Robust multi-model predictive control using LMIs

Abstract

This paper proposes a novel synthesis technique for robust predictive control of constrained nonlinear systems based on linear matrix inequalities (LMIs) formalism. Local discrete-time polytopic models are exploited for prediction of the system behavior. This design strategy can be applied to nonlinear systems provided a suitable embedding is available. The devised procedure guarantees constraint satisfaction and asymptotic stability. The proposed result extends previous works by handling less conservative input constraints and exploiting the different local descriptions of nonlinearity and uncertainty. The multi-model prediction together with the modified input constraints show significant improvements in terms of closed-loop performance and estimation of the feasibility domain.