Game-Theoretic Robust Reinforcement Learning Handles Temporally-Coupled Perturbations

TL;DR

This paper introduces GRAD, a game-theoretic method for robust reinforcement learning that effectively handles temporally-coupled uncertainties, outperforming prior methods in robustness across various attack scenarios.

Contribution

The paper presents a novel game-theoretic framework, GRAD, that addresses the challenge of temporally-coupled perturbations in robust RL, which was not previously tackled.

Findings

GRAD achieves higher robustness against various attacks.

It outperforms prior methods in continuous control tasks.

Effective against both temporally-coupled and decoupled perturbations.

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.