Variational Autoencoder-Enhanced Data Augmentation Method for Ship-motion Trajectory Prediction

Accurate prediction of ship-motion trajectories is of critical significance in modern maritime affairs, directly impacting navigation safety, traffic control, and resource management. Deep neural networks (DNNs) have emerged as a powerful tool in this domain, yet their performance hinges on the availability of extensive and high-quality training data to mitigate overfitting. In this paper, we present a new data augmentation method for ship-motion trajectory prediction, especially delving into the potential of variational autoencoder as the core enhancement means. By efficiently optimizing the quality and expanding the quantity of automatic identification system (AIS) data, the overall data quality is enhanced, and then the effectiveness of DNNs in maritime applications is fully exerted. Experiment results demonstarte that our VAE - enhanced data augmentation method can effectively improve the prediction accuracy