ABSTRACT
This chapter focuses on the issue of prediction mismatch. It shows how the move to infinite horizon reduces the prediction mismatch and hence facilitates a stability result. Prediction mismatch is expected in the setup of finite horizon Model Predictive Control (MPC) algorithms such as generalised predictive control (GPC) and dynamic matrix control (DMC) and prevents straightforward stability proofs. Irrespective of the choice of K, the dual mode MPC algorithm is guaranteed stable for the nominal case. In essence, fixing the d.o.f. beyond some point in the horizon does not affect the stability proof so long as any prediction assumptions are reproducible at subsequent sampling instants, that is, inclusion of the tail. One of the major weaknesses of the original predictive control algorithms such as DMC, GPC, IDCOM is the prediction structure used in the optimisation of performance.
