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

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Chen, Donghe, Peng, Yubin, Zheng, Tengjie, Wang, Han, Qu, Chaoran, Cheng, Lin
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866915137140555776
author Chen, Donghe
Peng, Yubin
Zheng, Tengjie
Wang, Han
Qu, Chaoran
Cheng, Lin
author_facet Chen, Donghe
Peng, Yubin
Zheng, Tengjie
Wang, Han
Qu, Chaoran
Cheng, Lin
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2502_02265
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Adviser-Actor-Critic: Eliminating Steady-State Error in Reinforcement Learning Control
Chen, Donghe
Peng, Yubin
Zheng, Tengjie
Wang, Han
Qu, Chaoran
Cheng, Lin
Machine Learning
Artificial Intelligence
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.
title Adviser-Actor-Critic: Eliminating Steady-State Error in Reinforcement Learning Control
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2502.02265