一类推广的两步鲁棒辨识算法.PDF

第17 卷第5 期 2002 年9 月 Vol. 17 No.5 Control and Decision 　Sept . 2002 : 1001-0920(2002) 05-0575-04 莫建林, 张卫东, 许晓鸣 ( , 200030) : 基于传统的两步辨识方法, 提出一类当频域数据样本在系统频率 上非均匀分布时的H 鲁棒 ∞ 辨识方法。分析了相应的最差情况的辨识误差, 并给出了采用Lidstone 插值样条算子和三角窗函数时 显式最差情况辨识误差上界。 : 鲁棒辨识; 鲁棒收敛; 插值样条; 窗函数 : TP 13: A A class of extended two step robust identification MO Jian-lin, ZH A N G Wei-dong , X U Xiao-ming ( epartment of Automation Shanghai Jiaotong University Shanghai, , 200030, China) A bstract: Based on the traditional two step robust identification, a class of robust identification algorithms are proposed for the system in which frequency data are sampled in unequally distributed . - , frequency intervals T he worst case identification error upper bound is analyzed and an explict error upper bound is obtained when the Lidstone interpolation spline and triangular window are used. Key words : robust identification; robust convergency; interpolation spline; window functions 1引　　言 , [ 11, 12] ; H H 1( FFT) ,2 NehariHankel , () [ 1, 2, 5] , , , , H [ 14] , [ 13] [5, 6] ,H ,