Hybrid Dual-Population Artificial Physics Optimization for Complex Constrained Multi-Objective Scenarios

Solving constrained multi-objective optimization problems (CMOPs) poses significant challenges in balancing convergence, diversity, and feasibility. This paper introduces a novel Hybrid Dual-Population Artificial Physics Optimization (HDAPO) algorithm designed to effectively tackle CMOPs. The proposed approach employs a dual-population structure: a primary population aimed at promoting convergence and diversity, and a secondary population dedicated to constraint management. A dynamic epsilon-based constraint-handling strategy is integrated, which adaptively adjusts the feasibility threshold according to the proportion of feasible solutions. The optimization is guided by two collaborative mechanisms: the Exploration-Oriented Controller (EOC) operating on the primary population, and the Feasibility-Oriented Enhancer (FOE) managing the secondary population for refined exploitation. HDAPO is benchmarked using two constrained test suites i.e. LIR-CMOP and MW within PlatEMO platform. Comparative analysis with state-of-the-art algorithms confirms the superior effectiveness of HDAPO across a variety of test problems.