增量自适应学习算法  

An incremental adaptive learning algorithms

作　　者：孙明轩[1] 徐晨晨 邹胜祥 Sun Mingxuan;Xu Chenchen;Zou Shengxiang(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》2020年第7期666-675,共10页Chinese High Technology Letters

基　　金：国家自然科学基金(61573320)资助项目。

摘　　要：提出适用于参数不确定系统的增量自适应学习算法,为连续时间自适应控制系统提供了一种新的设计方法。针对现有积分自适应算法,首先指出其近似实现方式,即实现时采用的离散算法与原算法存在着差别。对于非限幅增量自适应学习算法,证明了闭环系统收敛性,并给出其参数估值性质;为了保证参数估值本身有界,提出限幅增量自适应学习算法,借助类Barbalat引理,获得闭环系统的收敛性结果。与积分自适应系统相比较,所提出学习算法规避了积分自适应律在实现时离散化造成的近似问题,从而能够有效处理参数不确定性。针对运动控制系统设计了增量自适应鲁棒控制器,并用于电机位置跟踪,实验结果验证了所提控制方法的有效性。In this paper,an incremental adaptive learning algorithm for parametric uncertain systems is proposed,which puts forward a novel design methodology for continuous-time adaptive control systems.Firstly,the problem arisen from the implementation of the conventional integral adaptive system is pointed out,namely,the numerical integration procedure leading to the difference between the implemented and the original algorithms.The incremental adaptive learning algorithm is suggested to be used as an alternative,and for the unsaturated adaptation,the convergence of the closed-loop system and the estimation properties have been established.It has been shown that the saturation adaptation technique is able to assure bounded estimation,and the Barbalat-like lemma finalizes the analysis of convergence of the adaptive learning system undertaken.Compared with the integral adaptive system,the proposed learning algorithm avoids the problem arisen from the discretization for the integration,which can effectively deal with the parameter uncertainty.For the illustration,an incremental adaptive robust controller is designed for the uncertain motion systems,and a motor test bed is used for the verification of its feasibility and effectiveness.

关 键 词：自适应控制 自适应鲁棒控制 连续时间系统 

分 类 号：TP273.2[自动化与计算机技术—检测技术与自动化装置]