv:2107.04034, 2021.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"G. B. Margolis, G. Yang, K. Paigwar, T. Chen, and P. Agrawal. Rapid Locomotion via Reinforcement Learning. arXiv:2205.02824, 2022.","identifier":"arXiv:2205.02824","url":"https://arxiv.org/abs/2205.02824"},{"text":"L. Smith, I. Kostrikov, and S. Levine. A Walk in the Park: Learning to Walk in 20 Minutes With Model-Free Reinforcement Learning. arXiv:2208.07860, 2022.","identifier":"arXiv:2208.07860","url":"https://arxiv.org/abs/2208.07860"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. arXiv:1710.06537, 2017.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"},{"text":"G. Ji, J. Mun, H. Kim, and J. Hwangbo. Concurrent Training of a Control Policy and a State Estimator for Dynamic and Robust Legged Locomotion. arXiv:2202.05481, 2022.","identifier":"arXiv:2202.05481","url":"https://arxiv.org/abs/2202.05481"},{"text":"I. M. A. Nahrendra, B. Yu, and H. Myung. DreamWaQ: Learning Robust Quadrupedal Locomotion With Implicit Terrain Imagination via Deep Reinforcement Learning. arXiv:2301.10602, 2023.","identifier":"arXiv:2301.10602","url":"https://arxiv.org/abs/2301.10602"},{"text":"G. B. Margolis and P. Agrawal. Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior. arXiv:2212.03238, 2022.","identifier":"arXiv:2212.03238","url":"https://arxiv.org/abs/2212.03238"},{"text":"S. Gangapurwala, M. Geisert, R. Orsolino, M. Fallon, and I. Havoutis. RLOC: Terrain-Aware Legged Locomotion using Reinforcement Learning and Optimal Control. arXiv:2012.03094, 2020.","identifier":"arXiv:2012.03094","url":"https://arxiv.org/abs/2012.03094"},{"text":"Z. Fu, X. Cheng, and D. Pathak. Deep Whole-Body Control: Learning a Unified Policy for Manipulation and Locomotion. arXiv:2210.10044, 2022.","identifier":"arXiv:2210.10044","url":"https://arxiv.org/abs/2210.10044"},{"text":"R. Yang, M. Zhang, N. Hansen, H. Xu, and X. Wang. Learning Vision-Guided Quadrupedal Locomotion End-to-End with Cross-Modal Transformers. arXiv:2107.03996, 2021.","identifier":"arXiv:2107.03996","url":"https://arxiv.org/abs/2107.03996"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. arXiv:2211.07638, 2022.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. arXiv:2309.14341, 2023.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. arXiv:2309.05665, 2023.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. arXiv:2306.14874, 2023.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"M. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. arXiv:2305.08510, 2023.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. arXiv:2209.12827, 2022.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"L. Ljung. System Identification: Theory for the User, 2nd ed.. Prentice Hall, 1999.","identifier":"","url":""},{"text":"R. Pintelon and J. Schoukens. System Identification: A Frequency Domain Approach, 2nd ed.. Wiley-IEEE Press, 2012.","identifier":"","url":""},{"text":"M. Bloesch, M. Hutter, M. A. Hoepflinger, S. Leutenegger, C. Gehring, C. D. Remy, and R. Siegwart. State Estimation for Legged Robots: Consistent Fusion of Leg Kinematics and IMU. Robotics: Science and Systems (RSS), 2012.","identifier":"","url":""},{"text":"R. Hartley, M. Ghaffari, R. M. Eustice, and J. W. Grizzle. Contact-Aided Invariant Extended Kalman Filtering for Legged Robot State Estimation. arXiv:1805.10410, 2018.","identifier":"arXiv:1805.10410","url":"https://arxiv.org/abs/1805.10410"},{"text":"D. Wisth, M. Camurri, and M. Fallon. VILENS: Visual, Inertial, Lidar, and Leg Odometry for All-Terrain Legged Robots. arXiv:2107.07243, 2021.","identifier":"arXiv:2107.07243","url":"https://arxiv.org/abs/2107.07243"},{"text":"B. Calli, A. Walsman, A. Singh, S. Srinivasa, P. Abbeel, and A. M. Dollar. Benchmarking in Manipulation Research: The YCB Object and Model Set and Benchmarking Protocols. arXiv:1502.03143, 2015.","identifier":"arXiv:1502.03143","url":"https://arxiv.org/abs/1502.03143"},{"text":"P. Anderson et al.. On Evaluation of Embodied Navigation Agents. arXiv:1807.06757, 2018.","identifier":"arXiv:1807.06757","url":"https://arxiv.org/abs/1807.06757"},{"text":"N. Yokoyama, S. Ha, and D. Batra. Success Weighted by Completion Time: A Dynamics-Aware Evaluation Criteria for Embodied Navigation. arXiv:2103.08022, 2021.","identifier":"arXiv:2103.08022","url":"https://arxiv.org/abs/2103.08022"},{"text":"A. Kadian et al.. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance? arXiv:1912.06321, 2019.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2607.00004","slug":"2607.00004","registered_at":"2026-07-21T08:23:00Z","updated_at":"2026-07-23T10:38:00Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2607.00004","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2607.00004","tex_path":"articles/2607.00004/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-07-21T10:40:00Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2607.00004"},{"kagg_id":"KAGG:2607.00003","slug":"2607.00003","title":"Range-Gated Safety Envelopes for Reactive Quadruped Navigation","authors":[{"name":"Anika Roth","affiliation":"Robot Learning Lab, University of Zurich, Zurich, Switzerland"},{"name":"Selin Aydin","affiliation":"Robot Learning Lab, University of Zurich, Zurich, Switzerland"},{"name":"Lukas Brandt","affiliation":"Robot Learning Lab, University of Zurich, Zurich, Switzerland"}],"affiliation":null,"abstract":"Reactive navigation layers for legged robots commonly surround obstacles with a fixed inflation margin. Such a margin is conservative at low speed and can be unsafe at high speed, where velocity-controller lag lets the realized motion run past the point at which a command was judged safe. This paper proposes a range-gated safety envelope for reactive quadruped navigation. A configuration-based motion model, following the abstract robot configuration used by AgniNav, predicts the motion realized by each candidate velocity command under a bounded actuation delay, and the predicted free range along the planned direction is evaluated over the planning horizon. A command is admitted only if this predicted range envelope remains inside a speed-dependent corridor whose length and width grow with the worst-case stopping distance; otherwise the layer hands control to a braking maneuver. Under a bounded controller lag and a guaranteed braking deceleration, a certificate-style argument shows that every admitted command preserves the existence of a collision-free stop. The approach is evaluated in an Isaac Lab cluttered-corridor suite with doorway and stair-adjacent segments and on a Unitree Go2 traversing indoor hallways at setpoints from 1.5 to 4.0 m/s. Compared with a fixed inflation margin, the envelope raises collision-free success from 81.4 to 93.2 percent and mean speed from 2.70 to 3.05 m/s while reducing reaction overshoot from 14.2 to 12.8 cm. Ablations indicate that the controller-lag bound and range gating account for most of the gain.","publication_date":"2026-07-17","registered_at":"2026-07-22T08:07:00Z","updated_at":"2026-07-23T14:26:00Z","status":"published","html_path":null,"pdf_path":"articles/2607.00003/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"D. Fox, W. Burgard, and S. Thrun. The Dynamic Window Approach to Collision Avoidance. IEEE Robotics and Automation Magazine, 1997.","identifier":"doi:10.1109/100.580977","url":"https://doi.org/10.1109/100.580977"},{"text":"P. Fiorini and Z. Shiller. Motion Planning in Dynamic Environments Using Velocity Obstacles. International Journal of Robotics Research, 1998.","identifier":"doi:10.1177/027836499801700706","url":"https://doi.org/10.1177/027836499801700706"},{"text":"O. Khatib. Real-Time Obstacle Avoidance for Manipulators and Mobile Robots. International Journal of Robotics Research, 1986.","identifier":"doi:10.1177/027836498600500106","url":"https://doi.org/10.1177/027836498600500106"},{"text":"A. D. Ames, X. Xu, J. W. Grizzle, and