概率预测强化学习下非结构环境机械臂变阻抗力跟踪控制  

Probability Prediction Reinforcement Learning for Variable Impedance Force Tracking Control of Robotic Arms in Unstructured Environments

作　　者：董梓呈 胡伟石 邵辉[1] 郭霖 DONG Zicheng;HU Weishi;SHAO Hui;GUO Lin(College of Information Science and Engneering,Huaqiao University,Xiamen 361021,China;Department of Laboratory and Device Management,Huaqiao University,Xiamen 361021,China)

机构地区：[1]华侨大学信息科学与工程学院,福建厦门361021 [2]华侨大学实验室与设备管理处,福建厦门361021

出　　处：《华侨大学学报（自然科学版）》2024年第4期461-470,共10页Journal of Huaqiao University(Natural Science)

基　　金：福建省自然科学基金资助项目(2021J01291);华侨大学研究生教育教学改革研究项目(22YJG006)。

摘　　要：针对非结构环境下末端实时移动机械臂阻抗控制力跟踪问题,通过动态调节阻尼系数以应对接触环境的不确定性。为确保阻抗策略的高效搜索,利用机械臂与接触环境交互产生状态-动作序列构建概率预测模型(PPM)。学习过程中,机械臂仅需与非结构接触环境进行少量交互即可获得最优变阻抗策略,这使得该过程在真实机械臂上直接训练成为可能。仿真实验表明,在几种非结构环境下,所提出的方法使力跟踪动态和稳态性能均明显优于传统阻抗控制和自适应变阻抗控制。Aiming at the real time impedance control force tracking problems of the end mobile robotic arm in a unstructured environment,the damping coefficient is dynamically adjusted to cope with the uncertainty of the contact environment.To ensure efficient search of the impedance strategy,a probabilistic prediction model(PPM)is constructed by utilizing the interaction between the robotic arm and the contact environment to generate state action sequences.During the learning process,the robotic arm only needs to interact minimally with the unstructured contact environment to obtain the optimal variable impedance strategy.This makes it possible to directly train the process on a real robotic arm.Simulation results show that in several unstructured environments,the proposed method significantly outperforms the traditional impedance control and adaptive variable impedance control in both dynamic and steady state force tracking performance.

关 键 词：变阻抗控制 机械臂力跟踪 强化学习 非结构环境 概率预测模型 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]