A unified framework for optimality analysis of model predictive control

As well-known, model predictive control is closely related to optimal control. This paper studies relationships between them and provides a unified framework for optimality analysis of model predictive controllers (MPC). The optimality is evaluated by comparing total performance of MPC with finite and infinite horizon optimal cost. Based on relaxed value iteration method, upper and lower bounds of optimality evaluation functions are expressed explicitly in terms of optimization horizon. These results reveal detailed characteristics on performance of closed-loop MPC systems due to using “receding horizon optimization” implementation style.