Large-scale Multi-objective Evolutionary Algorithm Based on Dynamic Resource Allocation

In recent years, significant efforts have been dedicated to solving large-scale multi-objective optimization problems (LSMOPs) more efficiently. The large-scale multiobjective evolutionary algorithm based on decision variable analysis overlooks the contribution difference of the subcomponents to evolution, leading to inefficient computational resource allocation. Moreover, its single differential direction in the optimization operator struggles to balance convergence and diversity. To address these issues, a large-scale multi-objective evolutionary algorithm based on dynamic resource allocation (LMEA-DVA/DRA) is proposed, featuring: (1) an adaptive tridirectional differential evolution operator (ATDE) that maintains population diversity during subcomponent convergence; (2) a contribution-based resource allocation strategy to dynamically adjusts the optimization process of different subcomponents. Experimental results on the WFG test suits demonstrate that LMEA-DVA/DRA significantly outperforms the state-of-the-art algorithms (CCGDE3, MOEA-DVA, and DGEA).