Affinity Propagation Based Multiobjective Evolutionary Algorithm And Its Application

Evolutionary algorithm is a major approach to deal with the multiobjective optimization problems(MOPs) at present. The multiobjective evolutionary algorithm(MOEA) has become an important research topic in the field of computational intelligence. A MOEA is mainly consisted of two parts, i.e., variation operators(mating selection, recombination, mutation, etc.) and environmental selection operator. However, until now, in the contributions on MOEAs, most of the attention is paid on the second operator, and the works on variation schemes are much less. Especially, the research on the maing selection of MOEA is insufficient. Therefore, based on the affinity propagation(AP) clustering technique, this thesis studies the mating selection scheme of MOEA to improve the performance of the algorithm.At first, the research actuality of MOEA is analyzed, the shortages existing in the study on MOEA are summarized, and the motivation of this thesis is presented. In addition, the basic knowledge on genetic algorithm and MOEA is also stated.After that, the detailed procedures of AP approach is describled. Based on the AP approach, an AP based mating restriction strategy is designed, and an AP based multiobjective optimization algorithm(APMO) is proposed. In the APMO, at each generation, the AP is first used to cluster the population, then with a certain probability, the parents are only allowed to be selected from the similar solutions in the same cluster. Meanwhile, in order to adapt the variation of the balance between exploration and exploitation in the evolution of the algorithm, a probability update scheme of mating restriction is developed based on the strength Pareto dominance.At last, a test suite with complex Pareto sets and Pareto fronts is employed to test the APMO. Three state-of-the-art MOEAs, i.e., MOEA/D-DE, NSGA-II and SPEA2, are utilized to compare the performancce. Experimental results show that, while dealing with the test suites, APMO performs the best compared with the comparison algorithms.. APMO is also taken to optimize the aiming points of the missiles to test its efficiency on practical application. The optimization results indicate that APMO is promising to deal with this problem.