A Novel Archive Maintenance for Adapting Weight Vectors in Decomposition-based Multi-objective Evolutionary Algorithms

Abstract

This paper proposes a novel archive maintenance for adapting weight vectors to improve the performance of the decomposition-based evolutionary algorithms for multi- and many-objective optimization problems with different Pareto front shapes (called AMAWV). AMAWV adopts a novel archive maintenance strategy for avoiding the dominance resistant solutions, as well as retaining the good diversity of non-dominated solution set. In addition, guided from the information of the archive, an adaptive weight vector method is designed to solve problems with various Pareto fronts. The proposed algorithm is compared with state-of-the-art algorithms on a number of test problems with different Pareto front shapes (the simplex-like, the inverted, the disconnected, the degenerated, the scaled, the mixed, the high dimensional). The experimental results have shown the superiority and versatility of the proposed algorithm.