改进：2-SE‘inception -Stacked LSTM 新型网络模型可视化轨迹展示与结果分析

结合原模型与改进模型，这里选取五种模型进行对比

为了方便描述，我们编号简化模型如下表 ：

每种模型的结构图如下：

model：Ⅰ

略

model：Ⅱ

model：Ⅲ

model：Ⅳ

model：Ⅴ

可视化预测轨迹如下：

Ⅰ

Ⅱ

Ⅲ

Ⅳ

Ⅴ

残差图：

加粗样式

总分析：

首先航班M预测轨迹始终不理想，模型的改进并没有使得其误差减小，我们认为可能是数据集本身错误。即剔除不完全所致，所以在此后的分析中我们都舍弃了航班M而讨论。
对于轨迹的分析，我们选取了具有代表性的如上五个航班进行分析如下：
Ⅰ：曲线与实际偏离程度大，较为曲折，杂乱。预测效果并不是很好是不可取的。
Ⅱ：曲线与实际偏离程度与Ⅰ有明显差别，显著变小且更接近。可见改进的新型网络相较于Ⅰ有了很大效果。
Ⅲ：曲线与实际偏离程度与Ⅱ无明显差别，但根据总体残差来看略小于Ⅱ，可见增加注意力机制对于总体误差是由一定作用的，从轨迹上难以分辨可能是注意力机制选取不合适，后续仍有进一步研究的空间。
Ⅳ：曲线与实际偏离程度较Ⅱ略大，但不明显。但是总体残差比Ⅲ略小一点，但是相较于model：Ⅰ、Ⅱ、Ⅲ，从轨迹上来看，折线变多了，即“鬼画符”情况得到了改善，我们在此对堆叠通道可以改善残差持保留意见，但是可以发现，堆叠通道或许可以减少曲线的“曲折性”。
Ⅴ：曲线与实际偏离程度较Ⅱ有大有小，且飘忽不定，且总体残差比Ⅱ略大一点，可以看作一样。但相较于Ⅳ，从轨迹上来看，折线进一步变多，即“鬼画符”情况得到进一步改善，堆叠通道可以减少曲线的“曲折性”，同时，由模型Ⅳ我们持保留意见的：“堆叠通道可以改善残差或许是不对得”也得到了进一步印证。
注：以上分析仅针对在本模型中，并非普适结论。

直方图代码

% 创建数据
data = [175.643, 124.746, 119.475, 110.428, 122.473];

xvalues = {'Ⅰ', 'Ⅱ', 'Ⅲ', 'Ⅳ', 'Ⅴ'};

% 画直方图
bar(data)

% 设置横纵轴属性
set(gca, 'XTick', 1:numel(data), 'XTickLabel', xvalues)
xlabel('Model type');
ylabel('Unnomalized 2D-Error');
title('Error histograms of each model（Ⅰ-Ⅴ）');
% 显示各柱子的高度
% 显示各柱子的高度
for i = 1:length(data)
    text(i, data(i), num2str(data(i)), 'HorizontalAlignment', 'center', 'VerticalAlignment', 'bottom')
end

残差代码

%示例A、B为所要求残差的两个数组
A = [1, 3, 5, 7, 9];
B = [2, 3, 4, 2, 6];

% 计算绝对值之和
abs_sum = sum(abs(A - B));
disp(['对应位置元素相减的绝对值之和为：', num2str(abs_sum)]);

