Abstract

This paper addresses design of MPC controller for dynamic systems in the presence of disturbances. Model Predictive Control (MPC) based on nominal dynamic models cannot in general guarantee feasibility, in terms of constraint satisfaction, and stability of the system in the presence of disturbances. In this regard, it is common to use observers to estimate states and disturbances, allowing control algorithm to reject disturbances. However, it is important to guarantee feasibility and stability when using observer rather measurement. In this paper, we first provide conditions which guarantee feasibility and stability. Then, we demonstrate how one can extend model predictive control for systems with disturbances by applying full state and disturbance observer. To this aim, a two-stage approach is proposed where the MPC problem is decoupled from observer design procedure. The proposed approach is divided into two optimization problems. This simplifies design of robust model predictive control and improves the control performance (in some sense) compared to some existing robust model predictive control approaches.