P. Tabuada. Control Barrier Function Based Quadratic Programs for Safety Critical Systems. IEEE Transactions on Automatic Control, 2017. arXiv:1609.06408.","identifier":"arXiv:1609.06408","url":"https://arxiv.org/abs/1609.06408"},{"text":"A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada. Control Barrier Functions: Theory and Applications. European Control Conference, 2019. arXiv:1903.11199.","identifier":"arXiv:1903.11199","url":"https://arxiv.org/abs/1903.11199"},{"text":"T. He, C. Zhang, W. Xiao, G. He, C. Liu, and G. Shi. Agile But Safe: Learning Collision-Free High-Speed Legged Locomotion. Robotics: Science and Systems, 2024. arXiv:2401.17583.","identifier":"arXiv:2401.17583","url":"https://arxiv.org/abs/2401.17583"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Sridhar, D. Shah, C. Glossop, and S. Levine. NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration. arXiv:2310.07896, 2023.","identifier":"arXiv:2310.07896","url":"https://arxiv.org/abs/2310.07896"},{"text":"S. Kousik, S. Vaskov, F. Bu, M. Johnson-Roberson, and R. Vasudevan. Bridging the Gap Between Safety and Real-Time Performance in Receding-Horizon Trajectory Design for Mobile Robots. International Journal of Robotics Research, 2020. arXiv:1809.06746.","identifier":"arXiv:1809.06746","url":"https://arxiv.org/abs/1809.06746"},{"text":"A. Majumdar and R. Tedrake. Funnel Libraries for Real-Time Robust Feedback Motion Planning. International Journal of Robotics Research, 2017. arXiv:1601.04037.","identifier":"arXiv:1601.04037","url":"https://arxiv.org/abs/1601.04037"},{"text":"S. L. Herbert, M. Chen, S. Han, S. Bansal, J. F. Fisac, and C. J. Tomlin. FaSTrack: A Modular Framework for Fast and Guaranteed Safe Motion Planning. IEEE Conference on Decision and Control, 2017. arXiv:1703.07373.","identifier":"arXiv:1703.07373","url":"https://arxiv.org/abs/1703.07373"},{"text":"S. Bansal, M. Chen, S. Herbert, and C. J. Tomlin. Hamilton-Jacobi Reachability: A Brief Overview and Recent Advances. IEEE Conference on Decision and Control, 2017. arXiv:1709.07523.","identifier":"arXiv:1709.07523","url":"https://arxiv.org/abs/1709.07523"},{"text":"J. F. Fisac, A. K. Akametalu, M. N. Zeilinger, S. Kaynama, J. Gillula, and C. J. Tomlin. A General Safety Framework for Learning-Based Control in Uncertain Robotic Systems. IEEE Transactions on Automatic Control, 2019. arXiv:1705.01292.","identifier":"arXiv:1705.01292","url":"https://arxiv.org/abs/1705.01292"},{"text":"K. P. Wabersich and M. N. Zeilinger. Linear Model Predictive Safety Certification for Learning-Based Control. IEEE Conference on Decision and Control, 2018. arXiv:1803.08552.","identifier":"arXiv:1803.08552","url":"https://arxiv.org/abs/1803.08552"},{"text":"K.-C. Hsu, H. Hu, and J. F. Fisac. The Safety Filter: A Unified View of Safety-Critical Control in Autonomous Systems. Annual Review of Control, Robotics, and Autonomous Systems, 2024. arXiv:2309.05837.","identifier":"arXiv:2309.05837","url":"https://arxiv.org/abs/2309.05837"},{"text":"J. Zeng, B. Zhang, and K. Sreenath. Safety-Critical Model Predictive Control with Discrete-Time Control Barrier Function. American Control Conference, 2021. arXiv:2007.11718.","identifier":"arXiv:2007.11718","url":"https://arxiv.org/abs/2007.11718"},{"text":"R. Grandia, A. J. Taylor, A. D. Ames, and M. Hutter. Multi-Layered Safety for Legged Robots via Control Barrier Functions and Model Predictive Control. IEEE International Conference on Robotics and Automation, 2021. arXiv:2011.00032.","identifier":"arXiv:2011.00032","url":"https://arxiv.org/abs/2011.00032"},{"text":"T. G. Molnar, R. K. Cosner, A. W. Singletary, W. Ubellacker, and A. D. Ames. Model-Free Safety-Critical Control for Robotic Systems. IEEE Robotics and Automation Letters, 2022. arXiv:2109.09047.","identifier":"arXiv:2109.09047","url":"https://arxiv.org/abs/2109.09047"},{"text":"J. Tordesillas, B. T. Lopez, and J. P. How. FASTER: Fast and Safe Trajectory Planner for Flights in Unknown Environments. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2019. arXiv:1903.03558.","identifier":"arXiv:1903.03558","url":"https://arxiv.org/abs/1903.03558"},{"text":"X. Zhou, Z. Wang, H. Ye, C. Xu, and F. Gao. EGO-Planner: An ESDF-Free Gradient-Based Local Planner for Quadrotors. IEEE Robotics and Automation Letters, 2021. arXiv:2008.08835.","identifier":"arXiv:2008.08835","url":"https://arxiv.org/abs/2008.08835"},{"text":"A. Loquercio, E. Kaufmann, R. Ranftl, M. Mueller, V. Koltun, and D. Scaramuzza. Learning High-Speed Flight in the Wild. Science Robotics, 2021. arXiv:2110.05113.","identifier":"arXiv:2110.05113","url":"https://arxiv.org/abs/2110.05113"},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. Science Robotics, 2019. arXiv:1901.08652.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. Science Robotics, 2020. arXiv:2010.11251.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. Science Robotics, 2022. arXiv:2201.08117.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. Robotics: Science and Systems, 2021. arXiv:2107.04034.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. Conference on Robot Learning, 2021. arXiv:2109.11978.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2022. arXiv:2209.12827.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. Science Robotics, 2024. arXiv:2306.14874.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. Conference on Robot Learning, 2023. arXiv:2309.05665.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. IEEE International Conference on Robotics and Automation, 2024. arXiv:2309.14341.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"R. Grandia, F. Jenelten, S. Yang, F. Farshidian, and M. Hutter. Perceptive Locomotion Through Nonlinear Model-Predictive Control. IEEE Transactions on Robotics, 2023. arXiv:2208.08373.","identifier":"arXiv:2208.08373","url":"https://arxiv.org/abs/2208.08373"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. Robotics: Science and Systems, 2023. arXiv:2305.08510.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"D. D. Fan, K. Otsu, Y. Kubo, A. Dixit, J. Burdick, and A. Agha-Mohammadi. STEP: Stochastic Traversability Evaluation and Planning for Risk-Aware Off-road Navigation. Robotics: Science and Systems, 2021. arXiv:2103.02828.","identifier":"arXiv:2103.02828","url":"https://arxiv.org/abs/2103.02828"},{"text":"G. Kahn, P. Abbeel, and S. Levine. BADGR: An Autonomous Self-Supervised Learning-Based Navigation System. IEEE Robotics and Automation Letters, 2021. arXiv:2002.05700.","identifier":"arXiv:2002.05700","url":"https://arxiv.org/abs/2002.05700"},{"text":"W. Xu, Y. Cai, D. He, J. Lin, and F. Zhang. FAST-LIO2: Fast Direct LiDAR-Inertial Odometry. IEEE Transactions on Robotics, 2022. arXiv:2107.06829.","identifier":"arXiv:2107.06829","url":"https://arxiv.org/abs/2107.06829"},{"text":"H. Oleynikova, Z. Taylor, M. Fehr, R. Siegwart, and J. Nieto. Voxblox: Incremental 3D Euclidean Signed Distance Fields for On-Board MAV Planning. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2017. arXiv:1611.03631.","identifier":"arXiv:1611.03631","url":"https://arxiv.org/abs/1611.03631"},{"text":"M. Mittal et al. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. IEEE Robotics and Automation Letters, 2023. arXiv:2301.04195.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"V. Makoviychuk et al. Isaac Gym: High Performance GPU-Based Physics Simulation for Robot Learning. NeurIPS Datasets and Benchmarks Track, 2021. arXiv:2108.10470.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2017. arXiv:1703.06907.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"},{"text":"A. Kadian et al. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance? arXiv:1912.06321, 2019.