Robust model predictive control

Project members: H. Chen (Jilin University), C.W. Scherer

Model predictive control (MPC) is among the most successful regulation techniques in process industry. As the essential theoretical benefit, MPC achieves feedback-control action by solving open-loop optimization problems on-line. Industrial systems typically suffer from a large plant-model mismatch which can lead to considerable performance degradation in on-line optimization schemes if not taken into account. Unfortunately the systematic incorporation of (structured) uncertainties in the MPC framework is still rather immature. The main goal of this research is to understand how feedback-control action can be generated for uncertain models by solving minimax problems for open-loop strategies.

As an initial step we specifically consider the -disturbance attenuation problem with hard constraints on the control inputs. Based on game-theoretic techniques we investigate schemes that allow to relax or tighten performance specifications during on-line optimization in order to avoid input saturation while still guaranteeing desired performance specifications. Our ultimate goal is a non-conservative solution of the -synthesis problem with hard constraints.