摘要
航空器的4D航迹预测作为基于航迹运行(TBO)的关键技术之一具有非常重要的意义。基于Transformer-GRU(T-GRU)网络,提出了一种新的航迹预测方法,结合Adamax优化器实现了4D航迹预测。利用Transformer网络的自注意力机制对输入序列进行建模,通过GRU网络获取时序数据的特征;对原始航迹数据进行重采样插值和中值滤波等预处理,以便消除数据缺失和异常值等对预测的影响;通过E E、E AT、E CT、E A等误差指标对实验结果进行评价,并与其他常用的航迹预测方法进行对比。研究结果表明:与传统深度学习方法相比,基于T-GRU网络的4D航迹预测模型在航迹预测中具有更高的准确性和鲁棒性。
The 4D trajectory prediction of aircraft is one of the key technologies based on trajectory-based operations(TBO),which has significant significance.Based on Transformer-GRU(T-GRU)network,a trajectory prediction method was proposed and 4D trajectory prediction was realized by combining with Adamax optimizer.Firstly,the self-attention mechanism of the Transformer network was used to model the input sequence,and the features of time-series data were obtained through the GRU network.Secondly,the original trajectory data was preprocessed by resampling interpolation and median filtering to eliminate the impact of data missing and outliers on prediction.Finally,the experimental results were evaluated through error indicators such as E E,E AT,E CT and E A,and compared with other commonly used trajectory prediction methods.The research results show that the proposed T-GRU network-based 4D trajectory prediction model has higher accuracy and robustness in trajectory prediction,compared with traditional deep learning methods.
作者
翟文鹏
宋一峤
张兆宁
ZHAI Wenpeng;SONG Yiqiao;ZHANG Zhaoning(College of Air Traffic Management,Civil Aviation University of China,Tianjin 300300,China)
出处
《重庆交通大学学报（自然科学版）》
CAS
CSCD
北大核心
2024年第6期94-101,共8页
Journal of Chongqing Jiaotong University(Natural Science)
基金
国家自然科学基金民航联合基金重点项目(U2233209)。
关键词
交通工程
空中交通管理
TBO
4D航迹预测
深度学习
traffic engineering
air traffic management
TBO
4D trajectory prediction
deep learning
