An Enhanced Domination Based Evolutionary Algorithm for Multi-objective Problems

Abstract

We proposed a new evolutionary algorithm for multiobjective optimization problems. The influence of constraints on search space and Pareto front are analyzed first. According to the analysis, a new clustering method based on domination is proposed, in which the infeasible solutions are employed. Then, aiming to converge to Pareto fronts of the multiobjective problems quickly, a differential evolution based crossover operator is designed. In the designed crossover operator, uniform design method was used. At last, a square search method is employed to update the feasible nondominated solutions to improve the precision. Experiments on 10 selected test problems and comparisons with NSGA-II are made. Simulation results indicate that our proposal is effective and sound, and our proposal outperforms NSGA-II on the selected test problems.