随着海上船舶日益增多，海情急剧复杂化，及时准确地预测船舶的下一步动向成为海事监管的迫切需求。针对现有船舶轨迹预测算法提取轨迹特征能力较差、预测精度不高的问题，提出了添加Attention注意力机制的序列到序列船舶轨迹预测算法(sequence-to-sequence with attention，Seq2Seq-Att)。通过改进Seq2Seq的编码器结构和添加Attention机制，提高模型对轨迹特征的记忆能力，从而提升算法的预测精度。以东海海域的AIS数据为样本训练模型，预测船舶未来一段时间的经度、纬度、航速和航向。实验结果表明，相较于传统算法，该算法的预测精度更高，且均方根误差明显降低，可以为海事监管和智能航行提供依据。
Ship Trajectory Prediction Algorithm Based on Seq2Seq-Att
With the increasing number of ships at sea and the dramatic complexity of the sea situation,timely and accurate prediction of the next movement of ships has become an urgent need for maritime supervision.Aiming at the problem that the existing ship trajectory prediction algorithm has poor ability to extract trajectory features and low prediction accuracy,a sequence-to-sequence with attention(Seq2Seq-Att)ship trajectory prediction algorithm is proposed.By improving the encoder structure of Seq2Seq and adding the attention mechanism,the memory ability of the model on trajectory features is improved,so as to improve the prediction accuracy of the algorithm.The AIS data in the East China Sea is used as a sample to train the model to predict the longitude,latitude,speed and course of the ship in the future period.The experimental results show that compared with traditional algorithms,the prediction accuracy of the proposed algorithm is higher,and the root mean square error(RMSE)is significantly reduced,which can provide a basis for the maritime supervision and intelligent navigation.
trajectory predictionattention mechanismsequence to sequenceAIS data
国家科技期刊平台
NSTL
万方数据
