Feasibility, optimality and computational complexity of Robust Model Predictive Control

For robust model predictive control (RMPC), many efforts have been taken on enlarging the region of attraction, enhancing the optimality and reducing the computational burden. For open-loop min-max MPC, it may be unable to enlarge the region of attraction by increasing the computational burden. Hence, an improvement of the initial optimization of the open-loop min-max MPC is specially given. By choosing the closed-loop optimal control, more on-line decision variables and larger optimization horizon, the feasibility and optimality can be improved but the computational burden may be intensified. However, promising strategies for improvement can be formulated by modifying the performance index, the type of invariant sets and control modes.