Adaptive Neighborhood Grid-Boundary Multi-Objective Mayfly Algorithm for Improved Pareto Front Approximation

Multi-objective optimization algorithms are widely used for solving engineering design problems involving conflicting objectives. The Multi-Objective Mayfly Algorithm (MOMA) is a promising swarm-based optimizer, but its performance can be limited by fixed search parameters, premature convergence, weak archive distribution, and loss of extreme trade-off solutions during archive truncation. This paper proposes an Adaptive Neighborhood Grid-Boundary Multi-Objective Mayfly Algorithm (ANGB-MOMA) to improve the convergence and diversity behavior of the original MOMA. The proposed method introduces four mechanisms: adaptive parameter control, neighborhood-guided leader selection, grid-based archive and leader selection, and boundary-preserving archive truncation. The algorithm was evaluated using ZDT and DTLZ benchmark functions and compared with Original MOMA, Levy-flight MOMA, Stochastic-inertia MOMA, NSGA-II, MOPSO, and MOEA/D. The experimental study also included ablation analysis, parameter sensitivity analysis, Holm-corrected Wilcoxon statistical tests, metric reproducibility settings, and runtime assessment. The results show that ANGB-MOMA significantly improves the original MOMA and related MOMA variants on most tested functions. Strong improvements were observed on ZDT1, ZDT2, ZDT5, ZDT6, DTLZ4, and DTLZ7, with particularly large IGD+ reductions on ZDT5 and ZDT6. ANGB-MOMA also achieved the best overall performance on DTLZ4 among all compared algorithms. However, the comparison shows that the proposed method is not universally superior to established optimizers, as MOPSO and MOEA/D performed better on some discontinuous, discrete, and multimodal benchmark cases. The results therefore support ANGB-MOMA as a reproducible and effective improvement to the MOMA framework, while also identifying ZDT3 and ZDT4 as important limitation cases for future development.