时变系统辨识方法及收敛定理_丁锋.pdfVIP

第5 卷第 1期 江南大学 学 报(自然 科学 版) Vol.5 No.1 2006年2 月 Journal of Southern Yangtze University(Natural Science Edition) Feb. 2006 : 1671- 7147( 2006)01- 0115- 12 时变系统辨识方法及其收敛定理 丁 锋 , 杨慧中, 纪志成 ( 江南大学 通信与控制工程学院, 江苏无锡214122) : 介绍了用于辨识方法性能研究的 收敛定理和 超收敛定理, 阐述了其应用范围;讨论了 研究辨识算法收敛性的各种激励条件; 综述了时变随机系统的各种辨识方法, 包括最小二乘类辨 识方法( 如遗忘因子最小二乘算法卡尔曼滤波算法有限数据窗最小二乘算法等)和随机梯度类 辨识方法(如遗忘梯度算法广义投影算法等); 同时阐述了时变参数系统辨识领域的一些值得深 入研究的课题; 最后给出了遗忘梯度算法在不同条件下参数估计误差上界的几 个定理, 说明数据 的平稳性可以改善参数估计精度. : 时变系统;辨识; 参数估计; 收 性能; 最小二乘; 随机梯度;鞅收 定理; 鞅超收 定理 : TP 273 : A TimeOVarying System Identification Methods and onvergence Theorems DING Feng, YANG HuiOzhong, JI ZhiOcheng (School of Communication and Control Engineering, Southern Yangtze University, Wuxi 214122, China) Abstract: In this paper, we introduce the martingale convergence theorem and martingale hyperconvergence theorem for analyzing performances of identification methods and states their application ranges. Thepaper also discusses excitation conditionsfor studying the convergenceof identification algorithms and surveys main identification approaches for timeOvarying systems, including least squares type algorithms ( e. g., the least squares algorithm with a forgetting factor, Kalman filtering algorithm, finiteOdataOwidow least squares algorithm) and stochastic gradient type algorithms ( e. g., the forgetting gradient algorithm and generalized projection algorithm). Furthermore, some further research topics for timeOvarying systems are presented. Finally, we give several convergence theorems of the forgetting gradient algorithm for timeO varying systems by using the stochastic process theory and martinga