Ship trajectory prediction using AIS data with TransFormer-based AI

Ship trajectory prediction is important to maintain maritime travel and logistics for both security and efficiency. Nowadays we can access public data from Automatic Identification System(AIS), however, there have remained challenges to predicting precise trajectories by using AIS data due to its complexity and diversity. Recent significant developments in AI technology including generative AI, such as TransFormer-based algorithms, have made us realise that those technologies have the potential to produce more accurate trajectory prediction. This study aims to develop a method of trajectory prediction using AI from AIS data. We used the data of fishing vessels tracked in Japanese territorial waters of the Pacific Ocean, which includes latitude, longitude, ship over ground, and course over ground, and converted them into four-hot vectors to analyse the complex trajectories. The accuracy of the prediction was evaluated by accumulating the distance between the predicted value and the actual measurement value for each time. Our results showed that the TransFormer-based method can predict the fishing vessels’ trajectories which had been failed to predict by the LSTM-based method in previous studies. We conclude that the use of AI technology can contribute to the improvement of maritime traffic safety and sustainability, and could lead to more effective and versatile trajectory prediction algorithms.