工作总结：

本文针对基于 ADS-B 数据的飞机轨迹预测问题，开展了较为深入的研究，在预测模型设计等方面提供了思路与方法，并初步取得了一些研究成果与结论，具体如下：
一 、对ADS-B数据集进行了深入的分析，发现安全性正面临重大的挑战，并指出传统方法在预测中缺点越来越明显，深度学习在此方面优势越来越到，大有可为。
二 、复现了一种基于 N-Iception-LSTM 的轨迹预测模型。并对ADS-B 数据集进行了特别的处理，使其具有时间与空间特征，Iception适合空间扩展，并且对于提取局部空间特征也是非常有优势，再加上 LSTM 具有一定的内存容量，多用于处理时间序列数据，结合二者优势的情况下，设计出的N-Iception-LSTM预测模型。
实验结果表明：该模型相较于 LSTM、Bi-directional LSTM、Stacked LSTM，可以达到更好的预测性能。
三 、改进了这种方法并做出多种对比，【包括：2-SE’Inceptio-LSTM，2-SE’Inceptio-Stacked LSTM（1）/（2）】。
实验结果表明：2-SE’Inception-LSTM优于2-Iception-LSTM，但计算量更大，为更多实际应用场景下的轨迹预测提供选择，2-SE’Inceptio-Stacked LSTM（1）/（2）模型之间经度与前面差别不大，但是对于轨迹的曲折性明显减少，对智能空中交通管制产生深远影响总体而言，本文研究成果具有重要的学术意义以及潜在的工程应用价值。

未来工作展望

本文的预测实验是基于 ADS-B 数据进行研究的，通过 LSTM、N-Iception-LSTM、2-SE’Inceptio-LSTM、2-SE’Inceptio-Stacked LSTM（1）、2-SE’Inceptio-Stacked LSTM（2） 这几个模型进行预测，虽然取得一定的研究成果，但由于篇幅限制，本研究还有一些需要进一步完善的地方：
本文所采用的是基于安全的 ADS-B 数据（只用了信号强度）的飞机轨迹预测研究，但传统空中交通管制系统监视部分的主要手段还是依赖一二次雷达，而 ADS-B 技术作为辅助监视手段，但基于 ADS-B 数据的轨迹预测仍有一定的研究意义。
在改进时，由于时间的原因，本文只加入了一种通道注意力机制，还有更多注意力机制值得探索，或将取得更好的效果。
本文为了充分利用数据特征，将数据进行了较为繁琐的预处理，由于涉及到切片因此要进行维度还原。本文只是采取重复测量去评价的办法 来来进行还原，事实上，时间窗格的不同位置有着不同的影响因子，可以考虑设计新式模型来进一步提升精度。
最后本文研究都是基于深度学习方法，没有使用机器学习的方法、基于传统滤波器的方法、基于空气动力学的研究方法，虽然实验模型丰富多彩，但是实验种类比较单一，不能很好地从实验层面展示深度学习的优势。

