北京信息科技大学学报（自然科学版）2026，Vol.41Issue(1)：1-11,11.DOI:10.16508/j.cnki.11-5866/n.2026.01.001
时域检索与频域扰动双域联合增强的船舶轨迹预测
Dual-domain joint augmentation via temporal retrieval and frequency perturbation for vessel trajectory forecasting
摘要
Abstract
Vessel trajectory forecasting is of significant importance for ensuring maritime safety,optimizing traffic management,and achieving intelligent decision-making.Existing methods are often limited by insufficient data diversity and inadequate use of historical information,making them difficult to deal with complex and dynamic marine environments.To address this issue,a dual-domain joint augmentation method via temporal retrieval and frequency perturbation for vessel trajectory forecasting was proposed.By expanding sample diversity through frequency perturbation and enhancing the model's ability to represent long-term dependencies by drawing on similar historical trajectory sequences,the proposed method significantly improves the generalization and robustness of the prediction model while maintaining high accuracy.Experimental results on multiple real-world vessel trajectory datasets show that the proposed method outperforms the mainstream time-series forecasting models,including LSTM,BiLSTM,DLinear,Transformer,Informer,and iTransformer,in terms of prediction accuracy and stability,verifying the effectiveness and superiority of this method in vessel trajectory forecasting tasks.关键词
轨迹预测/检索增强生成/数据增广/自动识别系统Key words
trajectory forecasting/retrieval-augmented generation/data augmentation/automatic identification system(AIS)分类
交通工程引用本文复制引用
郭亚男,王晨腾,张本奎,常颖,刘志哲,曹林,杜康宁..时域检索与频域扰动双域联合增强的船舶轨迹预测[J].北京信息科技大学学报（自然科学版）,2026,41(1):1-11,11.基金项目
国家自然科学基金项目(U20A20163,62201066) （U20A20163,62201066）
目标认知与应用技术重点实验室开放基金(2023-CXPT-LC-005) （2023-CXPT-LC-005）
北京市自然科学基金项目(4264103) （4264103）
