基于增强上下文信息长短期记忆网络的弹道轨迹预测

doi:10.12382/bgxb.2021.0489

摘要/Abstract

摘要：

根据己方观测数据进行弹道轨迹预测是现代陆军实施精准打击的重要一环。针对现有弹道轨迹预测方法存在精度不足且实时性不强的问题,提出一种新的增强上下文信息长短期记忆(CE-LSTM)网络轨迹预测模型,进行弹道轨迹的长期精准预测。在LSTM网络可逼近任意非线性函数且具备长期记忆能力的基础上,构建隐藏层输出混合单元提取短时上下文信息,进一步逼近弹体运动状态;通过建立不同条件下的弹道轨迹的数据集,训练得到具备最优超参数的CE-LSTM网络。实验结果表明,与弹道微分方程组的数值积分解法以及高斯混合模型相比,CE-LSTM网络在预测的精度上优于其他2种方法,预测速度提高了3~10倍,且具备较强的泛化能力。

关键词: 轨迹预测, 增强上下文信息, 长短期记忆网络, 弹道微分方程组, 高斯混合模型

Abstract:

Trajectory prediction based on observed data is critical to the modern army's precision strike capability. Yet, the existing trajectory prediction methods have suffer from low accuracy and poor real-time performance. Thus, this study proposes a new model called Context-enhanced Long Short-Term Memory(CE-LSTM) Network to make an accurate long-term prediction of the exterior trajectory of incoming projectiles. The proposed method inherits the LSTM network's advantage in that it can approximate any nonlinear function and it has a long-term memory. Furthermore, we create a mixture output unit of the hidden layer to extract short-term context information and approximate the motion states of the incoming projectiles more accurately. The CE-LSTM network is trained with a large-scale dataset consisting of exterior trajectories under different initial conditions to obtain the optimal hyper-parameters. The experimental results show that compared with the methods like external ballistic differential equations and the Gaussian mixture model, the CE-LSTM network performs significantly better in prediction accuracy, and its prediction speed increases by three to ten times. Moreover, the proposed method is highly generalizable.

Key words: trajectory prediction, context-enhanced, long short-term memory network, ballistic differential equations, gaussian mixture model

中图分类号:

TJ302

任济寰, 吴祥, 薄煜明, 吴盘龙, 何山. 基于增强上下文信息长短期记忆网络的弹道轨迹预测[J]. 兵工学报, 2023, 44(2): 462-471.

REN Jihuan, WU Xiang, BO Yuming, WU Panlong, HE Shan. Ballistic Trajectory Prediction Based on Context-enhanced Long Short-Term Memory Network[J]. Acta Armamentarii, 2023, 44(2): 462-471.

图/表 13

图1 LSTM单元

Fig.1 LSTM unit

图2 CE-LSTM单元

Fig.2 CE-LSTM unit

图3 模型训练及预测流程

Fig.3 Model training & prediction flowchart

表1 数据集构成

Table 1 Composition of datasets

口径/mm	风况	弹头初速/(m·s^-1)	高低角/(°)	数量/条
155	恒定随机线性	680~720	24~66	1500 1500 1500
122	恒定随机线性	670~710	24~66	1500 1500 1500
35	恒定随机线性	1030~1070	29~81	1500 1500 1500

图4 CE-LSTM网络结构

Fig.4 Schematic diagram of the CE-LSTM network

图5 不同数量神经元对应RMSE

Fig.5 RMSE with different number of neurons

图6 不同数量神经元对应损失

Fig.6 Loss with different number of neurons

图7 基于前100个坐标值的预测结果

Fig.7 Prediction results based on the first 100 coordinate values

图8 基于前50个坐标值的预测结果

Fig.8 Prediction results based on the first 50 coordinate values

图9 不同坐标轴上的绝对误差

Fig.9 Absolute errors for different coordinate axises

表2 不同时长坐标值下均方根误差表

Table 2 RMSE with different time-length coordinate values

模型	观测值数量	RMSE
模型	观测值数量	x轴	y轴	z轴	平均
CE-LSTM	50 100	6.909 6.830	0.254 0.173	3.717 3.711	3.627 3.571
GMM	50 100	64.107 61.765	1.435 1.352	35.784 36.613	33.775 33.243
弹道方程	50 100	41.955 24.916	3.970 0.378	19.946 8.905	21.957 11.399