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"},{"text":"N. Yokoyama, S. Ha, and D. Batra. Success Weighted by Completion Time: A Dynamics-Aware Evaluation Criteria for Embodied Navigation. arXiv:2103.08022, 2021.","identifier":"arXiv:2103.08022","url":"https://arxiv.org/abs/2103.08022"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2607.00003","metadata":{"title":"Range-Gated Safety Envelopes for Reactive Quadruped Navigation","authors":[{"name":"Anika Roth","affiliation":"Robot Learning Lab, University of Zurich, Zurich, Switzerland"},{"name":"Selin Aydin","affiliation":"Robot Learning Lab, University of Zurich, Zurich, Switzerland"},{"name":"Lukas Brandt","affiliation":"Robot Learning Lab, University of Zurich, Zurich, Switzerland"}],"publication_date":"2026-07-17","abstract":"Reactive navigation layers for legged robots commonly surround obstacles with a fixed inflation margin. Such a margin is conservative at low speed and can be unsafe at high speed, where velocity-controller lag lets the realized motion run past the point at which a command was judged safe. This paper proposes a range-gated safety envelope for reactive quadruped navigation. A configuration-based motion model, following the abstract robot configuration used by AgniNav, predicts the motion realized by each candidate velocity command under a bounded actuation delay, and the predicted free range along the planned direction is evaluated over the planning horizon. A command is admitted only if this predicted range envelope remains inside a speed-dependent corridor whose length and width grow with the worst-case stopping distance; otherwise the layer hands control to a braking maneuver. Under a bounded controller lag and a guaranteed braking deceleration, a certificate-style argument shows that every admitted command preserves the existence of a collision-free stop. The approach is evaluated in an Isaac Lab cluttered-corridor suite with doorway and stair-adjacent segments and on a Unitree Go2 traversing indoor hallways at setpoints from 1.5 to 4.0 m/s. Compared with a fixed inflation margin, the envelope raises collision-free success from 81.4 to 93.2 percent and mean speed from 2.70 to 3.05 m/s while reducing reaction overshoot from 14.2 to 12.8 cm. Ablations indicate that the controller-lag bound and range gating account for most of the gain.","references":[{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"D. Fox, W. Burgard, and S. Thrun. The Dynamic Window Approach to Collision Avoidance. IEEE Robotics and Automation Magazine, 1997.","identifier":"doi:10.1109/100.580977","url":"https://doi.org/10.1109/100.580977"},{"text":"P. Fiorini and Z. Shiller. Motion Planning in Dynamic Environments Using Velocity Obstacles. International Journal of Robotics Research, 1998.","identifier":"doi:10.1177/027836499801700706","url":"https://doi.org/10.1177/027836499801700706"},{"text":"O. Khatib. Real-Time Obstacle Avoidance for Manipulators and Mobile Robots. International Journal of Robotics Research, 1986.","identifier":"doi:10.1177/027836498600500106","url":"https://doi.org/10.1177/027836498600500106"},{"text":"A. D. Ames, X. Xu, J. W. Grizzle, and P. Tabuada. Control Barrier Function Based Quadratic Programs for Safety Critical Systems. IEEE Transactions on Automatic Control, 2017. arXiv:1609.06408.","identifier":"arXiv:1609.06408","url":"https://arxiv.org/abs/1609.06408"},{"text":"A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada. Control Barrier Functions: Theory and Applications. European Control Conference, 2019. arXiv:1903.11199.","identifier":"arXiv:1903.11199","url":"https://arxiv.org/abs/1903.11199"},{"text":"T. He, C. Zhang, W. Xiao, G. He, C. Liu, and G. Shi. Agile But Safe: Learning Collision-Free High-Speed Legged Locomotion. Robotics: Science and Systems, 2024. arXiv:2401.17583.","identifier":"arXiv:2401.17583","url":"https://arxiv.org/abs/2401.17583"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Sridhar, D. Shah, C. Glossop, and S. Levine. NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration. arXiv:2310.07896, 2023.","identifier":"arXiv:2310.07896","url":"https://arxiv.org/abs/2310.07896"},{"text":"S. Kousik, S. Vaskov, F. Bu, M. Johnson-Roberson, and R. Vasudevan. Bridging the Gap Between Safety and Real-Time Performance in Receding-Horizon Trajectory Design for Mobile Robots. International Journal of Robotics Research, 2020. arXiv:1809.06746.","identifier":"arXiv:1809.06746","url":"https://arxiv.org/abs/1809.06746"},{"text":"A. Majumdar and R. Tedrake. Funnel Libraries for Real-Time Robust Feedback Motion Planning. International Journal of Robotics Research, 2017. arXiv:1601.04037.","identifier":"arXiv:1601.04037","url":"https://arxiv.org/abs/1601.04037"},{"text":"S. L. Herbert, M. Chen, S. Han, S. Bansal, J. F. Fisac, and C. J. Tomlin. FaSTrack: A Modular Framework for Fast and Guaranteed Safe Motion Planning. IEEE Conference on Decision and Control, 2017. arXiv:1703.07373.","identifier":"arXiv:1703.07373","url":"https://arxiv.org/abs/1703.07373"},{"text":"S. Bansal, M. Chen, S. Herbert, and C. J. Tomlin. Hamilton-Jacobi Reachability: A Brief Overview and Recent Advances. IEEE Conference on Decision and Control, 2017. arXiv:1709.07523.","identifier":"arXiv:1709.07523","url":"https://arxiv.org/abs/1709.07523"},{"text":"J. F. Fisac, A. K. Akametalu, M. N. Zeilinger, S. Kaynama, J. Gillula, and C. J. Tomlin. A General Safety Framework for Learning-Based Control in Uncertain Robotic Systems. IEEE Transactions on Automatic Control, 2019. arXiv:1705.01292.","identifier":"arXiv:1705.01292","url":"https://arxiv.org/abs/1705.01292"},{"text":"K. P. Wabersich and M. N. Zeilinger. Linear Model Predictive Safety Certification for Learning-Based Control. IEEE Conference on Decision and Control, 2018. arXiv:1803.08552.","identifier":"arXiv:1803.08552","url":"https://arxiv.org/abs/1803.08552"},{"text":"K.-C. Hsu, H. Hu, and J. F. Fisac. The Safety Filter: A Unified View of Safety-Critical Control in Autonomous Systems. Annual Review of Control, Robotics, and Autonomous Systems, 2024. arXiv:2309.05837.","identifier":"arXiv:2309.05837","url":"https://arxiv.org/abs/2309.05837"},{"text":"J. Zeng, B. Zhang, and K. Sreenath. Safety-Critical Model Predictive Control with Discrete-Time Control Barrier Function. American Control Conference, 2021. arXiv:2007.11718.","identifier":"arXiv:2007.11718","url":"https://arxiv.org/abs/2007.11718"},{"text":"R. Grandia, A. J. Taylor, A. D. Ames, and M. Hutter. Multi-Layered Safety for Legged Robots via Control Barrier Functions and Model Predictive Control. IEEE International Conference on Robotics and Automation, 2021. arXiv:2011.00032.","identifier":"arXiv:2011.00032","url":"https://arxiv.org/abs/2011.00032"},{"text":"T. G. Molnar, R. K. Cosner, A. W. Singletary, W. Ubellacker, and A. D. Ames. Model-Free Safety-Critical Control for Robotic Systems. IEEE Robotics and Automation Letters, 2022. arXiv:2109.09047.","identifier":"arXiv:2109.09047","url":"https://arxiv.org/abs/2109.09047"},{"text":"J. Tordesillas, B. T. Lopez, and J. P. How. FASTER: Fast and Safe Trajectory Planner for Flights in Unknown Environments. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2019. arXiv:1903.03558.","identifier":"arXiv:1903.03558","url":"https://arxiv.org/abs/1903.03558"},{"text":"X. Zhou, Z. Wang, H. Ye, C. Xu, and F. Gao. EGO-Planner: An ESDF-Free Gradient-Based Local Planner for Quadrotors. IEEE Robotics and Automation Letters, 2021. arXiv:2008.08835.","identifier":"arXiv:2008.08835","url":"https://arxiv.org/abs/2008.08835"},{"text":"A. Loquercio, E. Kaufmann, R. Ranftl, M. Mueller, V. Koltun, and D. Scaramuzza. Learning High-Speed Flight in the Wild. Science Robotics, 2021. arXiv:2110.05113.","identifier":"arXiv:2110.05113","url":"https://arxiv.org/abs/2110.05113"},{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. Science Robotics, 2019. arXiv:1901.08652.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. Science Robotics, 2020. arXiv:2010.11251.