Effcient robust output feedback MPC

This paper extends an effcient robust Model Predictive Control (MPC) methodology based on offline optimization of prediction dynamics in two respects. Firstly the case of output feedback via an observer is considered. Secondly, a method is proposed for imposing a bound on worst case performance using dynamic feedback which allows a nominal optimal feedback law to be implemented when this is feasible. The method is applicable to the case of uncertain systems which are not robustly stabilized by nominal LQ-optimal control. An effcient online optimization with feasibility and stability guarantees is proposed.