A Real-Time Solution to Fuzzy Model Based Predictive Control

This paper addresses optimization issues in fuzzy model-based predictive control. The success of linear predictive control in controlling constrained linear processes is mainly due to the fact that the on-line optimization problem is a convex one, usually a quadratic programme. When the process model is nonlinear fuzzy one, non-convex time consuming optimization is necessary, with no guarantee of finding an optimal solution. To overcome this drawback, we propose a predictive control algorithm that employs linear parameter-varying models, obtained by local linearization of the fuzzy model along the output and input trajectories of the process. Different schemes for obtaining local linear models are discussed and compared using both simulations and real-time experiments.