MOEA/D with Archive-Based Adaptive Reference Vector Regeneration

Multi-objective evolutionary algorithms (MOEAs) have demonstrated remarkable efficacy in solving multi-objective optimization problems. However, when addressing high-dimensional objectives and irregular Pareto fronts, conventional approaches often struggle to balance the trade-off between convergence and diversity. To mitigate this challenge, this paper proposes an archive-guided adaptive reference vector regeneration-based MOEA/D (MOEAD-ARVR). The proposed method integrates archive information of infeasible solutions within the MOEA/D framework and employs adaptive crossover operations, dynamic neighbourhood size adjustment, and an adaptive reference vector update mechanism to enhance both convergence and diversity. Experimental evaluations across various benchmark problems demonstrate that MOEAD-ARVR consistently outperforms existing approaches in achieving well-distributed and convergent solutions.