参考文献：

[1] K. Yu, L. Lin, M. Alazab, L. Tan, and B. Gu, “Deep learningbased traffific safety solution for a mixture of autonomous and manual vehicles in a 5G-enabled intelligent transportation system,” IEEE Trans. Intell. Transp. Syst., early access, Dec. 22, 2020, doi: 10.1109/ TITS.2020.3042504.
[2] N. Cheng et al., “A comprehensive simulation platform for space-air-ground integrated network,” IEEE Wireless Commun., vol. 27, no. 1, pp. 178–185, Feb. 2020.
[3] N. Kato et al., “Optimizing space-air-ground integrated networks by arti- fificial intelligence,” IEEE Wireless Commun., vol. 26, no. 4, pp. 140–147, Aug. 2019.
[4] N. Zhang, S. Zhang, P. Yang, O. Alhussein, W. Zhuang, and X. S. Shen, “Software defifined space-air-ground integrated vehicular networks: Challenges and solutions,” IEEE Commun. Mag., vol. 55, no. 7, pp. 101–109, Jul. 2017.
[5] L. Zhen, A. K. Bashir, K. Yu, Y. D. Al-Otaibi, C. H. Foh, and P. Xiao, “Energy-effificient random access for LEO satellite-assisted 6G internet of remote things,” IEEE Internet Things J., vol. 8, no. 7, pp. 5114–5128, Apr. 2021.
[6] R. Slattery and Y. Zhao, “Trajectory synthesis for air traffific automation,” J. Guid. Control. Dyn., vol. 20, no. 2, pp. 232–238, Mar. 1997.
[7] J. Post, “The next generation air transportation system of the united states: Vision, accomplishments, and future directions,” Engineering, vol. 7, no. 4, pp. 427–430, Apr. 2021, doi: 10.1016/j.eng.2020.05.026.
[8] I. Hwang and C. E. Seah, “Intent-based probabilistic conflflict detection for the next generation air transportation system,” Proc. IEEE, vol. 96, no. 12, pp. 2040–2059, Dec. 2008.
[9] D. McCallie, J. Butts, and R. Mills, “Security analysis of the ADS-B implementation in the next generation air transportation system,” Int. J. Crit. Infrastruct. Prot., vol. 4, no. 2, pp. 78–97, Aug. 2011.
[10] P. Pellegrini and J. Rodriguez, “Single European sky and single European railway area: A system level analysis of air and rail
transportation,” Transp. Res. A, Policy Pract., vol. 57, pp. 64–76, Nov. 2013.
[11] A. Sipe and J. Moore, “Air traffific functions in the NextGen and SESAR airspace,” in Proc. IEEE/AIAA 28th Digit. Avionics Syst. Conf., Oct. 2009, pp. 2.A.6-1–2.A.6-7.
[12] S. Ramasamy, R. Sabatini, A. G. Gardi, and Y. Liu, “Novel flflight management system for real-time 4-dimensional trajectory based opera ions,” in Proc. AIAA Guid., Navigat., Control (GNC) Conf., Aug. 2013, pp. 1–5.
[13] J. Lopez-Leones, M. A. Vilaplana, E. Gallo, F. A. Navarro, and C. Querejeta, “The aircraft intent description language: A key enabler for air-ground synchronization in trajectory-based operations,” in Proc. IEEE/AIAA 26th Digit. Avionics Syst. Conf., Oct. 2007, pp. 1–5.
[14] E. Kim and K. Sivits, “Blended secondary surveillance radar solutions to improve air traffific surveillance,” Aerosp. Sci. Technol., vol. 45, pp. 203–208, Sep. 2015.
[15] Li Dong Bin, Liu Yu Mei, Xu Jian Wu, Zhou Yuan, Song Yu. Risk Assessment on the Transi ion Program for Air Traffic Control Automation System Upgrade[J]. MATEC Web of Conferences,2016,63.
[16] Jarvis (Trey) J. Arthur, Randall E. Bailey, Steven P. Williams, Lawrence J. Prinzel, Kevin J.Shelton, Denise R. Jones, Vincent E. Houston. Review of head-worn displays for the Next Generation Air Transportation System[J]. Optical Engineering,2017,56(5).
[17] Inseok Hwang, Chze Eng Seah. Intent-Based Probabilistic Conflict Detection for the Next Generation Air Transportation System[J]. Proceedings of the IEEE,2008,96(12).
[18] Marayong Panadda, Strybel Thomas Z, Robles Jose, O’Connor Ryan, Vu Kim Phuong L, Battiste Vernol. Force-Feedback Integration with NASA’s Next Generation Air Transportation System Cockpit Situation Display[J]. Journal of Air Transportation,2017,25(1).
[19] Leonardi Mauro, Strohmeier Martin, Lenders Vincent. On Jamming Attacks in Crowdsourced Air Traffic Surveillance[J]. IEEE AEROSPACE AND ELECTRONIC SYSTEMS MAGAZINE,2021,36(6).
