A Bi-population Multi-objective Algorithm for Continuous Multi-objective Optimization Problem

摘要

The decomposition based multi-objective evolution algorithm (MOEA) had been successfully applied into solving many kinds of multi-objective optimization problems. MOEA/D is the most classical MOEA based on decomposition approach, which has received an increasing interesting from computational intelligent in the past few years. In the MOEA/D, a multi-objective optimization problem would be decomposed into a set of scalar single-objective subproblems and then the evolutionary algorithm is utilized to address these subproblems simultaneously. In order to improve the quality and distribution of the nondominated solutions, the differential evolution is used to evolve the external archive based on the Pareto rule, where the global and local evolution is designed to improve the quality and distribution, adaptively. A set of experiments are carried out to investigate the strength and weakness of our proposed algorithm on a series of benchmark test problems in comparison with the original MOEA/D.