Genetic algorithms and fuzzy situations for sequential optimization of control surfaces

We outline a new controller concept, which exploits the general structure of a control surface as if it is induced by a fuzzy controller. In addition to this we show how one can use a genetic algorithm to optimize the controller and we present the concept of fuzzy situations for a sequential optimization. Thus we consider in one optimization phase the control behavior belonging to only one starting condition. We obtain a controller optimized for this situation. After optimizing two situations we combine the two controllers by a fusion algorithm, which is based on the different activation degrees in the last test runs. Our approach leads to a quite good control behavior, also when the control task is very complex. We show some results on the simulation of the well-known cart-pole-problem