Multi Objective Evolutionary Algorithm

In real life there are many optimization problems, which often is multi-objective innature, mutual restraint between the various objectives, in one of the objectives for theoptimization, is to reduce the other goals of performance. So overall, multi-objectiveoptimization problem does not exist in the case of a single optimal solution, but a set ofapproximate optimal compromise solution set. Traditional optimization algorithms can onlyget one run a compromise solution, which is low for multi-objective optimization problemsolving efficiency, can not meet the actual application requirements. Evolutionary units withpopulations of multi-objective evolutionary algorithm can get a group of a near optimalsolution, and the evolution of multiple individuals at the same time, the importance of a singleindividual can be reduced, thereby reducing the probability of falling into local optimum"trap". In this paper, multi-objective evolutionary algorithm in-depth research and analysis,main contents are as follows:1) NSGA-II (non dominated sorting genetic algorithm) has good time complexity, but at theexpense of the distributed algorithm as the cost, in keeping with its advantages at thesame time, in order to maintain the diversity of population, using the differentialevolution algorithm, add the difference of local search, the theoretical analysis andsimulation results show that the the time complexity of DELS_NSGA2algorithm withNSGA-II, and the solution set is more breadth and uniformity.2) Particle swarm optimization algorithm for solving multi-objective optimization problem,with fewer parameters, simple features. However, due to the particle swarm optimizationapplications in the field of multi-target relatively short time, the existing multi-objectiveparticle swarm optimization, while maintaining the diversity of the group and theconvergence of the solution in terms still to be improved. In this paper, the classicmulti-objective particle swarm algorithm has been improved by adding the individualperturbation vector particles dynamically select the optimal adjustment, optimal particleand new ways of handling the boundary with time-update mechanism. Comparative testsshow improved MOPSO-II effectively strengthen the population’s ability to escape fromlocal optima, and in terms of convergence and distribution have received greatlyimproved.