Adaptive decision alternatives in fuzzy predictive control

In fuzzy predictive control, goals and constraints are described by fuzzy sets, and techniques from fuzzy multicriteria decision making are applied to find the optimal control actions. Unfortunately the resulting optimization problem is nonconvex. By discretizing the control actions, the search space for the optimal solution is limited and branch-and-bound optimization can be used. The discretization, however introduces a tradeoff between the number of discrete actions and the performance. A possible solution to this problem is proposed in this paper where an adaptive set of discrete control alternatives based on the fulfilment of fuzzy criteria is introduced. The adaptation is performed by a scaling factor multiplied by a dynamic set of control actions. By using this approach, the number of alternatives is kept low, while performance is increased. The proposed method is applied to temperature control