基于网络逆系统的迭代学习初始控制量确定  

Initial Control Value Determination of Iterative Learning Based on Network Inverse System

作　　者：刘娣[1] 李宏胜[1] 林健[1] 

机构地区：[1]南京工程学院自动化学院,南京211167

出　　处：《系统仿真学报》2013年第5期1041-1045,1052,共6页Journal of System Simulation

基　　金：江苏省自然科学基金(BK2009350);南京工程学院引进人才科研启动基金(YKJ201012)

摘　　要：鉴于初始控制量对迭代学习控制(ILC)算法收敛速度及跟踪精度的重要影响,为保证ILC算法对任意期望轨迹的跟踪性能,提出了一种基于小波神经网络(WNN)逆系统的ILC初始控制量确定方法。首先分析了研究对象的可逆性,在此基础上建立了对象的WNN逆模型,然后根据该逆模型求得任意期望轨迹下的网络输出,将其作为ILC算法的理想初始控制量进行迭代学习。仿真结果表明,新算法辅助的ILC能利用先前的控制经验,在面临新的期望轨迹时能有效减少迭代次数,提高跟踪精度。In view of the important influence of initial control value to convergence rate and tracking accuracy of iterative learning control （ILC）, in order to ensure tracking performance of lLC for every desired trajectory, the initial control value determination method based on wavelet neural network （WNN） inverse system was proposed. Firstly, the reversibility analysis of research object was analyzed. On this basis, a WNN inverse system model of research object was established. Then, the network outputs were obtained by the inverse model for any desired trajectory which used as the ideal initial control value of 1LC algorithm. Simulation results show that the new algorithm assisted ILC can use previous control experience, and thus reduce iteration times and improve tracking precision efficiency when facing new desired trajectory.

关 键 词：迭代学习控制 初始控制量 小波神经网络 逆系统 

分 类 号：TP13[自动化与计算机技术—控制理论与控制工程]