A Strategy-Adaptive Surrogate Regulation Algorithm for Expensive Constrained Multi-Objective Optimization

Expensive constrained multi-objective optimization is common in industrial design, engineering system configuration, and complex decision-making. Since objective and constraint evaluations are expensive, existing surrogate-assisted methods still struggle to adaptively balance objective advancement and constraint satisfaction across different feasible-region structures and evolutionary stages. To address this issue, this paper proposes SASR-EA, which introduces an evolutionary state-responsive surrogate regulation mechanism to adaptively select surrogate modes according to population convergence, diversity, feasibility, and search stage, reducing the limitations of fixed rules and experience-based switching. It further develops a collaborative search framework that integrates objective-only optimization, objective optimization with overall constraint violation modeling, and objective optimization with key-constraint modeling, so that objective, overall constraint violation, and explicit constraint information contribute differently across stages. Experiments on the CF, MW, and DAS-CMOP benchmark suites show that SASR-EA achieves competitive performance in IGD, solution-set quality, diversity, and generalization, confirming the effectiveness of the proposed surrogate regulation mechanism and collaborative search framework.