Research On Multimodal Multi-objective Optimization Method Based On Particle Swarm Optimization Algorithm

Evolutionary computation is widely used in multi-objective optimization problems.In many practical multi-objective optimization problems,there are often multiple equivalent optimal solution sets,which can map to the optimal frontier of the objective function,and this type of problem is called the multimodal multi-objective optimization problem.Searching for multiple equivalent optimal solution sets in the decision space can not only help to analyze the nature of the problem,but also provide decision-makers with a wider range of solutions.Hence,it is very vital to have a research on multimodal multiobjective optimization methods.In this thesis,optimization method of multimodal multiobjective optimization is studied based particle swarm optimization,the main work and research results are as follows:The selection process of particle swarm optimization algorithms is an important part of evolution,determining which individuals can participate in the evolution of the next generation.And the quality of the selection operator will have a great influence on the optimization results.In this thesis,the selection operator of the algorithm is studied,and the weighted distance measurement method based on harmonic mean distance is proposed,which is first used to calculate the harmonic mean distance between the points in the decision space and the target space,and then the two spatial distance values are weighted.Compared with the existing crowded distance method,the validity of the density measurement method is verified.In order to meet the demand of the algorithm for diversity and the change of exploration and development ability in different periods,a dynamic topological particle swarm optimization framework is designed based on the analysis of the niche particle swarm which can dynamically adjust the neighborhood by neighborhood switching coefficient and specify the size of neighborhood,and then dynamically adjust the exploration and development ability of particle swarm.And it is applied to solve multimodal problems.The simulation result shows that the proposed algorithm framework has significant advantages in solving multimodal multi-objective optimization problems.In order to further expand the idea of solving multimodal problems,the multisubpopulation particle swarm optimization is introduced,and the advantages of multigroup optimization in solving complex problems are analyzed.Affinity propagation clustering is used to classify the population,so that the subpopulation is evenly distributed in the decision space.Topological neighborhood is used in the subpopulation to promote the information exchange and shared among particles within the subpopulation.The global optimal is used to replace the internal neighborhood optimal of the subpopulation called Probabilistic boot update,which facilitates the transmission of information between subpopulations.The simulation results show that the proposed multi-population strategy can be used to solve multimodal multi-objective optimization problems,and it has certain advantages to obtain a series of equivalent Pareto optimal solution sets.