Efficient Adversarial Training without Attacking: Worst-Case-Aware Robust Reinforcement LearningDownload PDF

Published: 31 Oct 2022, Last Modified: 12 Mar 2024NeurIPS 2022 AcceptReaders: Everyone
Keywords: Reinforcement Learning, Robustness, Worst-case Aware, Adversarial Learning
TL;DR: We propose a strong and efficient robust training framework for RL, WocaR-RL, that directly estimates and optimizes the worst-case reward of a policy under bounded $\ell_p$ attacks without requiring extra samples for learning an attacker.
Abstract: Recent studies reveal that a well-trained deep reinforcement learning (RL) policy can be particularly vulnerable to adversarial perturbations on input observations. Therefore, it is crucial to train RL agents that are robust against any attacks with a bounded budget. Existing robust training methods in deep RL either treat correlated steps separately, ignoring the robustness of long-term rewards, or train the agents and RL-based attacker together, doubling the computational burden and sample complexity of the training process. In this work, we propose a strong and efficient robust training framework for RL, named Worst-case-aware Robust RL (WocaR-RL) that directly estimates and optimizes the worst-case reward of a policy under bounded l_p attacks without requiring extra samples for learning an attacker. Experiments on multiple environments show that WocaR-RL achieves state-of-the-art performance under various strong attacks, and obtains significantly higher training efficiency than prior state-of-the-art robust training methods. The code of this work is available at https://github.com/umd-huang-lab/WocaR-RL.
Supplementary Material: zip
Community Implementations: [![CatalyzeX](/images/catalyzex_icon.svg) 1 code implementation](https://www.catalyzex.com/paper/arxiv:2210.05927/code)
18 Replies

Loading