Ship Trajectory Prediction Based on Attention in Bidirectional Recurrent Neural Networks

Ship Trajectory Prediction Based on Attention in Bidirectional Recurrent Neural Networks

연구 분야: Artificial Intelligence

학회: 2020 5th International Conference on Information Science, Computer Technology and Transportation (ISCTT)

Using AIS data to further improve the accuracy of ship trajectory prediction, a model based on Attention in Bidirectional Long Short- Term Memory Recurrent Neural Networks (BLSTM) is proposed. The model learns from AIS data in a certain area over a while. Final model performance comparing the learning results of the four Recurrent Neural Network models on the same data set, let them make track predictions on the same AIS data, and proved that the model has higher prediction accuracy. The prediction results can provide a reference for ship traffic organization and management in the detection of abnormal ship behavior, early warning of ship collision or grounding, etc.