Dynamic Priority-Guided Two-Stage Multi-Objective Optimization for UAV–Ship Collaborative Maritime Search and Rescue under Storm Uncertainty

Maritime search and rescue (SAR) operations in extreme weather, such as typhoons, face tremendous challenges due to the high dynamic uncertainty of the environment and the real-time kinematic drifting of targets. Traditional single-platform dispatching is prone to inefficiencies and low coverage. To address these issues, this paper proposes an enhanced NSGA-III-based two-stage adaptive multi-objective optimization framework for unmanned aerial vehicle (UAV) and ship collaborative SAR. In the first stage, a multi-UAV dynamic detection model is constructed using an Archimedean spiral search anchored to multiple drifting storm centers, converting physical drifts into dynamic distance-based priority weights. In the second stage, a mixed-integer nonlinear programming (MINLP) model is formulated for ship scheduling to concurrently optimize three conflicting objectives: minimizing priority-weighted access time, minimizing total path distance, and maximizing high-priority coverage. To overcome Pareto front degradation in high-dimensional dynamic environments, an enhanced adaptive NSGA-III is designed. It integrates K-Means-based adaptive reference point updating, differential evolution (DE) mutation operators, and a hypervolume-based early stopping mechanism. Extensive Monte Carlo simulations verify the robustness of the framework, demonstrating the average improved in high-priority coverage. Furthermore, a real-world case study based on Typhoon Doksuri in the South China Sea confirms that the proposed framework significantly outperforms baseline methods in convergence, diversity, and decision-making efficiency, providing a robust algorithmic tool for modern maritime emergency command systems.