六轴机械臂神经网络自适应终端滑模控制  

Adaptive-Neural-Network-Based Terminal Sliding Mode Control for Six-Axis Robotic Manipulators

作　　者：贾华 刘延俊[1,2,3,4] 王雨 薛钢[2,3,4] JIA Hua;LIU Yanjun;WANG Yu;XUE Gang(School of Mechanical Engineering,Shandong University,Jinan 250061,China;Key Laboratory of High-Efficiency and Clean Mechanical Manufacture of Ministry of Education,Shandong University,Jinan 250061,China;National Demonstration Center for Experimental Mechanical Engineering Education,Shandong University,Jinan 250061,China;Institute of Marine Science and Technology,Shandong University,Qingdao,Shandong 266237,China)

机构地区：[1]山东大学机械工程学院,济南250061 [2]山东大学高效洁净机械制造教育部重点实验室,济南250061 [3]山东大学机械工程国家级实验教学示范中心,济南250061 [4]山东大学海洋研究院,山东青岛266237

出　　处：《西安交通大学学报》2022年第11期21-30,共10页Journal of Xi'an Jiaotong University

基　　金：国家自然科学基金资助项目(52001186);山东省自然科学基金资助项目(ZR2020QE292)。

摘　　要：考虑模型不确定且存在外界干扰时六轴机械臂的轨迹跟踪控制问题,以快速、稳定地跟踪轨迹规划生成的各关节期望轨迹为控制目标,提出了一种基于机械臂动力学模型分块逼近的神经网络自适应终端滑模控制方法。为加快跟踪误差的收敛速度,避免传统终端滑模中存在的奇异值问题,采用了一种非奇异终端滑模面。针对系统模型不确定的情况,利用多组RBF神经网络分块逼近动力学模型参数,通过自适应律在线调整权值,实现模型的重构,并设计鲁棒项消除模型重构误差。在Simulink中开展仿真分析,结果表明:与RBF神经网络整体逼近算法相比,本文提出的控制策略可以使六轴机械臂的关节最大稳态误差减少83.7%;当末端加入时变的负载后,关节最大稳态误差减少了85.6%,具有抵抗末端负载变化的能力。本文为六轴机械臂提供了一种有效、可行的轨迹跟踪控制方法。This paper proposes an adaptive-neural-network-based terminal sliding mode control algorithm based on partitional approximation of dynamic models of robot manipulators to realize tracking of the desired trajectory of each joint generated by trajectory planning of six-axis robot manipulators when models are unknown and there is external disturbance.In addition,a non-singular terminal sliding mode manifold is proposed to accelerate the convergence of the tracking error and avoid the singular problem in the control with the conventional terminal sliding mode.Since the system model is unknown,multi-group RBF neural networks are utilized to approximate the dynamic model parameters,and the model is reconstructed by an adaptive weight update law.The model reconstruction errors are compensated by the designed robust terms.Simulation analysis is carried out in Simulink.According to the analysis results,compared with the RBF neural network global approximation algorithm,the proposed control method can reduce the maximum steady-state error of the manipulator joints by 83.7%.When the time-varying load is added to the end effector,the maximum steady-state error of the joints is reduced by 85.6%.Therefore,the proposed control is an effective and feasible trajectory tracking method for the six-axis robot manipulator under the condition of load variation.

关 键 词：六轴机械臂 非奇异终端滑模 动力学模型 RBF神经网络 轨迹跟踪 

分 类 号：TP241[自动化与计算机技术—检测技术与自动化装置]