Fuzzy Hybrid Systems modeling with application in decision making and control

Abstract

Hybrid Systems are systems containing both discrete event and continuous variable components. Many recent contributions address crisp situations, where ambiguity or subjectivity in the measured data is absent. In this paper, we propose Fuzzy Hybrid Systems to account for inaccurate measurements and uncertain dynamics. We present a strategy to determine the most appropriate control actions in a sampled data setting. The proposed approach is based on three basic steps that are performed in each sampling period. First, the current discrete fuzzy state of the system is determined by a sensor evaluation. Next, the future discrete fuzzy state is predicted for the possible control actions and the best action, in respect to desired continuous states, is selected. Finally, the decision is cross-evaluated by a limited horizon prediction of the continuous system variables. The proposed method is explained and demonstrated for a variation of the a well-known two-tank scenario.