Development Of a Versatile and Fast Algorithm for The Optimal Ship Routing

Abstract

Maritime transport efficiency is a critical factor in global trade, yet traditional routing often relies on static pathfinding that ignores the dynamic impact of ocean weather on fuel consumption. This paper presents an automated framework for Optimal Ship Routing that integrates machine learning with heuristic search to minimize fuel usage. Our system utilizes a specialized data pipeline to synchronize AIS (Automatic Identification System) ship telemetry with high-resolution wave and wind data from the Open-Meteo API. To model ship performance, we implement an XGBoost Regressor trained on engineered features, including Speed Through Water (STW) and relative wind angles, achieving precise fuel-burn predictions. These predictions serve as the cost-weighting for an A Search Algorithm* operating on a spherical coordinate grid. A key feature of our implementation is the automated integration of a Natural Earth GeoJSON land mask, ensuring all generated paths are restricted to navigable waters. Experimental results demonstrate that by intelligently diverting from the Great Circle path to avoid high-resistance weather cells, the framework provides a scalable, data-driven approach to reducing operational costs and carbon emissions in maritime logistics.