基于LSTM与1DCNN的导弹轨迹预测方法

doi:10.12305/j.issn.1001-506X.2023.02.22

摘要/Abstract

摘要：

针对弹道导弹等超远程攻击目标的轨迹难以预测的问题, 提出一种基于长短期记忆(long short-term memory, LSTM)网络与一维卷积神经网络(1-dimensional convolutional neural network, 1DCNN)的目标轨迹预测方法。首先, 建立三自由度导弹运动模型, 依据再入类型设计3种目标轨迹数据, 构建机动数据库, 解决轨迹数据的来源问题。其次, 采用重复分割与滑动窗口的方法对轨迹数据进行预处理。然后, 基于LSTM与1DCNN设计了一种目标类型分类网络, 对目标进行初步分类。最后, 基于1DCNN设计轨迹预测网络, 对目标轨迹进行预测。仿真结果表明, 提出的轨迹预测网络能够完成轨迹预测任务, 预测误差在合理范围内。

关键词: 弹道导弹, 目标分类, 轨迹预测, 长短期记忆网络, 一维卷积神经网络

Abstract:

Aiming at the problem that it is difficult to predict the trajectory of ultra-long-range attack targets such as ballistic missiles, a target trajectory prediction method based on long short-term memory (LSTM)network and 1-dimensional convolutional neural network (1DCNN) is proposed. Firstly, a three-degree-of-freedom missile movement model is established, and three target trajectory data are designed according to the type of reentry, and a maneuvering database is constructed to solve the problem of the source of trajectory data. Secondly, the method of repeated segmentation and sliding window is used to preprocess the trajectory data. Then, a target type classification network is designed based on LSTM and 1DCNN to perform preliminary classification of targets. Finally, a trajectory prediction network is designed based on 1DCNN to predict the target trajectory. The simulation results show that the proposed trajectory prediction network can complete the trajectory prediction task, and the prediction error is within a reasonable range.

Key words: ballistic missile, target classification, trajectory prediction, long short-term memory (LSTM) network, 1-dimensional convolutional neural network(1DCNN)

中图分类号:

V448.15

宋波涛, 许广亮. 基于LSTM与1DCNN的导弹轨迹预测方法[J]. 系统工程与电子技术, 2023, 45(2): 504-512.

Botao SONG, Guangliang XU. Missile trajectory prediction method based on LSTM and 1DCNN[J]. Systems Engineering and Electronics, 2023, 45(2): 504-512.

