Gait in Eight: Efficient On-Robot Learning for Omnidirectional Quadruped Locomotion
This work addresses the problem of real-time learning for quadruped robots, offering a more efficient approach compared to prior methods focused on simple forward gaits, though it appears incremental in extending to omnidirectional control.
The paper tackles the challenge of computational constraints in on-robot reinforcement learning for legged robots by presenting a framework that enables efficient learning of quadruped locomotion in just 8 minutes of real-time training, extending it to omnidirectional locomotion and demonstrating robustness in various environments.
On-robot Reinforcement Learning is a promising approach to train embodiment-aware policies for legged robots. However, the computational constraints of real-time learning on robots pose a significant challenge. We present a framework for efficiently learning quadruped locomotion in just 8 minutes of raw real-time training utilizing the sample efficiency and minimal computational overhead of the new off-policy algorithm CrossQ. We investigate two control architectures: Predicting joint target positions for agile, high-speed locomotion and Central Pattern Generators for stable, natural gaits. While prior work focused on learning simple forward gaits, our framework extends on-robot learning to omnidirectional locomotion. We demonstrate the robustness of our approach in different indoor and outdoor environments.