一类MIMO非线性系统的自适应神经网络鲁棒跟踪控制  

ADAPTIVE ROBUST TRACKING CONTROL BASED ON NEURAL NETWORKFOR A CLASS OF MIMO NONLINEAR SYSTEMS

作　　者：刘德玉[1,2] 邹坤霖 贺力克 Liu Deyu;Zou Kunlin;He Like(Hunan Industry Polytechnic,Changsha 410208,Hunan,China;Hunan University National Engineering Training Room for Robot Vision Perception and Control Technology,Changsha 410082,Hunan,China)

机构地区：[1]湖南工业职业技术学院,湖南长沙410208 [2]湖南大学机器人视觉感知与控制技术国家工程实训室,湖南长沙410082

出　　处：《计算机应用与软件》2022年第11期66-72,共7页Computer Applications and Software

基　　金：湖南省自然科学基金项目(2018JJ5017)。

摘　　要：针对一类具有不确定性和外部干扰的非线性MIMO系统,提出一种基于神经网络的鲁棒二次镇定反馈混合控制方案。该混合方案将神经网络与自适应鲁棒补偿器相结合,实现了滑模控制和H_(∞)最优控制。采用神经网络逼近不确定非线性系统中未知的理想反馈控制律,设计滑模控制器以消除神经网络的重构误差和逼近误差,并采用H_(∞)鲁棒控制器实现给定的跟踪性能。采用二次镇定方法有效降低传统鲁棒控制方法的保守性。理论证明了闭环系统内所有信号是有界的,采用所提出的智能混合控制器保证了给定的H_(∞)性能。仿真结果表明,与其他类似方案相比,所提出的智能鲁棒二次镇定反馈混合控制器具有更好的性能。A robust quadratic stabilization feedback hybrid control scheme based on neural network is proposed for controlling a class of nonlinear MIMO systems with uncertainties and external disturbances.The hybrid scheme combined neural network with an adaptive robust compensator,where sliding model control and H_(∞)optimal control methods were conducted.The neural network was used to approximate the unknown ideal feedback control law of uncertain nonlinear system,the sliding model controllers were designed to eliminate the reconstruction errors and approximation errors of the neural network,and H_(∞)controller was performed to achieve a certain robust tracking performance.A quadratic stability approach was used to reduce the conservatism of the conventional robust control approach.It was proven that the signals in closed-loop were bounded and a specified H_(∞)tracking performance was ensured by using the proposed intelligent hybrid controller.Simulation results show that the proposed intelligent hybrid controller performs better in comparison to other similar schemes.

关 键 词：神经网络 滑模控制 H_(∞)性能 二次镇定 非线性系统 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术] TP273