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. Science Robotics, 2022. arXiv:2201.08117.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. Robotics: Science and Systems, 2021. arXiv:2107.04034.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. Conference on Robot Learning, 2021. arXiv:2109.11978.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2022. arXiv:2209.12827.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, N. Rudin, D. Sako, and M. Hutter. ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robots. Science Robotics, 2024. arXiv:2306.14874.","identifier":"arXiv:2306.14874","url":"https://arxiv.org/abs/2306.14874"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. Conference on Robot Learning, 2023. arXiv:2309.05665.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. IEEE International Conference on Robotics and Automation, 2024. arXiv:2309.14341.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"R. Grandia, F. Jenelten, S. Yang, F. Farshidian, and M. Hutter. Perceptive Locomotion Through Nonlinear Model-Predictive Control. IEEE Transactions on Robotics, 2023. arXiv:2208.08373.","identifier":"arXiv:2208.08373","url":"https://arxiv.org/abs/2208.08373"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. Robotics: Science and Systems, 2023. arXiv:2305.08510.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"D. D. Fan, K. Otsu, Y. Kubo, A. Dixit, J. Burdick, and A. Agha-Mohammadi. STEP: Stochastic Traversability Evaluation and Planning for Risk-Aware Off-road Navigation. Robotics: Science and Systems, 2021. arXiv:2103.02828.","identifier":"arXiv:2103.02828","url":"https://arxiv.org/abs/2103.02828"},{"text":"G. Kahn, P. Abbeel, and S. Levine. BADGR: An Autonomous Self-Supervised Learning-Based Navigation System. IEEE Robotics and Automation Letters, 2021. arXiv:2002.05700.","identifier":"arXiv:2002.05700","url":"https://arxiv.org/abs/2002.05700"},{"text":"W. Xu, Y. Cai, D. He, J. Lin, and F. Zhang. FAST-LIO2: Fast Direct LiDAR-Inertial Odometry. IEEE Transactions on Robotics, 2022. arXiv:2107.06829.","identifier":"arXiv:2107.06829","url":"https://arxiv.org/abs/2107.06829"},{"text":"H. Oleynikova, Z. Taylor, M. Fehr, R. Siegwart, and J. Nieto. Voxblox: Incremental 3D Euclidean Signed Distance Fields for On-Board MAV Planning. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2017. arXiv:1611.03631.","identifier":"arXiv:1611.03631","url":"https://arxiv.org/abs/1611.03631"},{"text":"M. Mittal et al. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. IEEE Robotics and Automation Letters, 2023. arXiv:2301.04195.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"V. Makoviychuk et al. Isaac Gym: High Performance GPU-Based Physics Simulation for Robot Learning. NeurIPS Datasets and Benchmarks Track, 2021. arXiv:2108.10470.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2017. arXiv:1703.06907.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"},{"text":"A. Kadian et al. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance? arXiv:1912.06321, 2019.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"},{"text":"N. Yokoyama, S. Ha, and D. Batra. Success Weighted by Completion Time: A Dynamics-Aware Evaluation Criteria for Embodied Navigation. arXiv:2103.08022, 2021.","identifier":"arXiv:2103.08022","url":"https://arxiv.org/abs/2103.08022"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2607.00003","slug":"2607.00003","registered_at":"2026-07-22T08:07:00Z","updated_at":"2026-07-23T14:26:00Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2607.00003","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2607.00003","tex_path":"articles/2607.00003/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-07-22T10:24:00Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2607.00003"},{"kagg_id":"KAGG:2607.00002","slug":"2607.00002","title":"Terrain-Aware Velocity Command Smoothing for Quadruped Navigation over Unstructured Ground","authors":[{"name":"Tomas Keller","affiliation":"Advanced Robotics Group, Toyota Research Institute, Palo Alto, CA, USA"},{"name":"Priya Raghunathan","affiliation":"Advanced Robotics Group, Toyota Research Institute, Palo Alto, CA, USA"},{"name":"Hana Sato","affiliation":"Advanced Robotics Group, Toyota Research Institute, Palo Alto, CA, USA"}],"affiliation":null,"abstract":"Local planners for legged robots emit velocity commands without regard for what the locomotion controller can smoothly realize. On unstructured ground this mismatch produces command chatter, degrades tracking, wastes energy, and triggers premature obstacle reactions. This paper presents a terrain-aware, differentiable command-smoothing layer placed between a configuration-driven local planner in the AgniNav style and an opaque quadruped velocity interface. The layer solves a short-horizon smoothing problem whose weights are scheduled by an online terrain-difficulty estimate, and it predicts realized motion with an interface model that is initialized by GAUGE-style active calibration and updated online. Its parameters are trained end-to-end in simulation through implicit differentiation, and its prediction of the realizable motion is fed back to the planner so that collision checking follows the motion the robot is expected to execute. The approach is evaluated in an Isaac Lab deformable-terrain suite with five terrain classes and in field trials with a Unitree B2 over gravel, exposed roots, and graded slopes at 1.0 to 2.5 m/s. Compared with unsmoothed planner output, it reduces command RMS from 9.8 to 9.1 cm/s and 95th-percentile tracking error from 14.7 to 14.2 cm, raises success rate from 86.0 to 91.5 percent, lowers an energy proxy from 1519 to 1385 J/km, and increases mean speed from 1.48 to 1.68 m/s. It also outperforms a fixed gain scheduler and a GAUGE-calibrated fixed-gain smoother. Ablations and recorded failure cases show which components matter and where the approach remains fragile.","publication_date":"2026-07-16","registered_at":"2026-07-18T15:23:00Z","updated_at":"2026-07-18T18:44:00Z","status":"published","html_path":null,"pdf_path":"articles/2607.00002/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. Science Robotics, 2019. arXiv:1901.08652.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. Science Robotics, 2020. arXiv:2010.11251.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. Science Robotics, 2022. arXiv:2201.08117.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. Robotics: Science and Systems, 2021. arXiv:2107.04034.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"X. Xiao, B. Liu, G. Warnell, and P. Stone. Motion Planning and Control for Mobile Robot Navigation Using Machine Learning: a Survey. Autonomous Robots, 2022. arXiv:2011.13112.","identifier":"arXiv:2011.13112","url":"https://arxiv.org/abs/2011.13112"},{"text":"T. Zang, H. Huang, S. Cheng, and W. Zhang. GAUGE: Planner-Conditioned Active Calibration of Opaque Quadruped Velocity Interfaces. arXiv:2607.02841, 2026.","identifier":"arXiv:2607.02841","url":"https://arxiv.org/abs/2607.02841"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. Conference on Robot Learning, 2021. arXiv:2109.11978.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"G. Ji, J. Mun, H. Kim, and J. Hwangbo. Concurrent Training of a Control Policy and a State Estimator for Dynamic and Robust Legged Locomotion. IEEE Robotics and Automation Letters, 2022. arXiv:2202.05481.","identifier":"arXiv:2202.05481","url":"https://arxiv.org/abs/2202.05481"},{"text":"G. B. Margolis, G. Yang, K. Paigwar, T. Chen, and P. Agrawal. Rapid Locomotion via Reinforcement Learning. Robotics: Science and Systems, 2022. arXiv:2205.02824.","identifier":"arXiv:2205.02824","url":"https://arxiv.org/abs/2205.02824"},{"text":"G. B. Margolis and P. Agrawal. Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior. Conference on Robot Learning, 2022. arXiv:2212.03238.","identifier":"arXiv:2212.03238","url":"https://arxiv.org/abs/2212.03238"},{"text":"Z. Fu, A. Kumar, J. Malik, and D. Pathak. Minimizing Energy Consumption Leads to the Emergence of Gaits in Legged Robots. Conference on Robot Learning, 2021. arXiv:2111.01674.","identifier":"arXiv:2111.01674","url":"https://arxiv.org/abs/2111.01674"},{"text":"Y. Yang, T. Zhang, E. Coumans, J. Tan, and B. Boots. Fast and Efficient Locomotion via Learned Gait Transitions. Conference on Robot Learning, 2021. arXiv:2104.04644.","identifier":"arXiv:2104.04644","url":"https://arxiv.org/abs/2104.04644"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. Conference on Robot Learning, 2022. arXiv:2211.07638.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. IEEE International Conference on Robotics and Automation, 2024. arXiv:2309.14341.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. Conference on Robot Learning, 2023. arXiv:2309.05665.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"D. Kim, J. Di Carlo, B. Katz, G. Bledt, and S. Kim. Highly Dynamic Quadruped Locomotion via Whole-Body Impulse Control and Model Predictive Control. arXiv:1909.06586, 2019.","identifier":"arXiv:1909.06586","url":"https://arxiv.org/abs/1909.06586"},{"text":"S. Gangapurwala, M. Geisert, R. Orsolino, M. Fallon, and I. Havoutis. RLOC: Terrain-Aware Legged Locomotion Using Reinforcement Learning and Optimal Control. IEEE Transactions on Robotics, 2022. arXiv:2012.03094.","identifier":"arXiv:2012.03094","url":"https://arxiv.org/abs/2012.03094"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2022. arXiv:2209.12827.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, L. Wellhausen, F. Farshidian, and M. Hutter. Learning a State Representation and Navigation in Cluttered and Dynamic Environments. IEEE Robotics and Automation Letters, 2021. arXiv:2103.04351.","identifier":"arXiv:2103.04351","url":"https://arxiv.org/abs/2103.04351"},{"text":"J. Truong, D. Yarats, T. Li, F. Meier, S. Chernova, D. Batra, and A. Rai. Learning Navigation Skills for Legged Robots with Learned Robot Embeddings. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2021. arXiv:2011.12255.","identifier":"arXiv:2011.12255","url":"https://arxiv.org/abs/2011.12255"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. Robotics: Science and Systems, 2023. arXiv:2305.08510.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"D. D. Fan, K. Otsu, Y. Kubo, A. Dixit, J. Burdick, and A. Agha-mohammadi. STEP: Stochastic Traversability Evaluation and Planning for Risk-Aware Off-road Navigation. Robotics: Science and Systems, 2021. arXiv:2103.02828.","identifier":"arXiv:2103.02828","url":"https://arxiv.org/abs/2103.02828"},{"text":"G. Kahn, P. Abbeel, and S. Levine. BADGR: An Autonomous Self-Supervised Learning-Based Navigation System. IEEE Robotics and Automation Letters, 2021. arXiv:2002.05700.","identifier":"arXiv:2002.05700","url":"https://arxiv.org/abs/2002.05700"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. IEEE International Conference on Robotics and Automation, 2023. arXiv:2210.03370.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"G. Williams, A. Aldrich, and E. A. Theodorou. Model Predictive Path Integral Control using Covariance Variable Importance Sampling. arXiv:1509.01149, 2015.","identifier":"arXiv:1509.01149","url":"https://arxiv.org/abs/1509.01149"},{"text":"X. Xiao, J. Biswas, and P. Stone. Learning Inverse Kinodynamics for Accurate High-Speed Off-Road Navigation on Unstructured Terrain. IEEE Robotics and Automation Letters, 2021. arXiv:2102.12667.","identifier":"arXiv:2102.12667","url":"https://arxiv.org/abs/2102.12667"},{"text":"H. Karnan, K. S. Sikand, P. Atreya, S. Rabiee, X. Xiao, G. Warnell, P. Stone, and J. Biswas. VI-IKD: High-Speed Accurate Off-Road Navigation using Learned Visual-Inertial Inverse Kinodynamics. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2022. arXiv:2203.15983.","identifier":"arXiv:2203.15983","url":"https://arxiv.org/abs/2203.15983"},{"text":"B. Amos and J. Z. Kolter. OptNet: Differentiable Optimization as a Layer in Neural Networks. International Conference on Machine Learning, 2017. arXiv:1703.00443.","identifier":"arXiv:1703.00443","url":"https://arxiv.org/abs/1703.00443"},{"text":"B. Amos, I. D. J. Rodriguez, J. Sacks, B. Boots, and J. Z. Kolter. Differentiable MPC for End-to-end Planning and Control. Advances in Neural Information Processing Systems, 2018. arXiv:1810.13400.","identifier":"arXiv:1810.13400","url":"https://arxiv.org/abs/1810.13400"},{"text":"A. Agrawal, B. Amos, S. Barratt, S. Boyd, S. Diamond, and J. Z. Kolter. Differentiable Convex Optimization Layers. Advances in Neural Information Processing Systems, 2019. arXiv:1910.12430.","identifier":"arXiv:1910.12430","url":"https://arxiv.org/abs/1910.12430"},{"text":"A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada. Control Barrier Functions: Theory and Applications. European Control Conference, 2019. arXiv:1903.11199.","identifier":"arXiv:1903.11199","url":"https://arxiv.org/abs/1903.11199"},{"text":"K. P. Wabersich and M. N. Zeilinger. Linear Model Predictive Safety Certification for Learning-Based Control. IEEE Conference on Decision and Control, 2018. arXiv:1803.08552.","identifier":"arXiv:1803.08552","url":"https://arxiv.org/abs/1803.08552"},{"text":"F. Berkenkamp, M. Turchetta, A. P. Schoellig, and A. Krause. Safe Model-based Reinforcement Learning with Stability Guarantees. Advances in Neural Information Processing Systems, 2017. arXiv:1705.08551.","identifier":"arXiv:1705.08551","url":"https://arxiv.org/abs/1705.08551"},{"text":"V. Makoviychuk, L. Wawrzyniak, Y. Guo, M. Lu, K. Storey, M. Macklin, D. Hoeller, N. Rudin, A. Allshire, A. Handa, and G. State. Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning. NeurIPS Datasets and Benchmarks Track, 2021. arXiv:2108.10470.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal, C. Yu, Q. Yu, J. Liu, N. Rudin, D. Hoeller, J. L. Yuan, R. Singh, Y. Guo, H. Mazhar, A. Mandlekar, B. Babich, G. State, M. Hutter, and A. Garg. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. IEEE Robotics and Automation Letters, 2023. arXiv:2301.04195.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2017. arXiv:1703.06907.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. IEEE International Conference on Robotics and Automation, 2018. arXiv:1710.06537.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"},{"text":"J. Tan, T. Zhang, E. Coumans, A. Iscen, Y. Bai, D. Hafner, S. Bohez, and V. Vanhoucke. Sim-to-Real: Learning Agile Locomotion For Quadruped Robots. Robotics: Science and Systems, 2018. arXiv:1804.10332.","identifier":"arXiv:1804.10332","url":"https://arxiv.org/abs/1804.10332"},{"text":"A. Kadian, J. Truong, A. Gokaslan, A. Clegg, E. Wijmans, S. Lee, M. Savva, S. Chernova, and D. Batra. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance? IEEE Robotics and Automation Letters, 2020. arXiv:1912.06321.