基于CNN-BiGRU-ATTENTION的船舶航迹预测方法研究

Vessel Trajectory Prediction Method Based on CNN-BiGRU-ATTENTION

为了维护海上交通安全,实时掌握船舶航行动态,国际海事组织(International Maritime Organization,IMO)要求船舶必须配备船舶自动识别系统(Automatic Identification System,AIS),而AIS数据包含了船舶的许多航行特征,可以预测船舶的未来走向,便于海上交管部门的管理。根据船舶AIS数据,提出了基于船舶航迹聚类和CNN-BiGRU-ATTENTION模型的船舶航迹预测方法,构建了基于卷积神经网络(CNN)、双向门控循环单元(BiGRU)和注意力机制(ATTENTION)的船舶航迹预测模型。用CNN抽取AIS数据潜在特征,利用BiGRU提取时序数据历史和未来的信息,结合注意力机制来突出关键特征,从而预测船舶航迹。实验结果表明,基于CNN-BiGRU-ATTENTION模型的船舶航迹预测准确度更高,在经度、纬度、航向和航速的预测上,CNN-BiGRU-ATTENTION模型的平均绝对误差(MAE)和均方根误差(RMSE)相比BiGRU和LSTM更低。

In order to maintain maritime traffic safety and grasp the navigation dynamics of vessel in real time,the international maritime organization(IMO)requires vessels to be equipped with automatic identification system(AIS),and AIS data contains many navigation characteristics of vessels,which can be predict the future direction of the vessel,which is convenient for the management of the marine traffic management department.According to vessel AIS data,a vessel trajectory prediction method based on vessel trajectory clustering and CNN-BiGRU-ATTENTION model is proposed.A vessel trajectory prediction model based on convolutional neural network(CNN),bidi-rectional gated recurrent unit(BIGRU)and attention mechanism(ATTENTION)is constructed.CNN is used to extract latent features of AIS data,BiGRU is used to extract historical and future information of time series data,and attention mechanism is used to highlight key features to predict vessel trajectory.The experimental results show that the vessel trajectory prediction accuracy based on the CNN-BiGRU-ATTENTION model is higher.In the prediction of longitude,latitude,heading and speed,the mean absolute error(MAE)and root mean square error(RMSE)of the CNN-BiGRU-ATTENTION model is lower than BiGRU and LSTM.