Robust output feedback predictive controller with adaptive invariant tubes and observer gains

Abstract

This work aims to develop a robust output feedback predictive controller that can guarantee the recursive feasibility and the system stability as the invariant tubes and the observer gains can be updated at each sampling time. The recursive feasibility is guaranteed by using the invariant tubes with non-increasing size so the satisfactions of more tightened constraints ensure the satisfactions of less tightened constraints as time proceeds. The system stability is guaranteed by bounding the control error and the estimation error using the invariant tubes that can be updated at each sampling time. The online computational complexity can be reduced as the invariant tubes and the observer gains can be updated at each sampling time. This can be done by computing offline the invariant tubes for the estimation error, the invariant tubes for the control error and the observer gains. The smallest invariant tube containing the estimation error is determined at each sampling time and the corresponding observer gain is chosen as the real-time observer gain. The proposed algorithm can reduce the online computational complexity as compared with the case when the invariant tubes and the observer gains are computed online while the same level of control performance is obtained. In addition, the proposed algorithm can improve the control performance as compared with the case when the invariant tubes and the observer gains are constant.