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2607.00002","metadata":{"title":"Terrain-Aware Velocity Command Smoothing for Quadruped Navigation over Unstructured Ground","authors":[{"name":"Tomas Keller","affiliation":"Advanced Robotics Group, Toyota Research Institute, Palo Alto, CA, USA"},{"name":"Priya Raghunathan","affiliation":"Advanced Robotics Group, Toyota Research Institute, Palo Alto, CA, USA"},{"name":"Hana Sato","affiliation":"Advanced Robotics Group, Toyota Research Institute, Palo Alto, CA, USA"}],"publication_date":"2026-07-16","abstract":"Local planners for legged robots emit velocity commands without regard for what the locomotion controller can smoothly realize. On unstructured ground this mismatch produces command chatter, degrades tracking, wastes energy, and triggers premature obstacle reactions. This paper presents a terrain-aware, differentiable command-smoothing layer placed between a configuration-driven local planner in the AgniNav style and an opaque quadruped velocity interface. The layer solves a short-horizon smoothing problem whose weights are scheduled by an online terrain-difficulty estimate, and it predicts realized motion with an interface model that is initialized by GAUGE-style active calibration and updated online. Its parameters are trained end-to-end in simulation through implicit differentiation, and its prediction of the realizable motion is fed back to the planner so that collision checking follows the motion the robot is expected to execute. The approach is evaluated in an Isaac Lab deformable-terrain suite with five terrain classes and in field trials with a Unitree B2 over gravel, exposed roots, and graded slopes at 1.0 to 2.5 m/s. Compared with unsmoothed planner output, it reduces command RMS from 9.8 to 9.1 cm/s and 95th-percentile tracking error from 14.7 to 14.2 cm, raises success rate from 86.0 to 91.5 percent, lowers an energy proxy from 1519 to 1385 J/km, and increases mean speed from 1.48 to 1.68 m/s. It also outperforms a fixed gain scheduler and a GAUGE-calibrated fixed-gain smoother. Ablations and recorded failure cases show which components matter and where the approach remains fragile.","references":[{"text":"J. Hwangbo, J. Lee, A. Dosovitskiy, D. Bellicoso, V. Tsounis, V. Koltun, and M. Hutter. Learning Agile and Dynamic Motor Skills for Legged Robots. Science Robotics, 2019. arXiv:1901.08652.","identifier":"arXiv:1901.08652","url":"https://arxiv.org/abs/1901.08652"},{"text":"J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Quadrupedal Locomotion over Challenging Terrain. Science Robotics, 2020. arXiv:2010.11251.","identifier":"arXiv:2010.11251","url":"https://arxiv.org/abs/2010.11251"},{"text":"T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter. Learning Robust Perceptive Locomotion for Quadrupedal Robots in the Wild. Science Robotics, 2022. arXiv:2201.08117.","identifier":"arXiv:2201.08117","url":"https://arxiv.org/abs/2201.08117"},{"text":"A. Kumar, Z. Fu, D. Pathak, and J. Malik. RMA: Rapid Motor Adaptation for Legged Robots. Robotics: Science and Systems, 2021. arXiv:2107.04034.","identifier":"arXiv:2107.04034","url":"https://arxiv.org/abs/2107.04034"},{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"X. Xiao, B. Liu, G. Warnell, and P. Stone. Motion Planning and Control for Mobile Robot Navigation Using Machine Learning: a Survey. Autonomous Robots, 2022. arXiv:2011.13112.","identifier":"arXiv:2011.13112","url":"https://arxiv.org/abs/2011.13112"},{"text":"T. Zang, H. Huang, S. Cheng, and W. Zhang. GAUGE: Planner-Conditioned Active Calibration of Opaque Quadruped Velocity Interfaces. arXiv:2607.02841, 2026.","identifier":"arXiv:2607.02841","url":"https://arxiv.org/abs/2607.02841"},{"text":"N. Rudin, D. Hoeller, P. Reist, and M. Hutter. Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning. Conference on Robot Learning, 2021. arXiv:2109.11978.","identifier":"arXiv:2109.11978","url":"https://arxiv.org/abs/2109.11978"},{"text":"G. Ji, J. Mun, H. Kim, and J. Hwangbo. Concurrent Training of a Control Policy and a State Estimator for Dynamic and Robust Legged Locomotion. IEEE Robotics and Automation Letters, 2022. arXiv:2202.05481.","identifier":"arXiv:2202.05481","url":"https://arxiv.org/abs/2202.05481"},{"text":"G. B. Margolis, G. Yang, K. Paigwar, T. Chen, and P. Agrawal. Rapid Locomotion via Reinforcement Learning. Robotics: Science and Systems, 2022. arXiv:2205.02824.","identifier":"arXiv:2205.02824","url":"https://arxiv.org/abs/2205.02824"},{"text":"G. B. Margolis and P. Agrawal. Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior. Conference on Robot Learning, 2022. arXiv:2212.03238.","identifier":"arXiv:2212.03238","url":"https://arxiv.org/abs/2212.03238"},{"text":"Z. Fu, A. Kumar, J. Malik, and D. Pathak. Minimizing Energy Consumption Leads to the Emergence of Gaits in Legged Robots. Conference on Robot Learning, 2021. arXiv:2111.01674.","identifier":"arXiv:2111.01674","url":"https://arxiv.org/abs/2111.01674"},{"text":"Y. Yang, T. Zhang, E. Coumans, J. Tan, and B. Boots. Fast and Efficient Locomotion via Learned Gait Transitions. Conference on Robot Learning, 2021. arXiv:2104.04644.","identifier":"arXiv:2104.04644","url":"https://arxiv.org/abs/2104.04644"},{"text":"A. Agarwal, A. Kumar, J. Malik, and D. Pathak. Legged Locomotion in Challenging Terrains using Egocentric Vision. Conference on Robot Learning, 2022. arXiv:2211.07638.","identifier":"arXiv:2211.07638","url":"https://arxiv.org/abs/2211.07638"},{"text":"X. Cheng, K. Shi, A. Agarwal, and D. Pathak. Extreme Parkour with Legged Robots. IEEE International Conference on Robotics and Automation, 2024. arXiv:2309.14341.","identifier":"arXiv:2309.14341","url":"https://arxiv.org/abs/2309.14341"},{"text":"Z. Zhuang, Z. Fu, J. Wang, C. Atkeson, S. Schwertfeger, C. Finn, and H. Zhao. Robot Parkour Learning. Conference on Robot Learning, 2023. arXiv:2309.05665.","identifier":"arXiv:2309.05665","url":"https://arxiv.org/abs/2309.05665"},{"text":"D. Kim, J. Di Carlo, B. Katz, G. Bledt, and S. Kim. Highly Dynamic Quadruped Locomotion via Whole-Body Impulse Control and Model Predictive Control. arXiv:1909.06586, 2019.","identifier":"arXiv:1909.06586","url":"https://arxiv.org/abs/1909.06586"},{"text":"S. Gangapurwala, M. Geisert, R. Orsolino, M. Fallon, and I. Havoutis. RLOC: Terrain-Aware Legged Locomotion Using Reinforcement Learning and Optimal Control. IEEE Transactions on Robotics, 2022. arXiv:2012.03094.","identifier":"arXiv:2012.03094","url":"https://arxiv.org/abs/2012.03094"},{"text":"N. Rudin, D. Hoeller, M. Bjelonic, and M. Hutter. Advanced Skills by Learning Locomotion and Local Navigation End-to-End. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2022. arXiv:2209.12827.","identifier":"arXiv:2209.12827","url":"https://arxiv.org/abs/2209.12827"},{"text":"D. Hoeller, L. Wellhausen, F. Farshidian, and M. Hutter. Learning a State Representation and Navigation in Cluttered and Dynamic Environments. IEEE Robotics and Automation Letters, 2021. arXiv:2103.04351.","identifier":"arXiv:2103.04351","url":"https://arxiv.org/abs/2103.04351"},{"text":"J. Truong, D. Yarats, T. Li, F. Meier, S. Chernova, D. Batra, and A. Rai. Learning Navigation Skills for Legged Robots with Learned Robot Embeddings. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2021. arXiv:2011.12255.","identifier":"arXiv:2011.12255","url":"https://arxiv.org/abs/2011.12255"},{"text":"J. Frey, M. Mattamala, N. Chebrolu, C. Cadena, M. Fallon, and M. Hutter. Fast Traversability Estimation for Wild Visual Navigation. Robotics: Science and Systems, 2023. arXiv:2305.08510.","identifier":"arXiv:2305.08510","url":"https://arxiv.org/abs/2305.08510"},{"text":"D. D. Fan, K. Otsu, Y. Kubo, A. Dixit, J. Burdick, and A. Agha-mohammadi. STEP: Stochastic Traversability Evaluation and Planning for Risk-Aware Off-road Navigation. Robotics: Science and Systems, 2021. arXiv:2103.02828.","identifier":"arXiv:2103.02828","url":"https://arxiv.org/abs/2103.02828"},{"text":"G. Kahn, P. Abbeel, and S. Levine. BADGR: An Autonomous Self-Supervised Learning-Based Navigation System. IEEE Robotics and Automation Letters, 2021. arXiv:2002.05700.","identifier":"arXiv:2002.05700","url":"https://arxiv.org/abs/2002.05700"},{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. IEEE International Conference on Robotics and Automation, 2023. arXiv:2210.03370.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"G. Williams, A. Aldrich, and E. A. Theodorou. Model Predictive Path Integral Control using Covariance Variable Importance Sampling. arXiv:1509.01149, 2015.","identifier":"arXiv:1509.01149","url":"https://arxiv.org/abs/1509.01149"},{"text":"X. Xiao, J. Biswas, and P. Stone. Learning Inverse Kinodynamics for Accurate High-Speed Off-Road Navigation on Unstructured Terrain. IEEE Robotics and Automation Letters, 2021. arXiv:2102.12667.","identifier":"arXiv:2102.12667","url":"https://arxiv.org/abs/2102.12667"},{"text":"H. Karnan, K. S. Sikand, P. Atreya, S. Rabiee, X. Xiao, G. Warnell, P. Stone, and J. Biswas. VI-IKD: High-Speed Accurate Off-Road Navigation using Learned Visual-Inertial Inverse Kinodynamics. