基于在线学习的离散时间人机协作系统预定性能柔顺控制  

Discrete-time prescribed performance compliant control based on online-learning for human-robot collaboration system

作　　者：刘霞 王露 陈勇[2] LIU Xia;WANG Lu;CHEN Yong(School of Electrical Engineering&Electronic Information,Xihua University,Chengdu 610039,China;School of Automation Engineering,University of Electronic Science and Technology of China,Chengdu 611731,China)

机构地区：[1]西华大学电气与电子信息学院,成都610039 [2]电子科技大学自动化工程学院,成都611731

出　　处：《电子科技大学学报》2025年第1期52-61,共10页Journal of University of Electronic Science and Technology of China

基　　金：国家自然科学基金面上项目(61973257);四川省自然科学基金面上项目(2023NSFSC0510)。

摘　　要：为了使人机协作系统中机器人能够准确地顺应人类行为,提出了一种基于在线学习的离散时间预定性能柔顺控制方法。该方法在外环采用在线顺序极限学习机算法估计人类行为,并将估计结果结合参考阻抗模型来重建参考轨迹。在内环建立了离散时间预定性能控制器用于跟踪重建后的参考轨迹,并利用时间延迟估计来获得机器人复杂的未知动力学模型。分析了闭环系统的瞬态和稳态性能,通过对比仿真验证了该方法的有效性。所提的离散时间控制方法可更好地满足数字计算机的工作原理,在减少计算和内存负担的基础上,使得机器人末端执行器的跟踪误差能够满足预设性能要求。此外,该方法无需机器人精确的数学模型,同时还能减轻人类操作机器人的力量负担,保证人机协作的柔顺性。To enable robot compliant to human behavior in human-robot collaboration system accurately,a discrete-time prescribed performance compliant control based on online-learning is proposed.This method employs online sequential extreme learning machine in the outer loop to estimate the human behavior.The estimation results are combined with a reference impedance model to reconstruct the reference trajectory.A discrete-time prescribed performance controller is established in the inner loop to track the reconstructed reference trajectory,and time delay estimation is employed to obtain the unknown dynamics of the robot.The transient and steady performances of the closed-loop system are analyzed.The effectiveness of the proposed controller is verified by comparative simulations.The proposed discrete-time control method can better satisfy the working principle of digital computers,and it can make the tracking error of the end-effector meet the prescribed performance with less computation and memory burden.In addition,the proposed method does not require the accurate mode of the robot,reduces the force burden of human operating the robot,and guarantees the compliance to the human behavior.

关 键 词：柔顺控制 离散时间人机协作系统 人类行为估计 在线顺序极限学习机 预定性能 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]