Ship weather routing framework based on dual-mode deep reinforcement learning

Ship navigation is strongly influenced by the ocean environment. An adverse sea environment not only increases fuel consumption but can also endanger navigation safety. Therefore, developing a weather routing framework that balances safety and economy is of practical importance. This study proposes a ship weather routing framework based on deep reinforcement learning (DRL) with two main objectives: avoiding extreme weather conditions such as typhoons and reducing fuel consumption. Wind, wave, and current reanalysis data are used to construct the environment under normal conditions, while a high-resolution wave model simulates typhoon wave fields to enhance data accuracy under extreme conditions. Dangerous waters are identified by considering factors such as parametric rolling and synchronous rolling. The actual ship speed is corrected based on environmental loads under a fixed power condition. The method utilizes the Dueling Double Deep Q-Network (D3QN) algorithm, with dual modes developed to navigate under normal and extreme weather. This method provides recommended routes and speeds, overcoming the limitations of traditional methods that neglect temporal variations in the ocean environment. Multiple case studies confirm that the proposed method remains reliable in both normal and extreme conditions, ensuring navigation safety and reducing fuel consumption, offering valuable support for ship operation optimization and maritime decarbonization.