A many-objective evolutionary algorithm with population preprocessing and projection distance-assisted elimination mechanism

Abstract Pareto dominance-based many-objective evolutionary algorithms (MaOEAs) face a significant challenge from many-objective problems (MaOPs). The selection pressure reduces as the number of objectives rises, while the non-dominated solution grows exponentially. Pareto dominance-based MaOEA increases the selection pressure by designing diversity-related environmental strategies. However, it still struggles to strike a good balance between population diversity and convergence. Moreover, the diversity-selection method increases the likelihood that dominance-resistant solutions (DRSs) will be chosen, which is detrimental to the performance of MaOEAs. To address the aforementioned problems, a many-objective optimization algorithm based on population preprocessing and projection distance-assisted elimination mechanism (PPEA) is proposed. In PPEA, first, the population preprocessing method is designed to lessen the negative impacts of DRSs. Second, to further improve the ability to balance population diversity and convergence of Pareto dominance-based MaOEAs, a projection distance-assisted elimination mechanism is proposed to remove the poorer individuals one by one until the population size satisfies the termination condition. The performance of PPEA was compared with seven excellent MaOEAs on a series of benchmark problems with 3–15 objectives and a real-world application problem. The experimental results indicate that PPEA is competitive and can effectively balance the diversity and convergence of the population when dealing with MaOPs.