Generating a path-search graph based on ship-trajectory data: Route search via dynamic programming for autonomous ships

Voyage optimization is critical for voyage planning and an important approach for realizing the autonomous and green operation modes for ships. In recent years, voyage optimization (also referred to as weather routing) has been extensively investigated and commercial solutions have also been launched. It is used to find the shortest path (optimum path) for voyage planning, which can help reduce the operating costs of the ship. However, accurately estimating fuel costs and validating routes within an acceptable computational time remain challenging. In this study, we devised methods to generate a path-search graph based on ship-trajectory data and determined an optimal route based on dynamic programming. The trajectory data of the vessels were extracted from their automatic identification systems, and the errors contained in the data were discarded via preprocessing. The trajectory data were subsequently simplified using the Douglas-Peucker method, following which a path-search graph was created using Delaunay triangulation and the number of vertices were reduced using the quadtree method. The search graph was validated using a high-resolution bathymetric chart, and dynamic programming was adopted to determine the optimal route for minimizing fuel costs. Changes in the operational performance caused by meteorological conditions were calculated, and the fuel costs were estimated using the wind resistance coefficient and SHOPERA-NTUA-NTU-MARIC method. The effect on power was determined by the effect of current on the vessel speed; this methodology was simulated and tested on oil tankers, container ships, and car and gas carriers. The fuel saved for each vessel compared to that used for the shortest path was estimated when sailing the transpacific and transatlantic routes; a maximum fuel reduction of approximately 7.8% was estimated. The proposed methods can be applied to various ship types and maritime zones to reduce the computation time of path-search graph generation while considering the high-resolution map and accurately estimating the fuel-oil consumption.