A Kind Of Fuzzy Genetic Algorithm Based On Population Diversity Measures

Genetic Algorithm (GA) is a kind of highly paralel, stochastic, global probability search algorithm based on the evolutionism such as natural selection, genetic crossover and gene mutation. GA is widely applied because of its simple structure, strong robustness and excellent capability of solving nonlnear function optimization compared to those traditional search algorithms. The GA behavior is determined by the exploitation and exploration relationship(EER) kept throughout the run. Many factors affect the HER, the most important ones are those parameters related to the crossover operator and the mutation operator. If these parameters are designed unreasonably, si outstanding problem appears: the premature convergence. Fuzzy theory simulates the visual thinking of human being.It expresses the procedure of human's thinking in very simple mathematical format and need not precise mathematical modelinglt is a sort of efficient mathematical tool and solves some difficult problems perfectly in the field of classical control theory.In the past few years, the combination of GA and fuzzy logic theory has become the research focus. On the one hand, GA can be used to process the fuzzy information in the imprecise circumstance. On the other hand, fuzzy logic is regard as tool to solve some problems concerning GA. They learn from others' strong points to offset one's weakness, therefore two research direction-genetic fuzzy system and fuzzy genetic algorithm(FGA)-appear.At present, the development of FGA is far from complete. There are all kinds of definitions aboil it and opinions vary. Majority think that FGA is the GA resulting from the integration that the use of fuzzy tools and fuzzy logio-based techniques for modeling different GA components and adapting GA control parameters, with the goal of improving performance. On the basis of former work, author presents a FGA based on the diversity measure. In order to keep the dynamic balance between EER, t'uz/y logic controller (FLC) was used to adjust GA's important parameters dynamicly.This dissertation defines two parameters based on the diversity measure as the input of the FLC which is used to control the GA's crossover probability and mutation probability. The experiments prove that FGA's convergence rate and quality are improved greatly.