Multi-model One Step Ahead Control for Nonlinear System

Two multi-model based onestep ahead predictive controller algorithms are proposed for nonlinear systems. The proposed design employ Constrained Kalman filter algorithm to interpolate the submodels/controllers and determine which model/controller pair to be used according to the operating region of the system. Thus, it maintain the performance of the controllers over a wide range of operating regions. The algorithms is independent of the submodels, hence can be applied regardless of the submodels development. Two benchmark system examples are studied to demonstrate the effectiveness of the proposed predictive controller design on tracking problems.