Automatic design of a least complicated hierarchical fuzzy system

The aim of this work is to present a particular design technique of hierarchical fuzzy controllers. The method makes an easy way to control complex systems with an automatic low complicated design. It was based on the systems capability to generate output informations when exited outside any control instance. This fact helps constructing input-output learning and testing data. This type of fuzzy controller is made optimal by the use of hierarchical minimal association of input rule bases, learned fuzzy inferences and the configuration of aggregation processes. The proposed strategy of design with learned fuzzy rule bases guarantee an interpretable behavioral decomposition. The effect of the growth of fuzzy rules number in terms of the algorithmic complexity has been studied. This algorithm was tested on the stabilization of an inverted simple pendulum around a desired position.