图10 泛化能力交叉验证

Fig.10 Cross-validation for generalization capability

图11 预测平均用时

Fig.11 Average prediction duration

参考文献 20

[1]	薄煜明, 郭治, 钱龙军, 等. 现代火控理论与应用基础[M]. 北京: 科学出版社, 2012.
	BO Y M, GUO Z, QIAN L J, et al. Modern fire control theory and its application foundation[M]. Beijing: Science Press, 2012. (in Chinese)
[2]	何山, 吴盘龙, 李星秀, 等. 基于刚体弹道模型的防空火控解算方法[J]. 兵工学报, 2020, 41(8):1494-1501. doi: 10.3969/j.issn.1000-1093.2020.08.003
	HE S, WU P L, LI X X, et al. The antiaircraft fire control calculation method based on rigid body trajectory model[J]. Acta Armamentarii, 2020, 41(8):1494-1501. (in Chinese) doi: 10.3969/j.issn.1000-1093.2020.08.003
[3]	宋丕极. 枪炮与火箭外弹道学[M]. 北京: 兵器工业出版社, 1993.
	SONG P J. External ballistics of bullet,projectile and rocket[M]. Beijing: Publication House of Ordnance Industry, 1993. (in Chinese)
[4]	MANUEL S, DARIO A, ALESSIO B, et al. Recursive estimation for sparse Gaussian process regression[J]. Automatica, 2020, 120:109-127.
[5]	CHEN M H, YU C H, GAO J L, et al. Quantum algorithm for Gaussian process regression[J]. Physical Review A, 2022, 106(1):012406. doi: 10.1103/PhysRevA.106.012406 URL
[6]	SHUSTIN P F, AVRON H. Gauss-legendre features for Gaussian process regression[J]. Journal of Machine Learning Research, 2022, 23(92):1-47.
[7]	FU Y L, LIU X N, SURYADIPTO S, et al. Gaussian mixture model with feature selection: an embedded approach[J]. Computers & Industrial Engineering, 2021, 152:107000. doi: 10.1016/j.cie.2020.107000 URL
[8]	SAYAH M, GUEBLI D, NOUREDDINE Z, et al. Deep LSTM enhancement for RUL prediction using gaussian mixture models[J]. Automatic Control and Computer Sciences, 2021, 55(1):15-25. doi: 10.3103/S0146411621010089
[9]	AMIT S, PUSHPENDRA S, BREJESH L, et al. Modeling and prediction of COVID-19 pandemic using Gaussian mixture model[J]. Chaos,Solitons & Fractals, 2020, 138:110023. doi: 10.1016/j.chaos.2020.110023 URL
[10]	JIN H P, SHI L X, CHEN X G, et al. Probabilistic wind power forecasting using selective ensemble of finite mixture Gaussian process regression models[J]. Renewable Energy, 2021, 174:1-18. doi: 10.1016/j.renene.2021.04.028 URL
[11]	ZHANG Y, ZHENG X T, LU X Q. Pairwise comparison network for remote-sensing scene classification[J]. IEEE Geoscience and Remote Sensing Letters, 2022, 19:1-5.
[12]	张宏鹏, 黄长强, 唐上钦. 基于卷积神经网络的无人作战飞机飞行轨迹实时预测[J]. 兵工学报, 2020, 41(8):1894-1903.
	ZHANG H P, HUANG C Q, TANG S Q. CNN-based real-time prediction method of flight trajectory of unmanned combat aerial vehicle[J]. Acta Armamentarii, 2020, 41(9):1894-1903. (in Chinese) doi: 10.3969/j.issn.1000-1093.2020.09.022
[13]	PARK M K, LEE J M, KANG W H, et al. Predictive model for PV power generation using RNN (LSTM)[J]. Journal of Mechanical Science and Technology, 2021, 35(2):795-803. doi: 10.1007/s12206-021-0140-0
[14]	GREG V H, CARLOS M, GONZALO N. A review on the long short-term memory model[J]. Artificial Intelligence Review, 2020, 53(8):5929-5955. doi: 10.1007/s10462-020-09838-1
[15]	XIE A Q, YANG H, CHEN J, et al. A Short-term wind speed forecasting model based on a multi-variable long short-term memory network[J]. Atmosphere, 2021, 12(5):651. doi: 10.3390/atmos12050651 URL
[16]	LANDI F, BARADI L, CORNIA M, et al. Working memory connections for LSTM[J]. Neural Networks, 2021, 144:334-341. doi: 10.1016/j.neunet.2021.08.030 pmid: 34547671
[17]	ZHU Y, LIU J L, GUO C, et al. Prediction of battlefield target trajectory based on LSTM[C]// Proceedings of the 2020 IEEE 16th International Conference on Control & Automation. Singapore: IEEE, 2020:725-730.
[18]	SEPP H, JÜRGEN S. Long short-term memory[J]. Neural Computation, 1997, 9(8):1735-1780. doi: 10.1162/neco.1997.9.8.1735 pmid: 9377276
[19]	韩子鹏. 弹箭外弹道学[M]. 北京: 北京理工大学出版社, 2014:74-86.
	HAN Z P. Exterior ballistics of projectiles and rockets[M]. Beijing: Beijing Institute of Technology Press, 2014:74-86. (in Chinese)
[20]	MUKADDES Ö T, TURGUT Ö. Derivation of three-derivative Runge-Kutta methods[J]. Numerical Algorithms, 2017, 74(1):247-265. doi: 10.1007/s11075-016-0147-2 URL