An Integrated Constrained Multi-Objective Evolutionary Algorithm with Feasibility-Driven Repair and Adaptive Parameter Control for Narrow-Band Optimization

Constrained multi-objective optimization (CMOP) is especially difficult when the feasible region is very narrow. In this study, we introduce Integrated-CMOEA, a clear and structured framework that uses structure-aware seeding, a projection-based repair operator, dual-population evolution, adaptive parameter control, and reference vector archiving. For the DC2-DTLZ1 problem, the repair step is handled as a continuous one-dimensional root-finding problem along a feasible search ray. This method provides clear rules for restoring feasibility when a valid bracket is found. Our results show that the method quickly finds and maintains strict feasibility and produces a well-distributed set of solutions near the constrained Pareto front. In tests with five independent runs, Integrated-CMOEA outperformed four other CMOEAs in both IGD and hypervolume. An ablation study shows that deterministic repair is the main reason for its strong performance on this narrow-band benchmark. Integrated-CMOEA is a reliable framework for analytically structured narrow-band CMOPs, though it has some limits when applied more broadly.