基于改进二阶优化器并行学习的弹道导弹神经网络落点预测方法

doi:10.7527/S1000-6893.2022.27964

Abstract

Abstract:

To address the requirement for Impact Point Prediction （IPP） in precision guidance for the ballistic missile after high maneuver penetration， an IPP Neural Network （NN） model is proposed based on an improved second-order optimizer. The IPP of the current flight state is predicted based on the elliptical trajectory theory， and the impact deviation is decoupled. Furthermore， a sample set with flight state as input and impact deviation of elliptical trajectory as output is constructed， which greatly reduces the difficulty of NN learning. To improve the prediction accuracy， three NNs are applied to predict the three components of the impact deviation. An improved Levenberg-Marquardt optimizer for multi-GPU parallel learning is established by using the matrix block algorithm， which shortens the network learning time and reduces the demand for GPU memory. A simulation experiment is designed to analyze the advantages and computational complexity of the proposed method. Simulation results show that the proposed method has low learning difficulty， high prediction accuracy and good real-time performance. Among the 869 320 samples contained in the training set and the test set， the $3 σ$ prediction error is 4.97 m. In the learning environment with 2 GPUs， the learning time is reduced by about 49.18%. On the STM32F407 single-chip microcomputer， IPP takes 2.585 ms. The proposed method can provide support for the design of guidance algorithm， and is applicable for engineering practice.

Key words: ballistic missile, high maneuver penetration, impact point prediction, neural network, parallel learning, elliptical trajectory, single-chip microcomputer

CLC Number:

V448.131

Leliang REN, Yong XIAN, Shaopeng LI, Gang LEI, Wei WU, Bing LI. A neural network model for impact point prediction of ballistic missile based on improved second-order optimizer with parallel learning[J]. Acta Aeronautica et Astronautica Sinica, 2023, 44(14): 327964-327964.

Figures/Tables 23

Fig.1

Fig.2

Fig.3

Fig.4

Fig. 5

Table 1

Range of simulation parameters

参数	下界	上界	离散间隔
$H 0,1$ /km	100.0	200.0	10.0
$H m$ /km	0	5.0	1.0
$φ s$ /（°）	20.0	36.0	4.0
$V s$ $/ (m ⋅ s - 1)$	6 650.0	7 850.0	40.0
$σ s$ /（°）	0	360.0	30.0
$Θ s$ /（°）	-18.5	-12.5	1.0

Table 1

Fig.6

Fig.7

Fig.8

Fig.9

Fig.10

Fig.11

Table 2

Network architecture of the proposed method

模型	隐藏层1	隐藏层2
$f x N N ·$	11	6
$f y N N ·$	6	5
$f z N N ·$	5	2

Table 2

Fig.12

Fig.13

Table 3

Network architecture of Ref. ［5］

模型	隐藏层1	隐藏层2
$f x, D T N N ·$	16	9
$f y, D T N N ·$	11	8
$f z, D T N N ·$	8	5

Table 3

Table 4

Computational complexity of the proposed method and Ref. ［5］

模型	激活函数个数		乘法量
$f x, D T N N ·$	25	共57	8×16+16×9+9×1=281	共582
$f y, D T N N ·$	19		8×11+11×8+8×1=184
$f z, D T N N ·$	13		9×8+8×5+5×1=117
$f x N N ·$	17	共35	8×11+11×6+6×1=160	共300
$f y N N ·$	11		8×6+6×5+5×1=83
$f z N N ·$	7		9×5+5×2+2×1=57

Table 4

Fig.14

Fig. 15

Fig.16

Fig.17

Table 5

Training runtime and GPU memory consumption

网络模型	网络节点数		$N G P U$	1 000代训练平均耗时/s	耗时降低率/%	$ϑ 1 = ϑ 2 = 1$ 时的显存占用量/MB		$ϑ 1 = ϑ 2 = 2$ 时的显存占用量/MB
网络模型	隐藏层1	隐藏层2	$N G P U$	1 000代训练平均耗时/s	耗时降低率/%	GPU1	GPU2	GPU1	GPU2
$f z N N ·$	5	2	2	59.97	32.30	2 105	2 003	1 563	1 461
$f z N N ·$	5	2	1	88.58	32.30	3 233		2 127
$f y N N ·$	6	5	2	62.26	36.00	2 483	2 381	1 743	1 641
$f y N N ·$	6	5	1	97.28	36.00	3 907		2 501
$f x, D T N N ·$	16	9	2	233.68	47.46	6 509	6 405	3 793	3 689
$f x, D T N N ·$	16	9	1	444.73	47.46	12 043		6 527
$f N N 1$	17	16	2	1 511.55	48.86	19 531	19 429	10 201	10 099
$f N N 1$	17	16	1	2 955.88	48.86	显存不足		19 553
$f N N 1$	20	15	2	1 875.13	49.18	21 577	21 475	11 317	11 215
$f N N 1$	20	15	1	3 689.83	49.18	显存不足		21 599

