Research Of Evolutionary Multi-objective Optimization Algorithm Based On Immune Algorithm

Many problems are multi-objective optimization problems in reality, more and more experts and scholars use evolutionary algorithm to solve these problems and obtain good results. Because the evolutionary algorithm is a global search algorithm, it is poor in local search. The results of algorithm are difficult to converge evenly to global optimal solutions if using improper selection mechanism during the process of evolution. Immune algorithm simulates the self-regulation method of human immune system, it has the characteristics of antigen recognition, immune memory, immune regulation and immune clone selection, and it can provide good inspiration to evolutionary algorithm.IMM-NSGA-Ⅱalgorithm is proposed based on NSGA-Ⅱalgorithm combining with affinity calculation, immune memory and immune clone selection mechanisms. This algorithm has four innovations:firstly, it adds immune memory function to keep excellent solutions without losing, and improves convergence at the same time. Secondly, affinity calculation avoids the limitation of calculating crowding distance. Thirdly, it adds immune clone selection mechanism to enlarge search range of population, enhancing local search ability, and the mutation operation promotes diverse antibodies in this process. Finally, improved sorting strategy, arithmetic cross operator strategy, sorting rank according to the demand strategy and selecting strategy with the given threshold are proposed in order to improve the diversity and convergence of this algorithm.IMM-NSGA-Ⅱalgorithm is tested through typical benchmark functions and compared with NSGA-Ⅱalgorithm in diversity and convergence. The simulations prove that IMM-NSGA-Ⅱalgorithm has better performance than NSGA-Ⅱalgorithm in diversity, but it does not perform well enough in multimodal function and cheat function. Furthermore, IMM-NSGA-Ⅱalgorithm can converge to Pareto front quicker and better than NSGA-Ⅱalgorithm.