H-infinity theory has been widely accepted for robust control system design to cope with the inaccurate description of the mathematical models. However, there exist very few identification techniques which are compatible with current robust control design methodologies. The problem of system identification in H-infinity is thus formulated to obtain an identified system model with explicit upper bound on identification error in H-infinity norm based on experimental information (noisy data). This proposal aims to develop control oriented identification techniques in H-infinity framework for linear time-invariant systems. The proposed research will focus on following issues: 1. Establish practical noise models which fit the H-infinity framework; 2. Investigate the role of linear algorithms in system identification; 3. Develop robustly convergent identification algorithms for different noise models with priori upper bounds for identification error in H-infinity norm; 4. Optimize the performance of robustly convergent identification algorithms; 5. Study system identification in H-infinity based on experimental noisy time response data. The ultimate goal of this proposal is to update identification techniques so that they are compatible with modern robust control methodologies.