Table 5

Table 6

References 44

1	周啟航，刘延芳，齐乃明，等. 基于反预警的反拦截中段规避突防策略［J］. 航空学报， 2017， 38（1）： 319922.
	ZHOU Q H， LIU Y F， QI N M， et al. Anti-warning-based anti-interception avoiding penetration strategy in midcourse［J］. Acta Aeronautica et Astronautica Sinica， 2017， 38（1）： 319922 （in Chinese）.
2	YANG C J， WU J， LIU G Q， et al. Ballistic missile maneuver penetration based on reinforcement learning［C］∥ 2018 IEEE CSAA Guidance， Navigation and Control Conference （CGNCC）. Piscataway： IEEE Press， 2020： 1-5.
3	樊博璇，陈桂明，林洪涛. 弹道导弹中段反应式机动突防规避策略［J/OL］. 兵工学报，（2021-11-29）［2022-08-23］. .
	FAN B X， CHEN G M， LIN H T. Mid-course reactive maneuver penetration and evading strategy of ballistic missile［J/OL］. Acta Armamentarii，（2021-11-29）［2022-08-23］. （in Chinese）.
4	JIANG L， NAN Y， LI Z H. Realizing midcourse penetration with deep reinforcement learning［J］. IEEE Access， 2021， 9： 89812-89822.
5	XIAN Y， REN L L， XU Y J， et al. Impact point prediction guidance of ballistic missile in high maneuver penetration condition［J/OL］. Defence Technology，（2022-06-13）［2022-08-23］. .
6	DRENICK R. The perturbation calculus in missile ballistics［J］. Journal of the Franklin Institute， 1951， 251（4）： 423-436.
7	PADHI R. An optimal explicit guidance scheme for ballistic missiles with solid motors［C］∥ Guidance， Navigation， and Control Conference and Exhibit. Reston： AIAA， 1999： 1006-1016.
8	BURCHETT B， COSTELLO M. Model predictive lateral pulse jet control of an atmospheric rocket［J］. Journal of Guidance， Control， and Dynamics， 2002， 25（5）： 860-867.
9	ZHANG X， YAO X X， ZHENG Q S. Impact point prediction guidance based on iterative process for dual-spin projectile with fixed canards［J］. Chinese Journal of Aeronautics， 2019， 32（8）： 1967-1981.
10	杨泗智，龚春林，郝波，等. 基于落点预测的高旋火箭弹弹道修正算法［J］. 航空学报， 2020， 41（2）： 323421.
	YANG S Z， GONG C L， HAO B， et al. Ballistic trajectory correction algorithms of high-spin rocket based on impact point prediction［J］. Acta Aeronautica et Astronautica Sinica， 2020， 41（2）： 323421 （in Chinese）.
11	ZHANG X， LEI H M， LI J， et al. Ballistic missile trajectory prediction and the solution algorithms for impact point prediction［C］∥ Proceedings of 2014 IEEE Chinese Guidance， Navigation and Control Conference. Piscataway： IEEE Press， 2015： 879-883.
12	牛云. 椭圆弹道射程角和飞行时间公式的一种推导方法［J］. 国防科技大学学报， 1990， 12（2）： 55-57.
	NIU Y. A method for deducing the formulas of the range angle and the flying time with the elliptical orbit［J］. Journal of National University of Defense Technology， 1990， 12（2）： 55-57 （in Chinese）.
13	任萱. 自由飞行时摄动方程的状态转移矩阵的解析解［J］.中国空间科学技术，1983， 3（1）：1-16.
	REN X. An state transition analytical solution for free flight perturbation equations ［J］. Chinese Space Science and Technology， 1983，3（1）：1-16 （in Chinese）.
14	郑伟. 地球物理摄动因素对远程弹道导弹命中精度的影响分析及补偿方法研究［D］. 长沙：国防科学技术大学， 2006.
	ZHENG W. Analysis on the influence of geophysical perturbation factors on the hit accuracy of long-range ballistic missiles and research on compensation methods［D］. Changsha： National University of Defense Technology， 2006 （in Chinese）.
