机器人关节摩擦建模与自适应RBF神经网络补偿计算力矩控制  

Friction modeling and computed torque control of robot jointswith adaptive RBF neural network compensation

作　　者：沈晓斌 王斌锐 余芮 崔小红 SHEN Xiaobin;WANG Binrui;YU Rui;CUI Xiaohong(College of Mechanical and Electrical Engineering,China Jiliang University,Hangzhou 310018,China)

机构地区：[1]中国计量大学机电工程学院,浙江杭州310018

出　　处：《中国计量大学学报》2020年第1期71-78,共8页Journal of China University of Metrology

基　　金：国家重点研发计划专项课题(2017YFC0804609)。

摘　　要：目的:补偿机器人关节摩擦力矩,提高关节轨迹跟踪效果。方法:以直流伺服电机和谐波减速器组成机器人关节为对象,基于Stribeck模型建立摩擦模型;采用最小二乘法和Matlab编程辨识模型参数;利用自适应RBF神经网络算法对系统模型参数误差引起的未建模动态进行在线观测;将观测结果补偿到计算力矩控制器中,建立含摩擦模型的自适应RBF神经网络补偿计算力矩控制器,推导出控制参数的约束条件。结果:基于自主设计的机器人关节进行对比实验。实验表明,含摩擦模型的自适应RBF神经网络补偿计算力矩控制追踪误差是无摩擦模型的自适应RBF神经网络补偿计算力矩控制追踪误差的55.12%,是传统计算力矩控制追踪误差的31.28%。结论:本文的控制方法提高了机器人关节轨迹跟踪精度。Aims:In order to compensate the joint frictional torque and improve the trajectory tracking effect of the robot joint.Methods:Taking the robot joint composed of DC servo motor and harmonic reducer as the object,the friction model is established based on Stribeck model;The least square method and MATLAB programming are used to identify the model parameters;The unmodeled dynamics caused by the parameter error of the system are observed online using the adaptive RBF neural network algorithm;The observation results are compensated to the calculated torque controllor,thus an adaptive RBF neural network compensated calculated torque controller with friction model is established and the constraint conditions of control parameters are derived.Results:Based on the self-designed robot joint,a comparative experiment is carried out.The experimental show that the tracking error of the adaptive RBF neural network compensation torque control with friction model is 55.12%of the tracking error of the adaptive RBF neural network compensation torque control without friction model,and is 31.28%of the tracking error of the traditional computed torque control.Conclusions:The control method proposed in this paper improves the trajectory tracking accuracy of the robot single joint.

关 键 词：机器人 Stribeck模型 计算力矩控制 RBF神经网络 自适应控制 轨迹跟踪 

分 类 号：TB242[一般工业技术—工程设计测绘]