A New Multi-objective Optimization Evolutionary Algorithm Based on Geometrical Pareto Selection and Double Neighbored Crossover

Multi-objective Optimization Evolutionary algorithm (MOEA) is an effective method to solve Multi-objective Optimization Problem. Currently, most of MOEAs have room for improvement in terms of the number of approximated Pareto front points, the approximation to the true Pareto front, the uniformity of the distribution of approximated Pareto front points and the complete of coverage. Here, a new MOEA named DNGPS which is combined with multi-subpopulation strategy, double neighbored crossover operator and a fast archiving algorithm named Geometrical Pareto Selection (GPS) is proposed. In this paper, nine widely used test problems are employed to test DNGPS's performance and experimental results show that DNGPS can reduce the expense on archiving, at the same time, can obtain enough approximated Pareto front points, improve the approximation, the uniformity, the complete of coverage, are better than the comparative algorithms such as SPEA, NSGA, SPEA2 and NSGAII.