Enhancing Multi-vessel Trajectory Prediction Using a Multiscale Hypergraph Neural Network: Capturing Complex Maritime Interactions

Marine transportation constitutes a vital component in the operation of the global economy. Recent global crises have profoundly disrupted maritime transportation, escalating costs, reducing connectivity, and intensifying global economic and supply chain challenges. The accurate prediction of vessel trajectories is crucial for optimizing operational efficiency, enhancing safety measures, and minimizing maritime costs. In this study, the application of a multiscale hypergraph neural network to multi-vessel trajectory prediction is proposed, aiming to model complex interactions between vessels at both the pair-wise and group levels. This model enhances prediction accuracy by capturing maritime-specific relational dynamics, which can improve navigational safety and efficiency. Compared to Long Short-Term Memory (LSTM) and Conditional Variational Autoencoder (CVAE), the model demonstrates superior path prediction performance, achieving the best results in terms of a number of key metrics. This showcases its strong potential for applications in multi-vessel trajectory prediction.