为解决以往船舶轨迹预测模型在训练过程中未能合理应对多步预测任务间关联性的问题，构建了一种融合频域分析方法和Seq2Seq模型的船舶轨迹预测算法（Time-Frequency based Seqence to Seqence Model，TF-Seq2Seq）。基于自动识别系统数据特点设计了系列预处理操作；用融合频域分析方法和Seq2Seq模型的改进算法捕获船舶轨迹序列的关联性。基于真实数据进行实例分析，结果表明基于长短时记忆单元的Seq2Seq改进算法多种指标上表现最优，均方误差较原算法降低了12.9%。算法改进能更好地发挥模型优势，提高船舶轨迹预测精度。
To address the inadequate handling of inter-task correlations in multi-step prediction during model training in existing approaches, a multi-step vessel trajectory prediction algorithm integrating frequency domain analysis with Seq2Seq architecture (TF-Seq2Seq) is proposed. Systematic preprocessing operations were designed considering characteristics of Automatic Identification System data, while an improved algorithm combining frequency domain analysis and Seq2Seq model was developed to capture temporal correlations in trajectory sequences. Experimental analysis using real-world data demonstrates that the enhanced Seq2Seq model with Long Short-Term Memory units achieves optimal performance across multiple evaluation metrics, with the mean squared error reduced by 12.9% compared with the original algorithm. The proposed improvements effectively enhance model advantages and significantly improve prediction accuracy for vessel trajectories.
2026,48(4): 141-147 收稿日期：2025-5-9
DOI：10.3404/j.issn.1672-7649.2026.04.022
分类号：U675.79
基金项目：国家自然科学基金青年科学基金资助项目（52301402）；上海交通大学深蓝计划（SL2022MS002）；广东省基础与应用基础研究基金项目（2022A1515110574）
作者简介：曹建鑫(2002-),男,硕士研究生,研究方向为智能船舶
参考文献：
[1] 刘钰, 彭鹏菲. 基于 CNN-BiGRU-ATTENTION 的船舶航迹预测方法研究[J]. 计算机与数字工程, 2024, 52(9): 2667-2674.
LIU Y, PENG P F. Vessel trajectory prediction method based on CNN-BiGRU-ATTENTION[J]. Computer & Digital Engineering, 2024, 52(9): 2667-2674.
[2] BEST R A, NORTON J P, et al. A new model and efficient tracker for a target with curvilinear motion[J]. IEEE Transactions on Aerospace and Electronic Systems, 1997, 33(3): 1030-1037.
[3] 徐铁, 蔡奉君, 胡勤友, 等. 基于卡尔曼滤波算法船舶AIS轨迹估计研究[J]. 现代电子技术, 2014, 37(5): 97-100.
XU T, CAI F J, HU Q Y, et al. Research on estimation of AIS vessel trajectory data based on Kalman filter algorithm[J]. Modern Electronics Technique, 2014, 37(5): 97-100.
[4] 甄荣, 金永兴, 胡勤友, 等. 基于AIS信息和BP神经网络的船舶航行行为预测[J]. 中国航海, 2017, 40(2): 6-10.
ZHEN R, JIN Y X, HU Q Y, et al. Vessel behavior prediction based on AIS data and BP neural network[J]. Navigation of China, 2017, 40(2): 6-10.
[5] 权波, 杨博辰, 胡可奇, 等. 基于LSTM的船舶航迹预测模型[J]. 计算机科学, 2018, 45(S2): 126-131.
QUAN B, YANG B C, HU K Q, et al. Prediction model of ship trajectory based on LSTM[J]. Computer Science, 2018, 45(S2): 126-131.
[6] 叶文哲. 基于机器学习和AIS数据的船舶轨迹预测模型的设计与实现[D]. 成都: 电子科技大学, 2021.
[7] GAO D W, ZHU Y S, ZHANG J F, et al. A novel MP-LSTM method for ship trajectory prediction based on AIS data[J]. Ocean Engineering, 2021, 228: 108956.
[8] 陈立家, 周乃祺, 李世刚, 等. 基于C-Informer模型的船舶轨迹预测方法[J]. 交通信息与安全, 2023, 41(6): 51-60.
CHEN L J, ZHOU N Q, LI S G, et al. A method of ship trajectory prediction based on a C-Informer model[J]. Journal of Transport Information and Safety, 2023, 41(6): 51-60.
[9] 唐光旭. 基于Transformer和Bi-LSTM模型的船舶轨迹预测[J]. 珠江水运, 2024(17): 109-111.
TANG G X. Ship trajectory prediction based on transformer and Bi-LSTM models[J]. Pearl River Water Transport, 2024(17): 109-111.
[10] 苗靖, 李晓婷. 基于Seq2Seq-Att的船舶轨迹预测算法[J]. 火力与指挥控制, 2024, 49(4): 71-76.
MIAO J, LI X T. Ship trajectory prediction algorithm basedonSeq2Seq-Att[J]. Fire Control & Command Control, 2024, 49(4): 71-76.
[11] 谢海波, 乔冠洲, 代程, 等. 基于NGO-Bi-GRU的船舶轨迹预测模型[J]. 舰船科学技术, 2025, 47(4): 14-20.
XIE H B, QIAO G Z, DAI C, et al. Ship trajectory prediction model based on NGO-Bi-GRU[J]. Ship Science and Technology, 2025, 47(4): 14-20.
[12] ZHOU T, MA Z Q, WEN Q S, et al. FEDformer: frequency enhanced decomposed transformer for long-term series forecasting[C]// 39th International Conference on Machine Learning (ICML), 2022.
[13] WANG H, PAN L, CHEN Z, et al. FreDF: learning to forecast in frequency domain[J]. arXiv preprint arXiv: 2402.02399, 2024.
[14] HOCHREITER S, SCHMIDHUBER J. long short-term memory[J]. Neural Comput, 1997 (9): 1735-1780.
[15] SUTSKEVER I, VINYALS O, LE Q V. Sequence to sequence learning with neural networks[J]. arXiv preprint arXiv: 1409.3215, 2014.
[16] CHO K, VAN MERRIENBOER B, GULCEHRE C, et al. Learning phrase representations using RNN encoder-decoder for statistical machine translation[J]. arXiv preprint arXiv: 1406.1078, 2014.
