舰船电子工程2026，Vol.46Issue(4)：47-52,6.DOI:10.3969/j.issn.1672-9730.2026.04.009
基于混合Transformer模型的飞行轨迹预测
Flight Trajectory Prediction Based on a Hybrid Transformer Model
摘要
Abstract
Efficient and accurate flight trajectory prediction technology is central to achieving intelligent and information-driv-en air traffic management,significantly enhancing the operational efficiency and safety of air traffic.To address the limitations of tra-ditional trajectory prediction methods in handling high-dimensional data and complex spatiotemporal dependencies,this paper pro-poses a hybrid neural network model that combines self-attention mechanisms,convolutional neural networks(CNN),and a Trans-former structure to improve the accuracy and real-time performance of flight trajectory prediction.The proposed model effectively captures spatiotemporal dependencies in trajectory data.Experimental results indicate that,compared to self-attention CNN-LSTM and traditional CNN-LSTM models,this model achieves notable improvements in prediction accuracy for longitude,latitude,and altitude,particularly exhibiting superior predictive performance at critical trajectory points.关键词
轨迹预测/空中交通管理/自注意力机制/Transformer/时空依赖性Key words
trajectory prediction/air traffic management/self-attention mechanism/Transformer/spatiotemporal dependency分类
航空航天引用本文复制引用
黄晋,赵隆懿,高震,李欣洋..基于混合Transformer模型的飞行轨迹预测[J].舰船电子工程,2026,46(4):47-52,6.基金项目
民航局安全能力建设项目(编号:FY2024MHBZ-12)资助. （编号:FY2024MHBZ-12）
