References & Citations
Computer Science > Machine Learning
Title: Game-Theoretic Robust Reinforcement Learning Handles Temporally-Coupled Perturbations
(Submitted on 22 Jul 2023 (v1), last revised 25 Apr 2024 (this version, v3))
Abstract: Deploying reinforcement learning (RL) systems requires robustness to uncertainty and model misspecification, yet prior robust RL methods typically only study noise introduced independently across time. However, practical sources of uncertainty are usually coupled across time. We formally introduce temporally-coupled perturbations, presenting a novel challenge for existing robust RL methods. To tackle this challenge, we propose GRAD, a novel game-theoretic approach that treats the temporally-coupled robust RL problem as a partially observable two-player zero-sum game. By finding an approximate equilibrium within this game, GRAD optimizes for general robustness against temporally-coupled perturbations. Experiments on continuous control tasks demonstrate that, compared with prior methods, our approach achieves a higher degree of robustness to various types of attacks on different attack domains, both in settings with temporally-coupled perturbations and decoupled perturbations.
Submission history
From: Yongyuan Liang [view email][v1] Sat, 22 Jul 2023 12:10:04 GMT (8382kb,D)
[v2] Tue, 9 Apr 2024 07:18:56 GMT (6330kb,D)
[v3] Thu, 25 Apr 2024 04:07:20 GMT (6330kb,D)
Link back to: arXiv, form interface, contact.