15	郑伟，汤国建. 弹道导弹自由段解算的等高约束解析解［J］. 宇航学报， 2007， 28（2）： 269-272.
	ZHENG W， TANG G J. Contour restricted analytical solution for free flight trajectory of ballistic missile［J］. Journal of Astronautics， 2007， 28（2）： 269-272 （in Chinese）.
16	WANG L， ZHENG W， ZHOU X. Orbit state deviation prediction model with second-order correction due to the J₂ term［J］. Journal of Physics： Conference Series， 2018， 1074： 012104.
17	王磊. 基于状态空间摄动法的战略导弹弹道快速预报与制导方法研究［D］. 长沙：国防科技大学， 2018.
	WANG L. Fast trajectory prediction and guidance algorithm for strategic missiles based on state space perturbation method［D］. Changsha： National University of Defense Technology， 2018 （in Chinese）.
18	李晓明. 经典f、g级数的修正法［J］. 国防科技大学学报， 1991， 13（2）： 59-67.
	LI X M. A correcting method for classical f and g series［J］. Journal of National University of Defense Technology， 1991， 13（2）： 59-67 （in Chinese）.
19	朱龙根. 改进的Barrar型中间轨道：远程弹道飞行器自由飞行段的解析解［J］. 国防科技大学学报， 1985， 7（2）： 67-82.
	ZHU L G. An improved barrar—type intermediate orbit analytic solutions for free flight trajectory of long—range ballistic vehicle［J］. Journal of National University of Defense Technology， 1985， 7（2）： 67-82 （in Chinese）.
20	李连仲. 弹道飞行器自由飞行轨道的解析解法［J］.宇航学报，1982，13 （1）：1-17.
	LI L Z. An analytic method for solving the equations of free flight trajectory of ballistic vehicle［J］. Journal of Astronautics，1982，13（1）： 1-17 （in Chinese）.
21	常晓华. 考虑地球非球形引力摄动影响的自由段弹道解析解［J］. 国防科技大学学报， 2018， 40（4）： 80-86.
	CHANG X H. Analytical solution for free flight trajectory considering earth non-spherical gravitation perturbation［J］. Journal of National University of Defense Technology， 2018， 40（4）： 80-86 （in Chinese）.
22	KOZAI Y. The motion of a close earth satellite［J］. The Astronomical Journal， 1959， 64（9）： 367-377.
23	WANG Z J， ZHANG J Z， WEI W. Deep learning based missile trajectory prediction［C］∥ 2020 3rd International Conference on Unmanned Systems （ICUS）. Piscataway： IEEE Press， 2020： 474-478.
24	王森. 基于机器学习的弹道落点预测研究［D］. 南京：南京理工大学， 2020.
	WANG S. Study on impact point prediction based on machine learning［D］. Nanjing： Nanjing University of Science and Technology， 2020 （in Chinese）.
25	余跃，王宏伦. 基于深度学习的高超声速飞行器再入预测校正容错制导［J］. 兵工学报， 2020， 41（4）： 656-669.
	YU Y， WANG H L. Deep learning-based reentry predictor-corrector fault-tolerant guidance for hypersonic vehicles［J］. Acta Armamentarii， 2020， 41（4）： 656-669 （in Chinese）.
26	赵捍东，黄鑫，马焱. 基于神经网络补偿的线性弹道落点预报方法［J］. 探测与控制学报， 2017， 39（4）： 96-102， 107.
	ZHAO H D， HUANG X， MA Y. Impact point prediction method of linear trajectory based on neural network compensation［J］. Journal of Detection & Control， 2017， 39（4）： 96-102， 107 （in Chinese）.
27	SHEN Z J， LU P. Onboard generation of three-dimensional constrained entry trajectories［J］. Journal of Guidance， Control， and Dynamics， 2003， 26（1）： 111-121.
