1996 Volume 8 Issue 6 Pages 1058-1065
Although genetic algorithms (GAs) are more and more employed as optimization tools of fuzzy logic based systems, the user is faced with the problem of selecting the right parameters of the GAs, such as the population size, the generation gap, the crossover and the mutation rate and so on. Unfortunately, the selection of those parameters require experience and knowledge. Moreover, it is evident, that those parameters are to change dynamically in accordance to the optimization stage. For example, in the beginning a large population is needed, while later near to convergence, most of the population member will be similiar, thus only wasting computing time. The idea proposed in this work is therefore a knowledge-based approachfordynamically adaptation of the GAs. The knowledge is expressed in form of fuzzy logic IF-THEN rules. It will be shown, that such a knowledge-based GAs is superior to a static GAs. Also, in the final stage of the optimization, we optimize the defuzzification.