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2022. arXiv:2203.15983.","identifier":"arXiv:2203.15983","url":"https://arxiv.org/abs/2203.15983"},{"text":"B. Amos and J. Z. Kolter. OptNet: Differentiable Optimization as a Layer in Neural Networks. International Conference on Machine Learning, 2017. arXiv:1703.00443.","identifier":"arXiv:1703.00443","url":"https://arxiv.org/abs/1703.00443"},{"text":"B. Amos, I. D. J. Rodriguez, J. Sacks, B. Boots, and J. Z. Kolter. Differentiable MPC for End-to-end Planning and Control. Advances in Neural Information Processing Systems, 2018. arXiv:1810.13400.","identifier":"arXiv:1810.13400","url":"https://arxiv.org/abs/1810.13400"},{"text":"A. Agrawal, B. Amos, S. Barratt, S. Boyd, S. Diamond, and J. Z. Kolter. Differentiable Convex Optimization Layers. Advances in Neural Information Processing Systems, 2019. arXiv:1910.12430.","identifier":"arXiv:1910.12430","url":"https://arxiv.org/abs/1910.12430"},{"text":"A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada. Control Barrier Functions: Theory and Applications. European Control Conference, 2019. arXiv:1903.11199.","identifier":"arXiv:1903.11199","url":"https://arxiv.org/abs/1903.11199"},{"text":"K. P. Wabersich and M. N. Zeilinger. Linear Model Predictive Safety Certification for Learning-Based Control. IEEE Conference on Decision and Control, 2018. arXiv:1803.08552.","identifier":"arXiv:1803.08552","url":"https://arxiv.org/abs/1803.08552"},{"text":"F. Berkenkamp, M. Turchetta, A. P. Schoellig, and A. Krause. Safe Model-based Reinforcement Learning with Stability Guarantees. Advances in Neural Information Processing Systems, 2017. arXiv:1705.08551.","identifier":"arXiv:1705.08551","url":"https://arxiv.org/abs/1705.08551"},{"text":"V. Makoviychuk, L. Wawrzyniak, Y. Guo, M. Lu, K. Storey, M. Macklin, D. Hoeller, N. Rudin, A. Allshire, A. Handa, and G. State. Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning. NeurIPS Datasets and Benchmarks Track, 2021. arXiv:2108.10470.","identifier":"arXiv:2108.10470","url":"https://arxiv.org/abs/2108.10470"},{"text":"M. Mittal, C. Yu, Q. Yu, J. Liu, N. Rudin, D. Hoeller, J. L. Yuan, R. Singh, Y. Guo, H. Mazhar, A. Mandlekar, B. Babich, G. State, M. Hutter, and A. Garg. Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments. IEEE Robotics and Automation Letters, 2023. arXiv:2301.04195.","identifier":"arXiv:2301.04195","url":"https://arxiv.org/abs/2301.04195"},{"text":"J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2017. arXiv:1703.06907.","identifier":"arXiv:1703.06907","url":"https://arxiv.org/abs/1703.06907"},{"text":"X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel. Sim-to-Real Transfer of Robotic Control with Dynamics Randomization. IEEE International Conference on Robotics and Automation, 2018. arXiv:1710.06537.","identifier":"arXiv:1710.06537","url":"https://arxiv.org/abs/1710.06537"},{"text":"J. Tan, T. Zhang, E. Coumans, A. Iscen, Y. Bai, D. Hafner, S. Bohez, and V. Vanhoucke. Sim-to-Real: Learning Agile Locomotion For Quadruped Robots. Robotics: Science and Systems, 2018. arXiv:1804.10332.","identifier":"arXiv:1804.10332","url":"https://arxiv.org/abs/1804.10332"},{"text":"A. Kadian, J. Truong, A. Gokaslan, A. Clegg, E. Wijmans, S. Lee, M. Savva, S. Chernova, and D. Batra. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance? IEEE Robotics and Automation Letters, 2020. arXiv:1912.06321.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2607.00002","slug":"2607.00002","registered_at":"2026-07-18T15:23:00Z","updated_at":"2026-07-18T18:44:00Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2607.00002","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2607.00002","tex_path":"articles/2607.00002/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-07-18T17:40:00Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2607.00002"},{"kagg_id":"KAGG:2607.00001","slug":"2607.00001","title":"Cross-Embodiment Transfer in Vision-Based Robot Navigation: Challenges, Methods, and Open Problems","authors":[{"name":"Yiliang Zhang","affiliation":"Eastern Institute of Technology, Ningbo, Zhejiang, China"}],"affiliation":null,"abstract":"Cross-embodiment transfer in visual navigation is often described as a domain-generalization problem, but this framing is incomplete: changing the robot simultaneously changes the camera viewpoint, collision geometry, motion constraints, action semantics, and the consequences of perception error. This survey argues that reliable transfer therefore requires a factorized navigation stack with four explicit components: an environment-centric representation, an embodiment descriptor, a feasibility layer that converts scene understanding into body-aware constraints, and a deployment layer that handles dynamics and sim-to-real mismatch. We organize recent work into three complementary families: heterogeneous-data generalist policies, embodiment-randomized or embodiment-conditioned policies, and geometric abstraction with configuration-driven planning. Representative systems include GNM, ViNT, NoMaD, RING, NavFoM, EA-Nav, CeRLP, and AgniNav. We then connect these methods to evidence from sim-to-real navigation research, showing why simulator fidelity alone is insufficient and why collision behavior, action interfaces, and evaluation design strongly affect real-world conclusions. Finally, we propose an evaluation protocol that separates environment, sensor, geometry, dynamics, and reality gaps, and identify open problems in online embodiment estimation, adaptive safety margins, heterogeneous sensing, formally constrained learning, and cross-embodiment foundation models.","publication_date":"2026-07-15","registered_at":"2026-07-29T14:45:00Z","updated_at":"2026-07-29T17:45:08Z","status":"published","html_path":"data/articles/2607.00001/index.html","pdf_path":"articles/2607.00001/paper.pdf","source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"references":[{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Sridhar, D. Shah, C. Glossop, and S. Levine. NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration. arXiv:2310.07896, 2023.","identifier":"arXiv:2310.07896","url":"https://arxiv.org/abs/2310.07896"},{"text":"A. Eftekhar et al. The One RING: a Robotic Indoor Navigation Generalist. arXiv:2412.14401, 2024.","identifier":"arXiv:2412.14401","url":"https://arxiv.org/abs/2412.14401"},{"text":"J. Zhang et al. Embodied Navigation Foundation Model. arXiv:2509.12129, 2025.","identifier":"arXiv:2509.12129","url":"https://arxiv.org/abs/2509.12129"},{"text":"J. Zhang, Y. Du, X. Guo, S. Sun, X. Liu, Y. Sun, G. Lu, W. Sui, and J. Li. EA-Nav: Learning Safe Visual Navigation Policies with Embodiment Awareness. arXiv:2607.19880, 2026.","identifier":"arXiv:2607.19880","url":"https://arxiv.org/abs/2607.19880"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"M. Savva et al. Habitat: A Platform for Embodied AI Research. arXiv:1904.01201, 2019.","identifier":"arXiv:1904.01201","url":"https://arxiv.org/abs/1904.01201"},{"text":"F. Xia, A. Zamir, Z.-Y. He, A. Sax, J. Malik, and S. Savarese. Gibson Env: Real-World Perception for Embodied Agents. arXiv:1808.10654, 2018.","identifier":"arXiv:1808.10654","url":"https://arxiv.org/abs/1808.10654"},{"text":"A. Kadian et al. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance? arXiv:1912.06321, 2019.