Risk-Averse Model Uncertainty for Distributionally Robust Safe Reinforcement Learning

TL;DR

This paper introduces a risk-averse deep reinforcement learning framework that handles model uncertainty without minimax optimization, ensuring robust and safe performance in uncertain environments.

Contribution

It proposes a novel distributionally robust safe reinforcement learning method using coherent distortion risk measures, avoiding minimax optimization for efficiency.

Findings

Demonstrates robustness and safety in continuous control tasks under environment perturbations.

Provides theoretical guarantees of robustness through equivalence to distributionally robust problems.

Achieves efficient, model-free implementation using standard data collection.

Abstract

Many real-world domains require safe decision making in uncertain environments. In this work, we introduce a deep reinforcement learning framework for approaching this important problem. We consider a distribution over transition models, and apply a risk-averse perspective towards model uncertainty through the use of coherent distortion risk measures. We provide robustness guarantees for this framework by showing it is equivalent to a specific class of distributionally robust safe reinforcement learning problems. Unlike existing approaches to robustness in deep reinforcement learning, however, our formulation does not involve minimax optimization. This leads to an efficient, model-free implementation of our approach that only requires standard data collection from a single training environment. In experiments on continuous control tasks with safety constraints, we demonstrate that our framework produces robust performance and safety at deployment time across a range of perturbed test environments.