Enhance the Safety in Reinforcement Learning by ADRC Lagrangian Methods

TL;DR

This paper introduces ADRC-Lagrangian methods for safe reinforcement learning, significantly reducing safety violations and constraint violations by leveraging Active Disturbance Rejection Control to enhance robustness and stability.

Contribution

The paper proposes a novel ADRC-Lagrangian framework that improves safety in reinforcement learning by mitigating oscillations and violations, unifying classical and PID methods.

Findings

Safety violations reduced by up to 74%

Constraint violation magnitudes decreased by 89%

Average costs lowered by 67%

Abstract

Safe reinforcement learning (Safe RL) seeks to maximize rewards while satisfying safety constraints, typically addressed through Lagrangian-based methods. However, existing approaches, including PID and classical Lagrangian methods, suffer from oscillations and frequent safety violations due to parameter sensitivity and inherent phase lag. To address these limitations, we propose ADRC-Lagrangian methods that leverage Active Disturbance Rejection Control (ADRC) for enhanced robustness and reduced oscillations. Our unified framework encompasses classical and PID Lagrangian methods as special cases while significantly improving safety performance. Extensive experiments demonstrate that our approach reduces safety violations by up to 74%, constraint violation magnitudes by 89%, and average costs by 67\%, establishing superior effectiveness for Safe RL in complex environments.