Research On Evolutionary Computation-based Multi-Objective Optimization Algorithms

Multi-objective optimization problems are widespread in scientific research and production applications,which have become more and more complex.These objectives are interrelated and conflicting,and it is difficult to find a solution that satisfies all objectives achieving optimal values at the same time.Evolutionary algorithms belong to the intelligent optimization algorithms based on population iterative optimization,which can obtain a set of non-dominated solutions in a single run.They are simple and have few control parameters,therefore,evolutionary algorithms have been widely used in solving multi-objective optimization problems in the past two decades.In this paper,focusing on the slow convergence speed and poor diversity of multi-objective evolutionary algorithms when solving multi-objective problems,three kinds of improved multi-objective evolutionary algorithms are proposed.The main research results and innovations of this thesis are as follows:1.Aiming at the shortcomings of evolutionary algorithm in solving multi-objective optimization problems with slow convergence speed and poor diversity,a multi-objective differential evolution algorithm with improved ranking-based mutation(MODE-IMR)is proposed.Based on the ranking mutation strategy,the proposed algorithm uses the optimal individual in the population as the base vector in the mutation process,and uses the strategy of opposition-based parameter controlling to mutate the individuals in the population.Furthermore,the improved crowding distance calculation formula is used to select the next evolutionary population.The simulation experiments are conducted on standard multi-objective optimization problems including ZDT1~ZDT4,ZDT6 and DTLZ6~DTLZ7.MODE-IMR is better than other comparison algorithms in terms of convergence and diversity.2.In order to further improve the optimizing performance of multi-objective evolutionary algorithms,a multi-objective differential evolution algorithm with fuzzy adaptive ranking-based mutation(MODE-FARM)is proposed.a fuzzy system-based indicator feedback strategy is incorporated into the ranking-based mutation process to balance the exploration and exploitation of the algorithm.In addition,the diversity of population in the evolutionary process is improved by adopting an improved population initialization method and a dual population-based selection strategy.The simulation experiments are conducted on 7 multi-objective optimization problems including ZDT and DTLZ,the results show that the proposed algorithm can converge to the true Pareto front better than other algorithms,and the effectiveness of fuzzy adaptive ranking-based mutation strategy is validated.3.The many-objective optimization problems with more than 3 objectives have become a research hotspot in the field of multi-objective optimization.In order to solve the many-objective optimization problems with irregular Pareto front,a many-objective optimization algorithm based on dynamic resource allocation(MaOEA-DRA)is proposed.The proposed algorithm uses the Tchebycheff method to decompose the objective function into some sub-objective functions,and uses the dynamic resource allocation method to allocate computing resources according to the different shapes of the Pareto front.In addition,the method of shift-based density estimation is used to rank the population.The simulation experiment results on 11 problems with irregular true Pareto front show that the proposed algorithm is better than other comparison algorithms in terms of convergence and diversity,and the effectiveness of the Tchebycheff method is validated.