Fault-Aware Robust Control via Adversarial Reinforcement Learning
This work is significant for robotics engineers and researchers, as it provides a method to improve the robustness of robots against physical damage, enabling more reliable deployment in challenging environments.
This paper addresses the problem of robot fragility due to joint damage, which limits their real-world application. The authors propose an adversarial reinforcement learning framework that significantly increases robot robustness against joint damage in both manipulation and locomotion tasks, achieving exceeding success rates.
Robots have limited adaptation ability compared to humans and animals in the case of damage. However, robot damages are prevalent in real-world applications, especially for robots deployed in extreme environments. The fragility of robots greatly limits their widespread application. We propose an adversarial reinforcement learning framework, which significantly increases robot robustness over joint damage cases in both manipulation tasks and locomotion tasks. The agent is trained iteratively under the joint damage cases where it has poor performance. We validate our algorithm on a three-fingered robot hand and a quadruped robot. Our algorithm can be trained only in simulation and directly deployed on a real robot without any fine-tuning. It also demonstrates exceeding success rates over arbitrary joint damage cases.