液压与气动2025，Vol.49Issue(11)：88-97,10.DOI:10.11832/j.issn.1000-4858.2025.11.010
液压位置伺服系统强化学习滑模控制研究
Research on Reinforcement Learning-based Sliding Mode Control for Hydraulic Position Servo Systems
摘要
Abstract
To address the limitations in control precision and robustness of hydraulic position servo systems,an intelligent adaptive control strategy combining the deep deterministic policy gradient algorithm with sliding mode control is proposed.A coupled electro-hydraulic asymmetric cylinder system model is established on the AMESim-Simulink platform,and the integration of the sliding mode control module with the reinforcement learning module is validated.The designed controller,combining deep deterministic policy gradient and sliding mode control,enables online self-tuning of sliding surface gains and chattering suppression factors.Simulation scenarios under three typical operating conditions-step input,sinusoidal input,and composite disturbances-are constructed.Results show that the proposed strategy achieves rise and settling times of 0.82 s and 0.83 s,respectively,in step tracking,outperforming radial basis function-based sliding mode control and conventional sliding mode control;under disturbance conditions,the maximum tracking error remains below 0.003 m,effectively suppressing system chattering.These findings demonstrate the proposed method's superior dynamic response and robustness in complex environments,providing significant implications for enhancing the intelligence and control performance of hydraulic servo systems.关键词
液压位置伺服系统/强化学习/位置跟踪控制/联合仿真Key words
hydraulic position servo system/reinforcement learning/position tracking control/co-simulation分类
机械工程引用本文复制引用
王天雷,王晨旭,辛增淼,贺跃帮,邱光繁,邓安安..液压位置伺服系统强化学习滑模控制研究[J].液压与气动,2025,49(11):88-97,10.基金项目
广西科技计划项目(2025GXNSFHA069191) （2025GXNSFHA069191）
江门市科技计划项目(2022JC01021) （2022JC01021）
