Adviser-Actor-Critic: Eliminating Steady-State Error in Reinforcement Learning Control

TL;DR

This paper introduces Adviser-Actor-Critic (AAC), a reinforcement learning method that combines feedback control with adaptive learning to improve precision and eliminate steady-state error in control tasks, especially in robotics.

Contribution

AAC innovatively integrates feedback control with reinforcement learning, using an Adviser to mentor the actor for enhanced goal precision and robustness.

Findings

AAC outperforms standard RL algorithms in precision-critical tasks

Demonstrates high accuracy and robustness in goal attainment

Effective in robotics and real-world applications

Abstract

High-precision control tasks present substantial challenges for reinforcement learning (RL) algorithms, frequently resulting in suboptimal performance attributed to network approximation inaccuracies and inadequate sample quality.These issues are exacerbated when the task requires the agent to achieve a precise goal state, as is common in robotics and other real-world applications.We introduce Adviser-Actor-Critic (AAC), designed to address the precision control dilemma by combining the precision of feedback control theory with the adaptive learning capability of RL and featuring an Adviser that mentors the actor to refine control actions, thereby enhancing the precision of goal attainment.Finally, through benchmark tests, AAC outperformed standard RL algorithms in precision-critical, goal-conditioned tasks, demonstrating AAC's high precision, reliability, and robustness.Code are available at: https://anonymous.4open.science/r/Adviser-Actor-Critic-8AC5.