Physics-informed multi-objective optimization for fuel consumption and air-pollutant exposure in ship operations

Maritime transportation poses severe public health risks in major port cities. Conventional strategies often rely on blanket speed reductions to lower total emission volumes, which can compromise logistics productivity and fail to mitigate peak pollutant concentrations driven by local meteorology. To address this, we propose a physics-informed multi-objective optimization framework that leverages the temporal dimension of navigation — adjusting speed profiles based on real-time local wind conditions — to simultaneously reduce fuel consumption and peak pollutant exposure in coastal zones. The framework first reconstructs high-resolution flow fields from sparse sensor data using a physics-informed deep operator network, then predicts downwind pollutant dispersion in real time, and finally derives Pareto-optimal route-and-speed policies through multi-objective Bayesian optimization. Validated on unsteady laminar flow scenarios, including the Busan Port vicinity, the method achieves 20%–35% hypervolume improvement and 34%–78% reductions in peak exposure compared to scenario-specific baselines. These results demonstrate that timing navigation to exploit favorable meteorological windows — rather than imposing uniform speed limits — can substantially reduce peak exposure without sacrificing operational efficiency.