Robust Design of Predictive Controllers in Presence of Unmodeled Dynamics*

Abstract

In this paper a predictive controller is designed in face of modeling errors. In particular, the robustness characteristics of the proposed control law will be investigated with respect to dynamic model uncertainty. Indeed, Model Predictive Controllers are widely used in the control of complex systems, which hardly can be exactly represented by a model of known structure. For the design procedure, we assume that the system is described by an uncertainty Model Set made up by a nominal approximate model and a bound on the unmodeled dynamics. Assuming that the plant S is stable we will compute a predictive control law guaranteeing satisfaction of hard input constraints and zero regulation properties despite the presence of unmodeled dynamics. The key feature of the proposed approach is that it relies on plant output measurements avoiding the use of the real system state that, in many practical cases, are not available and even may be infinite dimensional.