Fault-Aware Robust Control via Adversarial Reinforcement Learning
arXiv:2011.08728
Abstract
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.
arXiv comment7 pages, 7 figures
References in corpus (9)
- Continuous control with deep reinforcement learning
- Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor
- Robots that can adapt like animals
- Solving Rubik's Cube with a Robot Hand
- Robust Adversarial Reinforcement Learning
- Robust Deep Reinforcement Learning with Adversarial Attacks
- Reset-free Trial-and-Error Learning for Robot Damage Recovery
- Action Robust Reinforcement Learning and Applications in Continuous Control
- Fast Adaptation via Policy-Dynamics Value Functions