Iterative discrete forward-backward fuzzy predictive control

The main idea in this paper is to combine fuzzy controllers design with model predictive philosophy. In fact, this paper is an improvement of a previous work, where basic idea was to divide the initial optimization problem in a set of recursive optimization subproblems or decision stages. Each subproblem is raised as a fuzzy LQR design where the goal is to define the set of feedback gains of a fuzzy Parallel Distributed Compensator (PDC) that minimizes the function cost using Linear Matrix Inequalities (LMIs). Therefore, the global controller is a set of PDC controllers that satisfies the Bellman optimality principle, minimizing the cost function both locally and globally, and guarantees stability and satisfies the control action constraints. The new method described in this paper, which improves a previous work, reduces the upper bound of the cost function and, therefore, guarantees a better behavior of the controller from the optimization point of view.