Abstract
Design and analysis of robust control systems, especially the H∞ control, have been studied actively and many application examples have been reported in various areas. However, in order to apply the robust control theory to a real plant, it is necessary to evaluate model uncertainty or unmodeled dynamics. Therefore, research on robust identification which means system identification for robust control has become one of the hottest topics in system identification. Roughly speaking, there are two types of robust identification methods, that is, deterministic one and stochastic one. In both methods, some a priori information about model uncertainty is assumed, and sometimes unrealistic assumptions are included. In this paper, a large space structure (LSS) is chosen as an identification object. On the assumption that physical structure of an LSS is known a priori, a new practical robust identification method based on the least-squares estimation in cooperation with decimation is proposed. Effectiveness of the method is examined through experimental data.
