SoloParkour: Constrained Reinforcement Learning for Visual Locomotion from Privileged Experience

Elliot Chane-Sane, Joseph Amigo, Thomas Flayols, Ludovic Righetti, Nicolas Mansard
Proceedings of The 8th Conference on Robot Learning, PMLR 270:4268-4285, 2025.

Abstract

Parkour poses a significant challenge for legged robots, requiring navigation through complex environments with agility and precision based on limited sensory inputs. In this work, we introduce a novel method for training end-to-end visual policies, from depth pixels to robot control commands, to achieve agile and safe quadruped locomotion. We formulate robot parkour as a constrained reinforcement learning (RL) problem designed to maximize the emergence of agile skills within the robot’s physical limits while ensuring safety. We first train a policy without vision using privileged information about the robot’s surroundings. We then generate experience from this privileged policy to warm-start a sample efficient off-policy RL algorithm from depth images. This allows the robot to adapt behaviors from this privileged experience to visual locomotion while circumventing the high computational costs of RL directly from pixels. We demonstrate the effectiveness of our method on a real Solo-12 robot, showcasing its capability to perform a variety of parkour skills such as walking, climbing, leaping, and crawling.

Cite this Paper

BibTeX
@InProceedings{pmlr-v270-chane-sane25a, title = {SoloParkour: Constrained Reinforcement Learning for Visual Locomotion from Privileged Experience}, author = {Chane-Sane, Elliot and Amigo, Joseph and Flayols, Thomas and Righetti, Ludovic and Mansard, Nicolas}, booktitle = {Proceedings of The 8th Conference on Robot Learning}, pages = {4268--4285}, year = {2025}, editor = {Agrawal, Pulkit and Kroemer, Oliver and Burgard, Wolfram}, volume = {270}, series = {Proceedings of Machine Learning Research}, month = {06--09 Nov}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v270/main/assets/chane-sane25a/chane-sane25a.pdf}, url = {https://proceedings.mlr.press/v270/chane-sane25a.html}, abstract = {Parkour poses a significant challenge for legged robots, requiring navigation through complex environments with agility and precision based on limited sensory inputs. In this work, we introduce a novel method for training end-to-end visual policies, from depth pixels to robot control commands, to achieve agile and safe quadruped locomotion. We formulate robot parkour as a constrained reinforcement learning (RL) problem designed to maximize the emergence of agile skills within the robot’s physical limits while ensuring safety. We first train a policy without vision using privileged information about the robot’s surroundings. We then generate experience from this privileged policy to warm-start a sample efficient off-policy RL algorithm from depth images. This allows the robot to adapt behaviors from this privileged experience to visual locomotion while circumventing the high computational costs of RL directly from pixels. We demonstrate the effectiveness of our method on a real Solo-12 robot, showcasing its capability to perform a variety of parkour skills such as walking, climbing, leaping, and crawling.} }
Endnote
%0 Conference Paper %T SoloParkour: Constrained Reinforcement Learning for Visual Locomotion from Privileged Experience %A Elliot Chane-Sane %A Joseph Amigo %A Thomas Flayols %A Ludovic Righetti %A Nicolas Mansard %B Proceedings of The 8th Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2025 %E Pulkit Agrawal %E Oliver Kroemer %E Wolfram Burgard %F pmlr-v270-chane-sane25a %I PMLR %P 4268--4285 %U https://proceedings.mlr.press/v270/chane-sane25a.html %V 270 %X Parkour poses a significant challenge for legged robots, requiring navigation through complex environments with agility and precision based on limited sensory inputs. In this work, we introduce a novel method for training end-to-end visual policies, from depth pixels to robot control commands, to achieve agile and safe quadruped locomotion. We formulate robot parkour as a constrained reinforcement learning (RL) problem designed to maximize the emergence of agile skills within the robot’s physical limits while ensuring safety. We first train a policy without vision using privileged information about the robot’s surroundings. We then generate experience from this privileged policy to warm-start a sample efficient off-policy RL algorithm from depth images. This allows the robot to adapt behaviors from this privileged experience to visual locomotion while circumventing the high computational costs of RL directly from pixels. We demonstrate the effectiveness of our method on a real Solo-12 robot, showcasing its capability to perform a variety of parkour skills such as walking, climbing, leaping, and crawling.
APA
Chane-Sane, E., Amigo, J., Flayols, T., Righetti, L. & Mansard, N.. (2025). SoloParkour: Constrained Reinforcement Learning for Visual Locomotion from Privileged Experience. Proceedings of The 8th Conference on Robot Learning, in Proceedings of Machine Learning Research 270:4268-4285 Available from https://proceedings.mlr.press/v270/chane-sane25a.html.

Related Material