28	WILAMOWSKI B M， YU H. Improved computation for Levenberg-Marquardt training［J］. IEEE Transactions on Neural Networks， 2010， 21（6）： 930-937.
29	魏倩，蔡远利. 一种基于神经网络的中制导改进算法［J］. 西安交通大学学报， 2016， 50（7）： 125-130.
	WEI Q， CAI Y L. A modified algorithm on the midcourse guidance based on BP neural network［J］. Journal of Xi’an Jiaotong University， 2016， 50（7）： 125-130 （in Chinese）.
30	JI R P， LIANG Y， XU L F， et al. Trajectory prediction of ballistic missiles using Gaussian process error model［J］. Chinese Journal of Aeronautics， 2022， 35（1）： 458-469.
31	鲜勇，李邦杰，雷刚，等. 弹道导弹精度分析方法［M］. 长沙：国防科技大学出版社， 2012.
	XIAN Y， LI B J， LEI G， et al. Ballistic missile precision analysis method［M］. Changsha： National University of Defense Technology Press， 2012 （in Chinese）.
32	张金槐. 远程火箭精度分析与评估［M］. 长沙：国防科技大学出版社， 1995.
	ZHANG J H. Accuracy evaluation and analysis of long range rocket［M］. Changsha： National University of Defense Technology Press， 1995 （in Chinese）.
33	王文龙. 大气风场模型研究及应用［D］. 长沙：国防科技大学， 2009.
	WANG W L. Atmospheric wind field modeling and its application［D］. Changsha： National University of Defense Technology， 2009 （in Chinese）.
34	HEATH M T. Scientific computing： An introductory survey［M］. New York： McGraw-Hill Inc， 2005.
35	CURTIS H D. Orbital maneuvers［M］∥ Orbital Mechanics for Engineering Students. Amsterdam： Elsevier， 2014： 299-365.
36	张毅，肖龙旭，王顺宏. 弹道导弹弹道学［M］. 长沙，国防科技大学出版社， 1999 （in Chinese）.
	ZHANG Y， XIAO L X， WANG S H. Missile ballistic［M］. Changsha： National University of Defense Technology Press， 1999 （in Chinese）.
37	HE K M， ZHANG X Y， REN S Q， et al. Deep residual learning for image recognition［C］∥ 2016 IEEE Conference on Computer Vision and Pattern Recognition （CVPR）. Piscataway： IEEE Press， 2016： 770-778.
38	郭玮林，鲜勇，张大巧，等. 高超声速飞行器助推段弹道快速计算方法［J］. 中国惯性技术学报， 2018， 26（1）： 109-114.
	GUO W L， XIAN Y， ZHANG D Q， et al. Fast calculation method of booster trajectory for hypersonic vehicle［J］. Journal of Chinese Inertial Technology， 2018， 26（1）： 109-114 （in Chinese）.
39	姚琳怡. 基于强化学习的高速公路项目级养护决策研究［D］. 南京：东南大学， 2019.
	YAO L Y. Research on project level decision-making of highway asphalt pavement maintenance based on reinforcement learning［D］. Nanjing： Southeast University， 2019 （in Chinese）.
40	MARQUARDT D W. An algorithm for least-squares estimation of nonlinear parameters［J］. Journal of the Society for Industrial and Applied Mathematics， 1963， 11（2）： 431-441.
41	HAGAN M T， MENHAJ M B. Training feedforward networks with the Marquardt algorithm［J］. IEEE Transactions on Neural Networks， 1994， 5（6）： 989-993.
42	DUCHI J C， HAZAN E， SINGER Y. Adaptive subgradient methods for online learning and stochastic optimization［J］. Journal of Machine Learning Research， 2011， 12： 2121-2159.
43	RUDER S. An overview of gradient descent optimization algorithms［DB/OL］. arXiv preprint： 1609.04747，2016. .
44	KINGMA D P， BA J L. Adam： A method for stochastic optimization［C］∥ International Conference on Learning Representations. 2015： 1-13.

