We have combined trajectory optimization and reinforcement learning to achieve versatile and robust perceptive legged locomotion.

Published in Science Robotics: https://www.science.org/doi/10.1126/s...

arXiv: https://doi.org/10.48550/arXiv.2309.1...

Abstract: Legged locomotion is a complex control problem that requires both accuracy and robustness to cope with realworld challenges. Legged systems have traditionally been controlled using trajectory optimization with inverse dynamics. Such hierarchical modelbased methods are appealing due to intuitive cost function tuning, accurate planning, generalization, and most importantly, the insightful understanding gained from more than one decade of extensive research. However, model mismatch and violation of assumptions are common sources of faulty operation. Simulationbased reinforcement learning, on the other hand, results in locomotion policies with unprecedented robustness and recovery skills.
Yet, all learning algorithms struggle with sparse rewards emerging from environments where valid footholds are rare, such as gaps or stepping stones. In this work, we propose a hybrid control architecture that combines the advantages of both worlds to simultaneously achieve greater robustness, footplacement accuracy, and terrain generalization. Our approach utilizes a modelbased planner to roll out a reference motion during training. A deep neural network policy is trained in simulation, aiming to track the optimized footholds. We evaluate the accuracy of our locomotion pipeline on sparse terrains, where pure datadriven methods are prone to fail. Furthermore, we demonstrate superior robustness in the presence of slippery or deformable ground when compared to modelbased counterparts. Finally, we show that our proposed tracking controller generalizes across different trajectory optimization methods not seen during training. In conclusion, our work unites the predictive capabilities and optimality guarantees of online planning with the inherent robustness attributed to offline learning.

Authors: Fabian Jenelten, Junzhe He, Farbod Farshidian, and Marco Hutter

Video: Fabian Jenelten

#ANYmal #leggedrobot #robot #robotics #robotdog #AI #reinforcementlearning #rl #rescue #innovation #armasuisse #arche2023 #scienceresearch #stepping