Multi-step-ahead Prediction of Grenade Trajectory Based on CNN-LSTM Enhanced by Deep Learning and Self-attention Mechanism

doi:10.12382/bgxb.2024.0659

Abstract

Abstract:

Since the grenade flight trajectory presents the characteristics of complexity, temporal variability and sudden change, it brings great challenges to the close-range air-defense interception system. Aiming at the current problem that the spatio-temporal features of trajectory data are hardly captured and only a small number of steps can be predicted, a trajectory multi-step-ahead prediction model based on 1-dimension convolutional neural network-long short-term memory-attention (1D CNN-LSTM-ATT) with the introduction of self-attention mechanism is proposed. The proposed model is compared and analyzed with CNN-LSTM and LSTM models in single-step and multi-step predictions, respectively, and realizes a high-precision real-time multi-step-ahead prediction of target trajectory from T moment to any T+K future moments. The study shows that the evaluation indexes of the 1D CNN-LSTM-ATT model are significantly better than those of the other two models for both single-step and multi-step prediction; the cumulative prediction error of 500 steps (i.e., 10s) of the 1D CNN-LSTM-ATT model is 82.83m in the direction of range, 11.68m in the direction of altitude, and 0.07m in the direction of transverse deviation, which provides an important guarantee of timely response to the implementation of the projectile interception. This provides an important guarantee for the timely response to the implementation of projectile interception.

Key words: trajectory multi-step-ahead prediction, deep learning, self-attention mechanism, CNN-LSTM model

CLC Number:

TJ301

SUN Xichen, LI Weibing, HUANG Changwei, FU Jiawei, FENG Jun. Multi-step-ahead Prediction of Grenade Trajectory Based on CNN-LSTM Enhanced by Deep Learning and Self-attention Mechanism[J]. Acta Armamentarii, 2024, 45(S1): 51-59.

