In-Flight Attitude Control of a Quadruped using Deep Reinforcement Learning

Tarek El-Agroudi, Finn Gross Maurer, Jørgen Anker Olsen, Kostas Alexis
Proceedings of The 8th Conference on Robot Learning, PMLR 270:5014-5029, 2025.

Abstract

We present the development and real world demonstration of an in-flight attitude control law for a small low-cost quadruped with a five-bar-linkage leg design using only its legs as reaction masses. The control law is trained using deep reinforcement learning (DRL) and specifically through Proximal Policy Optimization (PPO) in the NVIDIA Omniverse Isaac Sim simulator with a GPU-accelerated DRL pipeline. To demonstrate the policy, a small quadruped is designed, constructed, and evaluated both on a rotating pole test setup and in free fall. During a free fall of 0.7 seconds, the quadruped follows commanded attitude steps of 45 degrees in all principal axes, and achieves an average base angular velocity of 110 degrees per second during large attitude reference steps.

Cite this Paper

BibTeX
@InProceedings{pmlr-v270-el-agroudi25a, title = {In-Flight Attitude Control of a Quadruped using Deep Reinforcement Learning}, author = {El-Agroudi, Tarek and Maurer, Finn Gross and Olsen, J\o{}rgen Anker and Alexis, Kostas}, booktitle = {Proceedings of The 8th Conference on Robot Learning}, pages = {5014--5029}, 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/el-agroudi25a/el-agroudi25a.pdf}, url = {https://proceedings.mlr.press/v270/el-agroudi25a.html}, abstract = {We present the development and real world demonstration of an in-flight attitude control law for a small low-cost quadruped with a five-bar-linkage leg design using only its legs as reaction masses. The control law is trained using deep reinforcement learning (DRL) and specifically through Proximal Policy Optimization (PPO) in the NVIDIA Omniverse Isaac Sim simulator with a GPU-accelerated DRL pipeline. To demonstrate the policy, a small quadruped is designed, constructed, and evaluated both on a rotating pole test setup and in free fall. During a free fall of 0.7 seconds, the quadruped follows commanded attitude steps of 45 degrees in all principal axes, and achieves an average base angular velocity of 110 degrees per second during large attitude reference steps.} }
Endnote
%0 Conference Paper %T In-Flight Attitude Control of a Quadruped using Deep Reinforcement Learning %A Tarek El-Agroudi %A Finn Gross Maurer %A Jørgen Anker Olsen %A Kostas Alexis %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-el-agroudi25a %I PMLR %P 5014--5029 %U https://proceedings.mlr.press/v270/el-agroudi25a.html %V 270 %X We present the development and real world demonstration of an in-flight attitude control law for a small low-cost quadruped with a five-bar-linkage leg design using only its legs as reaction masses. The control law is trained using deep reinforcement learning (DRL) and specifically through Proximal Policy Optimization (PPO) in the NVIDIA Omniverse Isaac Sim simulator with a GPU-accelerated DRL pipeline. To demonstrate the policy, a small quadruped is designed, constructed, and evaluated both on a rotating pole test setup and in free fall. During a free fall of 0.7 seconds, the quadruped follows commanded attitude steps of 45 degrees in all principal axes, and achieves an average base angular velocity of 110 degrees per second during large attitude reference steps.
APA
El-Agroudi, T., Maurer, F.G., Olsen, J.A. & Alexis, K.. (2025). In-Flight Attitude Control of a Quadruped using Deep Reinforcement Learning. Proceedings of The 8th Conference on Robot Learning, in Proceedings of Machine Learning Research 270:5014-5029 Available from https://proceedings.mlr.press/v270/el-agroudi25a.html.

Related Material