柔性关节机械臂的自适应命令滤波输出反馈控制  

Adaptive Command Filtered Output Feedback Control for Flexible Joint Manipulator

作　　者：潘昌忠[1] 费湘尹 周兰[1] 熊培银[1] 李智靖 PAN Changzhong;FEI Xiangyin;ZHOU Lan;XIONG Peiyin;LI Zhijing(School of Information and Electrical Engineering, Hunan University of Science and Technology, Xiangtan, Hunan 411100, China)

机构地区：[1]湖南科技大学信息与电气工程学院,湖南湘潭411100

出　　处：《西安交通大学学报》2022年第5期199-208,共10页Journal of Xi'an Jiaotong University

基　　金：国家自然科学基金资助项目(62173138);湖南省自然科学基金资助项目(2019JJ50157);湖南省教育厅科学研究重点资助项目(20A186)。

摘　　要：针对柔性关节机械臂轨迹跟踪控制中存在模型摄动、外界干扰以及部分状态信息不可测等问题,提出一种基于自适应神经网络观测器的命令滤波输出反馈控制方法。首先,给出含不确定性的柔性关节机械臂动力学方程,并选用RBF神经网络设计自适应律,在线逼近由模型摄动及未知干扰带来的不确定性项;然后,设计自适应神经观测器以对系统不可测状态进行实时估计,并将估计值用于反馈控制,解决部分状态信息不可测的问题;最后,基于Lyapunov理论设计反步跟踪控制器,并引入二阶命令滤波器获得中间虚拟控制量的导数,避免反步递推过程中“计算爆炸”的问题。仿真结果表明,当系统存在模型摄动和外界干扰时,所提自适应命令滤波反步控制方法与传统反步控制方法和误差补偿命令滤波控制方法相比,无需连杆的角速度以及电机轴的转角和角速度等状态信息,且轨迹跟踪精度分别提升78%和35%。To address such problems as model perturbation,external disturbance and state undetectability in the trajectory tracking control of a flexible joint manipulator(FJM),a command filtered output feedback control method based on adaptive neural network observer is proposed.Firstly,the dynamic equation of the FJM with uncertainties is given,and adaptive laws are designed by using RBF neural networks to approximate online the uncertainties caused by model perturbations and unknown disturbances.Then,an adaptive neural observer is designed to estimate undetectable states in a real time manner,and the estimated values are used for feedback controltosolvetheproblemofundetectability.Finally,abackstepping tracking controllerisdesigned basedonLyapunov theory,and second-order command filters are introduced to obtain the derivatives of the intermediate virtual control laws,where the problem of“calculation explosion”in the backstepping recursive process is avoided.Simulations are conducted for the FJM in the presences of model perturbations and external disturbances,and the results show that compared with the traditional backstepping control method and the command-filtered method with error compensation,the proposed method does not require some state information like the angular velocity of the link and the angle and the angular velocity of the motor shaft.Moreover,the trajectory tracking accuracy is improved by 78%and 35%respectively.

关 键 词：柔性关节机械臂 RBF神经网络 自适应观测器 命令滤波 反步法 

分 类 号：TP24[自动化与计算机技术—检测技术与自动化装置]