Deep bi-directional information-empowered ship trajectory prediction for maritime autonomous surface ships

It is critical to have accurate ship trajectory prediction for collision avoidance and intelligent traffic management of manned ships and emerging Maritime Autonomous Surface Ships (MASS). Deep learning methods for accurate prediction based on AIS data have emerged as a contemporary maritime transportation research focus. However, concerns about its accuracy and computational efficiency widely exist across both academic and industrial sectors, necessitating the discovery of new solutions. This paper aims to develop a new prediction approach called Deep Bi-Directional Information-Empowered (DBDIE) by utilising integrated multiple networks and an attention mechanism to address the above issues. The new DBDIE model extracts valuable features by fusing the Bi-directional Long Short-Term Memory (Bi-LSTM) and the Bi-directional Gated Recurrent Unit (Bi-GRU) neural networks. Additionally, the weights of the two bidirectional units are optimised using an attention mechanism, and the final prediction results are obtained through a weight self-adjustment mechanism. The effectiveness of the proposed model is verified through comprehensive comparisons with state-of-the-art deep learning methods, including Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Bi-LSTM, Bi-GRU, Sequence to Sequence (Seq2Seq), and Transformer neural networks. The experimental results demonstrate that the new DBDIE model achieves the most satisfactory prediction outcomes than all other classical methods, providing a new solution to improving the accuracy and effectiveness of predicting ship trajectories, which becomes increasingly important in the era of the safe navigation of mixed manned ships and MASS. As a result, the findings can aid the development and implementation of proactive preventive measures to avoid collisions, enhance maritime traffic management efficiency, and ensure maritime safety.