H∞ Identification with Truncated Model
Kouji TSUMURA, Kenichi HATSUNE, Yoshio SAITO
In this paper we study identification of discrete time systems using normalized coprime factorization. Robust identifications which agree with robust control theory have been actively investigated in frequency domain and time domain. We have proposed several new identification methods using normalized coprime factorization for time domain identification problems, in which the evaluated functions are given to have relation to the graph topology. In these methods, unknown terms of the identified rational functions, which are caused by I/O data finiteness, are optimally interpolated so that the evaluated functions are minimized. On the other hand, we can show other methods in which the unknown coefficients are approximately equal to 0. According to this idea, we propose some new identification methods using truncated model in this paper. Their algorithms contain iteration of some processes, to solve a linear matrix inequality or the 2-block problem of H∞-norm condition. We also give numerical examples and discuss the effectiveness and the problems of these methods.