模型	神经网络预测耗时/ms	椭圆弹道预测耗时/ms	其他耗时/ms	总耗时/ms
本文方法	1.250	0.634	0.696	2.585
文献［5］方法	2.299		0.696	2.995

[1]	Zhongzhi LI, Jinyi MA, Jianliang AI, Yiqun DONG. Fault detection and classification of aerospace sensors using deep neural networks finetuned from VGG16 [J]. Acta Aeronautica et Astronautica Sinica, 2023, 44(S1): 727615-727615.
[2]	Wu LIU, Yunyan WU, Wei LIU, Mingming TIAN, Tianpeng HUANG. Re-entry robust fault tolerant attitude control for RLVs considering unknown disturbances [J]. Acta Aeronautica et Astronautica Sinica, 2023, 44(S1): 727787-727787.
[3]	Chenyang LIU, Dawei WU, Yize GUO, Xinsai LV, Jiani ZHOU, Shuyi SHAO. Robust adaptive attitude control of quadrotor with uncertain strong coupling [J]. Acta Aeronautica et Astronautica Sinica, 2023, 44(S1): 727645-727645.
[4]	Zhikai WANG, Sheng CHEN, Wei FAN. Effect of neural network width on combustor emission prediction [J]. ACTA AERONAUTICAET ASTRONAUTICA SINICA, 2023, 44(5): 126816-126816.
[5]	Lei HE, Weiqi QIAN, Kangsheng DONG, Xian YI, Congcong CHAI. Aerodynamic characteristics modeling of iced airfoil based on convolution neural networks [J]. ACTA AERONAUTICAET ASTRONAUTICA SINICA, 2023, 44(5): 126434-126434.
[6]	Yucun SONG, Quanbo GE, Junlong ZHU, Zhenyu LU. Improved YOLOX object detection algorithm based on gradient difference adaptive learning rate optimization [J]. Acta Aeronautica et Astronautica Sinica, 2023, 44(14): 327951-327951.
[7]	Lei DONG, Hongbing CHEN, Xi CHEN, Changxiao ZHAO. Distributed multi-agent coalition task allocation strategy for single pilot operation mode based on DQN [J]. Acta Aeronautica et Astronautica Sinica, 2023, 44(13): 327895-327895.
[8]	HU Wei, WAN Wenzhang, CHEN Mou. Neural network and disturbance observer based control for automatic carrier landing of UAV [J]. ACTA AERONAUTICAET ASTRONAUTICA SINICA, 2022, 43(S1): 726963-726963.
[9]	CHEN Bo, YUE Kai, WANG Rusheng, HU Mingnan. Learning-based multi-rate multi-sensor fusion localization method [J]. ACTA AERONAUTICAET ASTRONAUTICA SINICA, 2022, 43(S1): 726904-726904.
[10]	WANG Yufei, LIU Xiaojia, LIU Huan, CAO Lijun, LUO Zhiqiang. Application of Siamese neural network in electrical performance test of aerospace products [J]. ACTA AERONAUTICAET ASTRONAUTICA SINICA, 2022, 43(S1): 727048-727048.
[11]	WANG Ziling, XIONG Zhenyu, GU Xiangqi. Correlation learning algorithm of visible light and SAR cross modal remote sensing images [J]. ACTA AERONAUTICAET ASTRONAUTICA SINICA, 2022, 43(S1): 727239-727239.
[12]	LIU Jiaqi, FENG Yunwen, LU Cheng, XUE Xiaofeng, PAN Weihuang. Safety analysis of aero-engine operation based on intelligent neural network [J]. ACTA AERONAUTICAET ASTRONAUTICA SINICA, 2022, 43(9): 625375-625375.
[13]	HAN Songyu, SHAO Haidong, JIANG Hongkai, ZHANG Xiaoyang. Intelligent fault diagnosis of aero-engine high-speed bearings using enhanced CNN [J]. ACTA AERONAUTICAET ASTRONAUTICA SINICA, 2022, 43(9): 625479-625479.
[14]	YANG Te, YANG Zhichun, LIANG Shuya, KANG Zaifei, JIA You. Feature extraction and identification of stationary random dynamic load using deep neural network [J]. ACTA AERONAUTICAET ASTRONAUTICA SINICA, 2022, 43(9): 225952-225952.
[15]	SHAO Yiwei, CHEN Jiayu, LIN Cuiying, WAN Cheng, GE Hongjuan, SHI Zhilong. Gearbox fault diagnosis with small training samples: An improved deep forest based method [J]. ACTA AERONAUTICAET ASTRONAUTICA SINICA, 2022, 43(8): 625429-625429.

A neural network model for impact point prediction of ballistic missile based on improved second-order optimizer with parallel learning

RichHTML

PDF (PC)

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 23

References 44

Related Articles 15

Recommended Articles

Metrics

Comments