图/表 17

图1

图2

图3

图4

图5

图6

图7

表1

图8

表2

表3

表4

图9

图10

图11

表5

表6

参考文献 24

25	ZHANGJ S,XINGM D,XIEY Y.FEC: a feature fusion framework for SAR target recognition based on electromagnetic scattering features and deep CNN features[J].IEEE Trans.on Geoscience and Remote Sensing,2020,59(3):2174-2187.
26	HUSSAIN M, BIRD J J, FARIA D R. A study on CNN transfer learning for image classification[C]//Proc. of the UK Workshop on Computational Intelligence, 2018: 191-202.
27	WANG Y H, NI Q J, ZHAO S, et al. A hybrid wind speed prediction model based on signal decomposition and deep 1DCNN[C]//Proc. of the International Conference on Swarm Intelligence, 2021: 272-281.
28	LIUZ L,WANGH,LIUJ J,et al.Multitask learning based on lightweight 1DCNN for fault diagnosis of wheelset bearings[J].IEEE Trans.on Instrumentation and Measurement,2020,70,3501711.
29	CAOJ,HEZ D,WANGJ H,et al.An antinoise fault diagnosis method based on multiscale 1DCNN[J].Shock and Vibration,2020,2020,8819313.
30	王新,杨任农,左家亮,等.基于HPSO-TPFENN的目标机轨迹预测[J].西北工业大学学报,2019,37(3):612-620.
	WANGX,YANGR N,ZUOJ L,et al.Target machine trajectory prediction based on HPSO-TPFENN[J].Journal of Northwes-tern Polytechnical University,2019,37(3):612-620.
1	PARKS T,LEEJ G.Design of a practical tracking algorithm with radar measurements[J].IEEE Trans.on Aerospace and Electronic Systems,1998,34(4):1337-1344. doi: 10.1109/7.722718
2	PARKS T,LEEJ G.Improved Kalman filter design for three-dimensional radar tracking[J].IEEE Trans.on Aerospace and Electronic Systems,2001,37(2):727-739. doi: 10.1109/7.937485
3	CHENB D,LIUX,ZHAOH Q,et al.Maximum correntropy Kalman filter[J].Automatica,2017,76,70-77. doi: 10.1016/j.automatica.2016.10.004
4	黄景帅,李永远,汤国建,等.高超声速滑翔目标自适应跟踪方法[J].航空学报,2020,41(9):297-310.
	HUANGJ S,LIY Y,TANGG J,et al.Adaptive tracking method of hypersonic gliding target[J].Journal of Aeronautics,2020,41(9):297-310.
5	YOUNW,MYUNGH.Robust interacting multiple model with modeling uncertainties for maneuvering target tracking[J].IEEE Access,2019,7,65427-65443. doi: 10.1109/ACCESS.2019.2915506
6	QIN Y, LIU Y, LI K J, et al. Maneuvering target tracking based on multiple-model multipath Bernoulli filter[C]//Proc. of the IOP Conference Series: Materials Science and Engineering, 2020: 012111.
7	HOUB W,HEZ M,LID,et al.Maximum corren-tropy unscented Kalman filter for ballistic missile navigation system based on SINS/CNS deeply integrated mode[J].Sensors,2018,18(6):1724. doi: 10.3390/s18061724
8	WANGG Q,ZHANGY G,WANGX D.Iterated maximum correntropy unscented Kalman filters for non-Gaussian systems[J].Signal Processing,2019,163,87-94. doi: 10.1016/j.sigpro.2019.05.015
9	HUANGY L,ZHANGY G,ZHAOY X,et al.A novel out-lier-robust Kalman filtering framework based on statistical similarity measure[J].IEEE Trans.on Automatic Control,2020,66(6):2677-2692.
10	GUIH,DER A H.Quaternion invariant extended Kalman filtering for spacecraft attitude estimation[J].Journal of Gui-dance, Control, and Dynamics,2018,41(4):863-878. doi: 10.2514/1.G003177
11	张炜,崔文,张育卫,等.稀疏数据情况下小椭圆轨道目标再入预报方法[J].空间碎片研究,2020,20(3):30-36.
	ZHANGW,CUIW,ZHANGY W,et al.Reentry prediction method for small elliptical orbit targets under sparse data[J].Space Debris Research,2020,20(3):30-36.
12	杜广洋,郑学合.雷达群目标跟踪条件下的弹道预报方法[J].系统工程与电子技术,2018,40(12):2683-2688. doi: 10.3969/j.issn.1001-506X.2018.12.10
	DUG Y,ZHENGX H.Trajectory prediction method under the condition of radar group target tracking[J].Systems Engineering and Electronics,2018,40(12):2683-2688. doi: 10.3969/j.issn.1001-506X.2018.12.10
13	马国兵,张楠.一种基于神经网络的机动目标轨迹预测方法[J].青岛理工大学学报,2006,27(5):108-111.
	MAG B,ZHANGN.A method of maneuvering target trajectory prediction based on neural network[J].Journal of Qing-dao University of Technology,2006,27(5):108-111.
14	钱夔,周颖,杨柳静,等.基于BP神经网络的空中目标航迹预测模型[J].指挥信息系统与技术,2017,8(3):54-58.
	QIANK,ZHOUY,YANGL J,et al.Air target track prediction model based on BP neural network[J].Command Information System and Technology,2017,8(3):54-58.
15	杨彬,贺正洪.一种GRNN神经网络的高超声速飞行器轨迹预测方法[J].计算机应用与软件,2015,32(7):239-243.
	YANGB,HEZ H.A GRNN neural network based hypersonic vehicle trajectory prediction method[J].Computer Applications and Software,2015,32(7):239-243.
16	张涛,郭基联,徐西蒙,等.基于Elman神经网络的战斗机空战轨迹预测[J].飞行力学,2018,36(5):89-94.
	ZHANGT,GUOJ L,XUX M,et al.Air combat trajectory prediction of fighter aircraft based on Elman neural network[J].Flight Dynamics,2018,36(5):89-94.
17	张共济. 滑翔式高超声速飞行器再入制导技术研究[D]. 南京: 南京航空航天大学, 2016.
	ZHANG G J. Research on reentry guidance technology of gliding hypersonic vehicle[D]. Nanjing: Nanjing University of Aeronautics and Astronautics, 2016.
18	SONG,KWONS,PARKN.Gender classification based on the non-lexical cues of emergency calls with recurrent neural networks (RNN)[J].Symmetry,2019,11(4):525.
19	ZHANGJ J,LIUH Y,CHANGQ,et al.Recurrent neural network for motion trajectory prediction in human-robot colla-borative assembly[J].CIRP Annals,2020,69(1):9-12.
20	WANGJ Y,ZHANGC.Software reliability prediction using a deep learning model based on the RNN encoder-decoder[J].Reliability Engineering & System Safety,2018,170,73-82.
21	SHERSTINSKYA.Fundamentals of re-current neural network (RNN) and long short-term memory (LSTM) network[J].Physica D: Nonlinear Phenomena,2020,404,132306.
22	SHI Z Y, XU M, PAN Q, et al. LSTM-based flight trajectory prediction[C]//Proc. of the International Joint Conference on Neural Networks, 2018.
23	ROSSIL,AJMARA,PAOLANTIM,et al.Vehicle trajectory prediction and generation using LSTM models and GANs[J].Plos One,2021,16(7):e0253868.
24	HUANGL,HEM L,TANC,et al.Jointly network image processing: multi-task image semantic segmentation of indoor scene based on CNN[J].IET Image Processing,2020,14(15):3689-3697.

层定义	层参数	层输出尺寸
LSTM层1	神经元数量32	(30, 32)
LSTM层2	神经元数量32	(30, 32)
1DCNN层1	卷积核数量64, 长度4	(27, 64)
1DCNN层2	卷积核数量128, 长度4	(26, 128)
最大池化层	核大小为4	(6, 128)
展开层	无	(768, 1)
全连接层	神经元数量3	(3, 1)

网络	训练集精度	测试集精度
1DCNN	72.4	71.8
LSTM	77.7	76.4
LSTM-1DCNN	86.0	84.8

层定义	层参数	层输出尺寸
1DCNN层1	卷积核数量32, 长度2	(29, 32)
最大池化层	核大小为2	(14, 32)
1DCNN层2	卷积核数量32, 长度2	(13, 32)
1DCNN层3	卷积核数量32, 长度2	(12, 32)
最大池化层	核大小为2	(6, 32)
展开层	无	(192, 1)
全连接层	神经元数量64	(64, 1)
全连接层	神经元数量32	(32, 1)
全连接层	神经元数量120	(120, 1)

目标类型	预测时长
目标类型	60 s	90 s	120 s
弹道式	569.3	1 591.5	2 016.5
半弹道式	417.5	3 137.7	3 153.1
升力式	457.2	1 047.4	1 576.7

目标类型	预测120 s误差
弹道式	2 635.7
半弹道式	3 860.1
升力式	2 079.8