基于SW-BiConvLSTM的船舶轨迹预测  

Ship Trajectory Prediction Based on SW-BiConvLSTM

作　　者：赵琦 许志远 葛佳薇 董婕 ZHAO Qi;XU Zhi-yuan;GE Jia-wei;DONG Jie(School of Navigation and Naval Architecture,Dalian Ocean University,Dalian Liaoning 116023,China)

机构地区：[1]大连海洋大学航海与船舶工程学院,辽宁大连116023

出　　处：《船海工程》2024年第4期36-42,47,共8页Ship & Ocean Engineering

基　　金：辽宁省教育厅2022年度高校基本科研项目(LJKMZ20221106)。

摘　　要：为了对船舶未来时刻的轨迹或航行趋势进行更加精准预测,进一步增强海上交通安全,提高海上安全航行水平,提出一种基于滑动窗口的双向卷积长短期记忆神经网络(SW-BiConvLSTM)的预测模型。该模型使用滑动窗口提取预处理过AIS数据,然后将提取到的数据输入到双向卷积长短期记忆神经网络中,通过滑动窗口进行输出,最终实现对船舶未来轨迹的预测。将该模型与LSTM、GRU及ConvLSTM等模型分别在直线航行、转弯航行及连续转弯航行3个场景进行对比,结果表明,相较于对比模型,本模型在单步及多步实验上均有更好表现。To more accurately predict the trajectory or navigation trend of the ship,further enhance maritime traffic safety,and improve the level of maritime safe navigation,a prediction model was presented based on a sliding window bidirectional convolutional long short-term memory neural network(SW-BiConvLSTM).The model used a sliding window to extract pre-processed AIS data,input the extracted data into the bidirectional convolutional long short-term memory neural network,and output it through the sliding window to finally predict the ship’s future trajectory.The prediction results of this model were compared with that of LSTM,GRU,and ConvLSTM models in three scenarios:straight sailing,turning,and continuous turning,showing that compared with the comparison model,this model performs better in single-and multi-step experiments.

关 键 词：AIS数据 轨迹预测 神经网络 滑动窗口 

分 类 号：U675.9[交通运输工程—船舶及航道工程]