Ship trajectory prediction based on transformer model

In view of the increasingly complex maritime traffic situation, in order to meet the demand of ship trajectory prediction accuracy, based on the large amount of ship trajectory data contained in the AIS(Automatic Identification System) and the encoder decoder construction mechanism of the transformer model, a transformer model is proposed to decode with the full connection layer instead of the decoder. According to the existing ship trajectory data, the model is trained to predict the future ship trajectory. The experimental results show that the error between the predicted trajectory information and the real trajectory information is small, which proves the effectiveness of the model.