Balance Reward and Safety Optimization for Safe Reinforcement Learning: A Perspective of Gradient Manipulation

TL;DR

This paper introduces a gradient manipulation approach to balance reward and safety in reinforcement learning, proposing a new optimization method and benchmark to improve safety without sacrificing reward performance.

Contribution

It presents a novel soft switching policy optimization method with convergence analysis and a new Safety-MuJoCo Benchmark for evaluating safe RL algorithms.

Findings

Our method outperforms state-of-the-art baselines in balancing reward and safety.

The proposed framework effectively manages the reward-safety trade-off.

Experimental results validate the theoretical advantages of gradient manipulation in safe RL.

Abstract

Ensuring the safety of Reinforcement Learning (RL) is crucial for its deployment in real-world applications. Nevertheless, managing the trade-off between reward and safety during exploration presents a significant challenge. Improving reward performance through policy adjustments may adversely affect safety performance. In this study, we aim to address this conflicting relation by leveraging the theory of gradient manipulation. Initially, we analyze the conflict between reward and safety gradients. Subsequently, we tackle the balance between reward and safety optimization by proposing a soft switching policy optimization method, for which we provide convergence analysis. Based on our theoretical examination, we provide a safe RL framework to overcome the aforementioned challenge, and we develop a Safety-MuJoCo Benchmark to assess the performance of safe RL algorithms. Finally, we evaluate the effectiveness of our method on the Safety-MuJoCo Benchmark and a popular safe RL benchmark, Omnisafe. Experimental results demonstrate that our algorithms outperform several state-of-the-art baselines in terms of balancing reward and safety optimization.