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"},{"text":"P. Anderson, A. Shrivastava, J. Truong, A. Majumdar, D. Parikh, D. Batra, and S. Lee. Sim-to-Real Transfer for Vision-and-Language Navigation. Proceedings of Machine Learning Research, 155:671-681, 2021.","identifier":"PMLR 155","url":"https://proceedings.mlr.press/v155/anderson21a.html"},{"text":"J. Truong, M. Rudolph, N. H. Yokoyama, S. Chernova, D. Batra, and A. Rai. Rethinking Sim2Real: Lower Fidelity Simulation Leads to Higher Sim2Real Transfer in Navigation. Proceedings of Machine Learning Research, 205:859-870, 2023.","identifier":"PMLR 205","url":"https://proceedings.mlr.press/v205/truong23a.html"},{"text":"Z. Wang, X. Li, J. Yang, Y. Liu, and S. Jiang. Sim-to-Real Transfer via 3D Feature Fields for Vision-and-Language Navigation. Proceedings of Machine Learning Research, 270:2982-2995, 2025.","identifier":"PMLR 270","url":"https://proceedings.mlr.press/v270/wang25i.html"},{"text":"P. Anderson et al. On Evaluation of Embodied Navigation Agents. arXiv:1807.06757, 2018.","identifier":"arXiv:1807.06757","url":"https://arxiv.org/abs/1807.06757"},{"text":"N. Yokoyama, S. Ha, and D. Batra. Success Weighted by Completion Time: A Dynamics-Aware Evaluation Criteria for Embodied Navigation. arXiv:2103.08022, 2021.","identifier":"arXiv:2103.08022","url":"https://arxiv.org/abs/2103.08022"},{"text":"J. Hu, S. Yuan, G. C. R. Bethala, A. Tzes, and Y. Fang. Learning Adaptive Safety Margins for Visual Navigation. arXiv:2607.18200, 2026.","identifier":"arXiv:2607.18200","url":"https://arxiv.org/abs/2607.18200"}],"canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2607.00001","metadata":{"title":"Cross-Embodiment Transfer in Vision-Based Robot Navigation: Challenges, Methods, and Open Problems","authors":[{"name":"Yiliang Zhang","affiliation":"Eastern Institute of Technology, Ningbo, Zhejiang, China"}],"publication_date":"2026-07-15","abstract":"Cross-embodiment transfer in visual navigation is often described as a domain-generalization problem, but this framing is incomplete: changing the robot simultaneously changes the camera viewpoint, collision geometry, motion constraints, action semantics, and the consequences of perception error. This survey argues that reliable transfer therefore requires a factorized navigation stack with four explicit components: an environment-centric representation, an embodiment descriptor, a feasibility layer that converts scene understanding into body-aware constraints, and a deployment layer that handles dynamics and sim-to-real mismatch. We organize recent work into three complementary families: heterogeneous-data generalist policies, embodiment-randomized or embodiment-conditioned policies, and geometric abstraction with configuration-driven planning. Representative systems include GNM, ViNT, NoMaD, RING, NavFoM, EA-Nav, CeRLP, and AgniNav. We then connect these methods to evidence from sim-to-real navigation research, showing why simulator fidelity alone is insufficient and why collision behavior, action interfaces, and evaluation design strongly affect real-world conclusions. Finally, we propose an evaluation protocol that separates environment, sensor, geometry, dynamics, and reality gaps, and identify open problems in online embodiment estimation, adaptive safety margins, heterogeneous sensing, formally constrained learning, and cross-embodiment foundation models.","references":[{"text":"D. Shah, A. Sridhar, A. Bhorkar, N. Hirose, and S. Levine. GNM: A General Navigation Model to Drive Any Robot. arXiv:2210.03370, 2022.","identifier":"arXiv:2210.03370","url":"https://arxiv.org/abs/2210.03370"},{"text":"D. Shah, A. Sridhar, N. Dashora, K. Stachowicz, K. Black, N. Hirose, and S. Levine. ViNT: A Foundation Model for Visual Navigation. arXiv:2306.14846, 2023.","identifier":"arXiv:2306.14846","url":"https://arxiv.org/abs/2306.14846"},{"text":"A. Sridhar, D. Shah, C. Glossop, and S. Levine. NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration. arXiv:2310.07896, 2023.","identifier":"arXiv:2310.07896","url":"https://arxiv.org/abs/2310.07896"},{"text":"A. Eftekhar et al. The One RING: a Robotic Indoor Navigation Generalist. arXiv:2412.14401, 2024.","identifier":"arXiv:2412.14401","url":"https://arxiv.org/abs/2412.14401"},{"text":"J. Zhang et al. Embodied Navigation Foundation Model. arXiv:2509.12129, 2025.","identifier":"arXiv:2509.12129","url":"https://arxiv.org/abs/2509.12129"},{"text":"J. Zhang, Y. Du, X. Guo, S. Sun, X. Liu, Y. Sun, G. Lu, W. Sui, and J. Li. EA-Nav: Learning Safe Visual Navigation Policies with Embodiment Awareness. arXiv:2607.19880, 2026.","identifier":"arXiv:2607.19880","url":"https://arxiv.org/abs/2607.19880"},{"text":"H. Xi, M. Tan, X. Zhang, S. Cheng, S. Wang, Y. Gu, X. Shen, and W. Zhang. CeRLP: A Cross-embodiment Robot Local Planning Framework for Visual Navigation. arXiv:2603.19602, 2026.","identifier":"arXiv:2603.19602","url":"https://arxiv.org/abs/2603.19602"},{"text":"T. Zang, S. Cheng, H. Huang, S. Wang, and W. Zhang. AgniNav: Configuration-Driven Cross-Embodiment Local Planning for Robot Navigation. arXiv:2606.10903, 2026.","identifier":"arXiv:2606.10903","url":"https://arxiv.org/abs/2606.10903"},{"text":"M. Savva et al. Habitat: A Platform for Embodied AI Research. arXiv:1904.01201, 2019.","identifier":"arXiv:1904.01201","url":"https://arxiv.org/abs/1904.01201"},{"text":"F. Xia, A. Zamir, Z.-Y. He, A. Sax, J. Malik, and S. Savarese. Gibson Env: Real-World Perception for Embodied Agents. arXiv:1808.10654, 2018.","identifier":"arXiv:1808.10654","url":"https://arxiv.org/abs/1808.10654"},{"text":"A. Kadian et al. Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance? arXiv:1912.06321, 2019.","identifier":"arXiv:1912.06321","url":"https://arxiv.org/abs/1912.06321"},{"text":"P. Anderson, A. Shrivastava, J. Truong, A. Majumdar, D. Parikh, D. Batra, and S. Lee. Sim-to-Real Transfer for Vision-and-Language Navigation. Proceedings of Machine Learning Research, 155:671-681, 2021.","identifier":"PMLR 155","url":"https://proceedings.mlr.press/v155/anderson21a.html"},{"text":"J. Truong, M. Rudolph, N. H. Yokoyama, S. Chernova, D. Batra, and A. Rai. Rethinking Sim2Real: Lower Fidelity Simulation Leads to Higher Sim2Real Transfer in Navigation. Proceedings of Machine Learning Research, 205:859-870, 2023.","identifier":"PMLR 205","url":"https://proceedings.mlr.press/v205/truong23a.html"},{"text":"Z. Wang, X. Li, J. Yang, Y. Liu, and S. Jiang. Sim-to-Real Transfer via 3D Feature Fields for Vision-and-Language Navigation. Proceedings of Machine Learning Research, 270:2982-2995, 2025.","identifier":"PMLR 270","url":"https://proceedings.mlr.press/v270/wang25i.html"},{"text":"P. Anderson et al. On Evaluation of Embodied Navigation Agents. arXiv:1807.06757, 2018.","identifier":"arXiv:1807.06757","url":"https://arxiv.org/abs/1807.06757"},{"text":"N. Yokoyama, S. Ha, and D. Batra. Success Weighted by Completion Time: A Dynamics-Aware Evaluation Criteria for Embodied Navigation. arXiv:2103.08022, 2021.","identifier":"arXiv:2103.08022","url":"https://arxiv.org/abs/2103.08022"},{"text":"J. Hu, S. Yuan, G. C. R. Bethala, A. Tzes, and Y. Fang. Learning Adaptive Safety Margins for Visual Navigation. arXiv:2607.18200, 2026.","identifier":"arXiv:2607.18200","url":"https://arxiv.org/abs/2607.18200"}],"affiliation":null,"source_type":"preprint","source_label":"Kaggleyes Academy Preprint","license":"CC BY 4.0","has_doi":false,"external_doi":null,"status":"published","kagg_id":"KAGG:2607.00001","slug":"2607.00001","registered_at":"2026-07-29T14:45:00Z","updated_at":"2026-07-29T17:45:08Z","canonical_url":"https://academy.kaggleyes.top/abs/KAGG:2607.00001","resolver_url":"https://academy.kaggleyes.top/id/KAGG:2607.00001","html_path":"data/articles/2607.00001/index.html","tex_path":"articles/2607.00001/source.tex","typesetting":{"class":"kaggpreprint","version":"v1.0","engine":"XeLaTeX","rendered_at":"2026-08-14T21:04:29Z"}},"resolver_url":"https://academy.kaggleyes.top/id/KAGG:2607.00001"}]