A Multi-population Evolutionary Algorithm for Multi-scenario Multi-objective Optimization Problems

The multi-scenario multi-objective optimization problem (MSMOP) is a new paradigm that is widely used in engineering due to its ability to provide a more comprehensive evaluation perspective. However, there are existing methods that pose challenges in finding a set of well-distributed and well-converged common compromise solutions. To address this issue, this paper develops a method for searching scenario-based Pareto optimal solutions. Firstly, we propose a combinatorial framework that co-evolves multiple populations to push the population toward the global optimal solutions, where each scenario employs a multi-objective evolutionary algorithm (MOEA) as an optimizer. Next, we present an adaptive decomposition-based MOEA with a two-layer selection strategy to derive different optimizers. The first layer selector focuses on convergence, and the second layer selector emphasizes diversity. Finally, extensive testing problems are used to validate the efficiency and effectiveness of the proposed framework. The results demonstrate that the proposed algorithm outperforms existing state-of-the-art algorithms, thereby offering a practical tool for solving MSMOPs.