Improving robust model predictive control via dynamic output feedback

Abstract

This paper considers robust model predictive control for systems with polytopic uncertainty and bounded disturbance, where the system state is unmeasurable, and the model at the current sampling time is an exact combination of the vertices of the polytope. A parameter-dependent dynamic output feedback is used for this problem. At each sampling time, the optimization problems can be solved via LMI techniques. By specifying quadratic boundedness, the closed-loop system is guaranteed to converge to a neighborhood of the origin. The primary contribution is the separation of the step for handling estimation error constraint in another always feasible optimization problem, being solved after the main optimization is performed. Thus, the recursive feasibility of the main optimization can be retrieved in a better manner. A numerical example is given to illustrate the effectiveness of the controller.