一类严反馈非线性系统的神经网络串级控制及应用  

Neural networks cascade control of a class of nonlinear systems with its applications

作　　者：王良勇[1] 

机构地区：[1]东北大学流程工业综合自动化国家重点实验室,辽宁沈阳110819

出　　处：《电机与控制学报》2013年第12期106-112,共7页Electric Machines and Control

基　　金：国家自然科学基金(61304035);辽宁省博士启动基金(20121011);中央高校基本科研业务费专向资金资助项目(N110308001)

摘　　要：针对一类结构和参数未知的严反馈多输入多输出非线性系统,提出一种基于神经网络的串级控制方法,该方法采用基于Backstepping(反步设计)的思想设计串级控制器,综合基于Backstepping的非线性控制方法控制精确度高以及串级控制结构简单的优势。采用神经网络逼近结构和参数未知的非线性系统,利用Lyapunov稳定性定理推导出神经网络的调节律。运用Lyapunov稳定性定理证明了闭环系统的所有信号均为半全局一致最终有界。为了验证本文方法的有效性,首先进行数值仿真实验,并进一步将方法应用于机械手系统,进行物理实验研究。数值实验和物理实验结果证明了所提方法的有效性。A class of strict feedback multi input multi output nonlinear systems with parameters and struc- ture unknown are considered in this paper. Neural networks based cascade control scheme was proposed to control the mentioned systems. In this control scheme, backstepping based method was adopted to de- sign this cascade controller which integtated the advantages of the good performance of backstepping based nonlinear control and simple structure of the cascade control. As parameters and structure of the nonlinear system are unkown, neural networks was introduced to approximate the nonlinear system. And the adap- tive law of the neural network was designed by Lyapunov method. Then semi global unform boundedness were analyzed based on Lyapunov method. Numerical simulations were also conducted. Finally, this pro- posed method was applied on a robotic manipulator, and physical experiments were completed. Results of numerical simulations and physical experiments demonstrate the effectiveness of the proposed method.

关 键 词：非线性系统 串级控制 反步设计 神经网络 机械手 

分 类 号：TP241[自动化与计算机技术—检测技术与自动化装置]