l1鲁棒辨识:一种递推插值方法

l1 robust identification:a recursive interpolatory algorithm

中文关键词  鲁棒辨识  模型集  插值方法  递推算法

英文关键词  robust identification  model set  interpolatory algorithm  recursive algorithm

中文摘要

      针对l¹鲁棒辨识不能有效利用试验数据和进行在线辨识的问题, 提出了一种在线递推插值辨识方法. 用几何方法描述试验信息, 利用系统可行集与新的试验信息所构成的半空间的包含关系判断数据信息, 有效地利用了试验数据, 提高了辨识精度. 同时提出了一种新的计算辨识误差紧界的方法. 仿真结果表明了算法的有效性和可行性.

英文摘要

      This paper proposes an on-line recursive interpolatory algorithm for the l¹ robust identification problem. The new idea is to describe the experiment information and apirio information about system and noise by geometric appraoch. Then we can determine whether or not the new experiment data cantains innovations by the containing relation of model set determined by old data and the set formed from new data, meanwhile, the uncertainty of model set is reduced by effectively using the innovations contained in new data. Thus, the identification algorithm proposed can effectively use the experiment data and improves the identification precision. Finally analysis of the proposed recursive algorithm is carried out and a simulation example is presented to demonstrate the effectiveness of the algorithm.