A novel framework of model predictive control for controlling a class of nonlinear system in the presence of uncertainty

This article addresses a novel framework of a model predictive control algorithm with providing robustness property for a class of nonlinear system. This structure uses both past knowledge and weighted predicted information of the process which lead to solving two optimization problems. The first one is related to the model predictive control optimization problem so as to obtain the optimal control input, and the second one is linked with another simple optimization problem to achieve the optimal weighting coefficients. Because the parameter uncertainties exist in real processes because of the limited amount of data or variation of parameters over time and so on, the robust monotonic convergent of the proposed model predictive control against model uncertainty is investigated. To validate the effectiveness of this structure, a nonlinear system and coupled tank system as an experimental simulation are implemented. Moreover, the comparison between this novel structure of model predictive control and the typical model predictive control algorithm is made. It is shown that by adjusting the weighting factors and control horizons properly in the proposed strategy, more satisfactory performance of the output signal in the proposed model predictive control rather than the typical model predictive control algorithm is obtained.