A method is disclosed for arranging or adapting a fuzzy controller or of a system of interlinked fuzzy controllers having at least two input variables and an output variable which controller or system has input membership functions with which truth values of linguistic values for the input variables are determined, and has a control mechanism in which truth values for the output variable are determined from the truth values for the input variable and has output membership functions with which the values of the output variable are determined from the truth values for the output variable. Pursuant to the method, value ranges are defined for the input variables and a number of classes are determined over the value ranges. Subsequently, in a learning phase, value combinations of the input variables and of the output variable are recorded data set by data set. The defined classes are altered, removed and/or new classes are generated as a function of the registered value combinations. Following the completion of the class adaptation, adapted membership functions and the rules of the fuzzy controller are set up on the basis of the newly formed classes.
展开▼