Study On Dynamic Multi-objective Optimization Evolutionary Algorithm

There are many dynamic multi-objective optimization problems in real life.Therefore,the study of dynamic multi-objective optimization problems has important application value.Due to the characteristics of dynamic multi-objective optimization problems,soving such problems becomes very difficult.Therefore,the study of dynamic multi-objective optimization has important theoretical value.Because of the outstanding performance of evolutionary algorithm in multi-objective optimization problems,the application of evolutionary algorithm in solving dynamic multi-objective optimization problems has attracted the attention of many scholars.In the process of dealing with dynamic multi-objective optimization problems,traditional evolutionary algorithms will encounter two difficult problems.One difficulty is how to enhance the search ability for the new environment,and the other is how to improve the convergence speed of the algorithm on the basis of maintaining the ability to search for the new environment.These two problems are considered as major challenges for evolutionary algorithms to solve dynamic multi-objective optimization problems.In this paper,around dynamic multi-objective optimization,dynamic multi-objective optimization test function and dynamic multi-objective optimization evolutionary algorithm are studied.Firstly,a set of new dynamic multi-objective optimization test function sets is designed.By studying the existing test function sets,the paper finds that the Pareto-optimal solutions shape changes affect the difficulty of dynamic multi-objective test functions.Based on the different moving modes of Pareto-optimal solutions and the different shape changes of Pareto-optimal solutions,the paper designs a set of test functions which combines Pareto-optimal solutions movement with Pareto-optimal solutions shape change.The simulation results show that the test functions are good for testing the algorithm performance.Then,an evolutionary algorithm based on new dynamic strategy is designed to deal with dynamic multi-objective optimization problems.The dynamic strategy consists of restart strategy and adjustment strategy.The restart strategy is,after the environmental changes,to predict the possible moving direction of Pareto-optimal solutions in the new environment by using the information of a small amount of new environment,and to reinitialize the population with the estimated direction and local search to make it close to the Pareto-optimal solutions of the new environment.This strategy is conducive to the rapid response of the algorithm to changes in the environment.The adjustment strategy is to adjust the current population after obtaining more new environmental information.So that the current population owes more individuals which are close to the new Pareto-optimal solutions,and the algorithm converges faster to the new Pareto-optimal solutions.The simulation results show that,in dynamic multi-objective optimization problems,the algorithm is very competitive in comparison with the existing algorithms.Finally,a hybrid immigrants strategy is proposed to deal with dynamic multi-objective optimization problems whose environmental changes cannot be detected.The hybrid immigrants strategy contains two strategies.One is the random immigrants strategy,which helps to maintain population diversity.The other is the immigrants strategy based on information guided.This strategy takes the difference between the optimal solutions of the previous generation and that of current generation as the guided direction,and produces some better solutions,thus improving the convergence rate.The algorithm proposed in this paper is tested on some different test functions.The empirical results show that the proposed algorithm is efficient and superior in maintaining diversity and tracking Pareto-optimal front.