[20] Subramanian Ramasamy, Roberto Sabatini, Alessandro Gardi, Trevor Kistan. Next Generation Flight Management System for Real-Time Trajectory Based Operations[J]. Applied Mechanics and Materials,2014,3446(629-629).
[21] Siqi HAO, Shaowu CHENG, Yaping ZHANG. A multi-aircraft conflict detection and resolution method for 4-dimensional trajectory-based operation[J]. Chinese Journal of Aeronautics,2018,31(7).
[22] Euiho Kim, Kevin Sivits. Blended secondary surveillance radar solutions to improve air traffic surveillance[J]. Aerospace Science and Technology, 2015, 45.
[23] Zhi-Jun Wu, Shan Tian, Lan Ma. A 4D Trajectory Prediction Model Based on the BP Neural Network[J]. Journal of Intelligent Systems,2019,29(1).
[24] Juan A. Besada, Guillermo Frontera, Jesus Crespo,Enrique Casado, Javier Lopez-Leones. Automated Aircraft Trajectory Prediction Based on Formal Intent-Related Language Processing.[J]. IEEE Trans. Intelligent Transportation Systems,2013,14(3).
[25] Marco Porretta, Marie-Dominique Dupuy, Wolfgang Schuster, Arnab Majumdar,Washington Ochieng. Performance Evaluation of a Novel 4D Trajectory Prediction Model for Civil Aircraft[J]. Journal of Navigation,2008,61(3).
[26] Jiang Gaoyang, Liu Genfeng, Yu Hansong. A Model Free Adaptive Scheme for Integrated Control of Civil Aircraft Trajectory and Attitude[J]. Symmetry,2021,13(2).
[27] I. Hwang and C. E. Seah, An Estimation Algorithm for Stochastic Linear Hybrid Systems with Continuous-State-Dependent Mode Transitions[J], Proceedings of the 45th IEEE Conference on Decision and Control, 2006:131-136.
[28] Urban Maeder, Manfred Morari, Thomas Ivar Baumgartner. Trajectory Prediction for Light Aircraft[J]. Journal of Guidance, Control, and Dynamics,2012,34(4).
[29] Sang-Hyo Arman Woo, Jong-Jin Shin, Jingyu Kim. Implementation and Analysis of Pattern Propagation Factor Based Radar Model for Path Planning[J]. Journal of Intelligent & Robotic Systems,2019,96(3-4).
[30] Tingting Gao, Xiaofeng Liu, Rui Zhang. Traffic safety evaluation based on fuzzy clustering analysis: case studies in the city-cluster region of Beijing-Tianjin-Hebei[J]. IOP Conference Series: Materials Science and Engineering,2019,688(4).
[31] Yi Min, Xie Wei, Mo Li. Short-Term Electricity Price Forecasting Based on BP Neural Network Optimized by SAPSO[J]. Energies,2021,14(20).
[32] Jing Wang, Yunkai Zou, Jianli Ding. ADS-B spoofing attack detection method based on LSTM[J]. EURASIP Journal on Wireless Communications and Networking,2020,2020(1).
[33] Bo-Sung Kwon, Rae-Jun Park, Kyung-Bin Song. Short-Term Load Forecasting Based on Deep Neural Networks Using LSTM Layer[J]. Journal of Electrical Engineering &Technology,2020,15(4).
[34] Han Ping, Yue Jucai, Fang Cheng, Shi Qingyan, Yang Jun. Short-term 4D trajectory prediction based on LSTM neural network[J]. SECOND TARGET RECOGNITION AND ARTIFICIAL INTELLIGENCE SUMMIT FORUM,2020,11427.
[35] Tang Huang, Yin Yong, Shen Helong. A model for vessel trajectory prediction based on long short-term memory neural network[J]. Journal of Marine Engineering & Technology,2019.
[36] Thomas Zeh, Judith Rosenow, Hartmut Fricke. Interdependent Uncertainty Handling in Trajectory Prediction[J]. Aerospace,2019,6(2).
[37] Juan A. Besada, Guillermo Frontera, Jesus Crespo, Enrique Casado, Javier Lopez-Leones. Automated Aircraft Trajectory Prediction Based on Formal Intent-Related LanguageProcessing.[J]. IEEE Trans. Intelligent Transportation Systems,2013,14(3).
[38] Wu Xiping, Yang Hongyu, Chen Hu, Hu Qinzhi, Hu Haoliang. Long-term 4D trajectory prediction using generative adversarial networks[J]. Transportation Research Part C,2022,136.
[39] 赵凯辉, 吴思成, 李涛, 贺才春, 查国涛. 基于Inception-BLSTM的滚动轴承故障诊断方法研究[J].振动与冲击,2021,40(17):290-297.
[40] 曹博, 刘文评, 刘志刚, 刘逸. 基于1090MHz ADS-B系统分析[J].中国科技信息,2019(17):62-63.
[41]暴佳伟,田小平,刘宇娜等.ADS-B安全问题研究综述[J].现代防御技术,2022,50(05):28-35.
[42]生柳振. 基于ADS-B数据的民航飞机轨迹预测研究[D].南京邮电大学,2022.DOI:10.27251/d.cnki.gnjdc.2022.001216.