一类非参数不确定系统的自适应神经网络重复学习控制  

Neural-network-based adaptive repetitive learning control for a class of non-parametric uncertain systems

作　　者：许昌源 谢树宗 陈强[1] XU Changyuan;XIE Shuzong;CHEN Qiang(College of Information Engineering,Zhejiang University of Technology,Hangzhou 310023)

机构地区：[1]浙江工业大学信息工程学院,杭州310023

出　　处：《高技术通讯》2022年第8期859-865,共7页Chinese High Technology Letters

基　　金：国家自然科学基金面上项目(62222315,61973274);浙江省自然科学基金(LZ22F030007);高端装备先进感知与智能控制教育部重点实验室开放课题(GDSC202010)资助项目。

摘　　要：本文针对一类非参数不确定系统提出一种自适应神经网络重复学习控制方法。利用期望轨迹的周期特性,构造周期性期望控制输入,并设计重复学习律估计未知期望控制输入并补偿系统周期不确定,实现系统对期望轨迹的高精度跟踪。在此基础上,利用神经网络估计系统未知状态和补偿非周期性不确定,进而提高系统鲁棒性。与已有的部分限幅学习律相比,本文提出的全限幅重复学习律可以保证估计值的连续性且能够被限制在指定的界内。最后,基于Lyapunov方法分析误差的收敛性能,并给出仿真结果验证了本文所提方法的有效性。An adaptive neural network repetitive learning control method is proposed for a class of non-parametric uncertain systems. A periodic desired control input is constructed by using the period characteristics of the desired trajectory,and a repetitive learning law is designed to estimate the unknown desired control input and compensate periodic uncertainty of the system, such that the high-precision tracking of the desired trajectory is achieved. On this basis,the neural network is employed to estimate the unknown state of the system and compensate for non-periodic uncertainties, such that the robustness of the whole system can be enhanced. Compared with the existing partial saturated learning law, the full saturated repetitive learning law proposed in this paper can ensure that the estimation is continuous and constrained within a prescribed region. Finally, the error convergence performance is analyzed based on the Lyapunov approach, and simulation results are given to verify the effectiveness of the proposed scheme.

关 键 词：重复学习控制 非参数不确定系统 神经网络 LYAPUNOV方法 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP13[自动化与计算机技术—控制科学与工程]