Figures/Tables 15

References 20

[1]	XIAN Y, REN L L, XU Y J, et al. Impact point prediction guidance of ballistic missile in high maneuver penetration condition[J]. Defence Technology, 2023, 26(3): 213-230.
[2]	TIAN Z G, YANG K P, DANINO M, et al. Launch point estimation with a single passive sensor without trajectory state estimation[J]. IEEE Transactions on Aerospace and Electronic Systems, 2021, 58(1): 318-327.
[3]	MONDOLONI S, ROZEN N. Aircraft trajectory prediction and synchronization for air traffic management applications[J]. Progress in Aerospace Sciences, 2020, 119: 100640.
[4]	胡星志, 王旭, 江雄, 等. 基于高斯过程回归的高超声速飞行器不确定轨迹预测[J]. 空天技术, 2022(4): 49-61.
	HU X Z, WANG X, JIANG X, et al. Uncertain trajectory prediction of hypersonic flight vehicles based on Gaussian process regression[J]. Aerospace Technology, 2022(4): 49-61. (in Chinese)
[5]	胡豹, 高永卫, 昔华倩. 基于误差反向传播神经网络的机弹分离轨迹预测研究[J]. 气动研究与试验, 2024, 2(2): 59-65.
	HU B, GAO Y W, XI H Q. Study on trajectory prediction of store separation based on error back-propagation neural network[J]. Aerodynamic Research & Experiment, 2024, 2(2): 59-65. (in Chinese)
[6]	肖宗鑫, 李晓杰, 肖宗烁, 等. 基于RBF神经网络优化的无人驾驶车辆增量线性模型预测轨迹跟踪控制研究[J]. 重庆理工大学学报(自然科学版), 2021, 35(3): 36-45.
	XIAO Z X, LI X J, XIAO Z S. Research on trajectory tracking control of driverless vehicles with incremental linear model predictive based on RBF neural network optimization[J]. Journal of Chongqing Institute of Technology, 2021, 35(3): 36-45. (in Chinese)
[7]	谭紫阳, 高忠文, 邓宇. 基于改进极限学习机和深度神经网络融合的车辆轨迹长期预测[J]. 汽车技术, 2020(11): 16-20. doi: 10.19620/j.cnki.1000-3703.20191396
	TAN Z Y, GAO Z W, DENG Y. The long term prediction of vehicle trajectory based on improved extreme learning machine and depth neural network fusion[J]. Automobile Technology, 2020(11): 16-20. (in Chinese)
[8]	RATHER A M, AGARWAL A, SASTRY V N. Recurrent neural network and a hybrid model for prediction of stock returns[J]. Expert Systems with Applications, 2015, 42(6):3234-3241
[9]	李明阳, 鲁之君, 曹东晶, 等. 一种基于Transformer编码器与LSTM的飞机轨迹预测方法[J]. 航天返回与遥感, 2024, 45(2): 163-176.
	LI M Y, LU Z J, CAO D J, et al. A predictive aircraft trajectory prediction method based on transformer encoder and LSTM[J]. Spacecraft Recovery & Remote Sensing, 2024, 45(2): 163-176. (in Chinese)
[10]	任济寰, 吴祥, 薄煜明, 等. 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. doi: 10.12382/bgxb.2021.0489
	REN J H, WU X, BO Y M, et al. Ballistic trajectory prediction based on context-enhanced long short-term memory network[J]. Acta Armamentarii, 2023, 44(2): 462-471. (in Chinese) doi: 10.12382/bgxb.2021.0489
[11]	张宏鹏, 黄长强, 唐上钦, 等. 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(9): 1894-1903. doi: 10.3969/j.issn.1000-1093.2020.09.022
	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
[12]	YIN X Y, WU G Z, WEL J Z, et al. Multi-stage attention spatial-temporal graph networks for traffic prediction[J]. Neurocomputing, 2020, 428(6): 42-53.
[13]	郑志伟, 管雪元, 傅健, 等. 基于卷积神经网络与长短期记忆神经网络的弹丸轨迹预测[J]. 兵工学报, 2023, 44(10): 2975-2983. doi: 10.12382/bgxb.2022.0511
	ZHENG Z W, GUAN X Y, FU J, et al. Projectile trajectory prediction based on CNN-LSTM model[J]. Acta Armamentarii, 2023, 44(10): 2975-2983. (in Chinese) doi: 10.12382/bgxb.2022.0511
[14]	MA L, TIAN S. A hybrid CNN-LSTM model for aircraft 4D trajectory prediction[J]. IEEE Access, 2020, 8(8): 134668-134680.
[15]	孔建国, 李亚彬, 张时雨, 等. 基于CNN-LSTM-attention模型航迹预测研究[J]. 航空计算技术, 2023, 53(1): 1-5.
	KONG J G, LI Y B, ZHANG S Y, et al. Research on track prediction based on CNN-LSTM-attention model[J]. Aeronautical Computing Technique, 2023, 53(1): 1-5. (in Chinese)
[16]	苗靖, 李晓婷. 基于Seq2Seq-Att的船舶轨迹预测算[J]. 火力与指挥控制, 2024, 49(4): 71-76.
	MIAO J, LI X T. Ship trajectory prediction algorithm based on Seq2Seq-Att[J]. Fire Control & Command Control, 2024, 49(4): 71-76. (in Chinese)
[17]	李战武, 张帅, 乔英峰, 等. 基于自注意力机制和 CNN-LSTM 的空战目标机动轨迹预测[J]. 兵器装备工程学报, 2023, 44(7):209-216.
	LI Z W, ZHANG S, QIAO Y F, et al. Maneuvering trajectory prediction of air combat targets based on self-attention mechanism and CNN-LSTM[J]. Journal of Ordnance Equipment Engineering, 2023, 44(7): 209-216. (in Chinese)
[18]	YOON K. Convolutional neural networks for sentence classification[C]// Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP). Doha, Qatar: Association for Computational Linguistics, 2014:1746-1751.
[19]	LE Q V, JAITLY N, HINTONG G E. A simple way to initialize recurrent networks of rectified linear units[J]. Computer Science, 2015, 1504.00941.
[20]	WERNERT P, THEODOULIS S, MOREL Y. Flight dynamics properties of 155 mm spin-stabilized projectiles analyzed in different body frames[C]//Proceedings of the Atmospheric Flight Mechanics Conference. Toronto, Ontario, Canada:AIAA, 2010:1-17.

参数	采样值	采样间隔
发射角/(°)	[30,60]	2
速度/(m·s^-1)	[600,900]	10

参数	采样值	采样间隔
发射角/(°)	[30,60]	2
速度/(m·s^-1)	[600,900]	10

参数	采样值
迭代次数	50
学习率	0.0001
批处理量	256
优化器	Adam

参数	采样值
迭代次数	50
学习率	0.0001
批处理量	256
优化器	Adam

模型	参数	层信息
	LSTM层	神经元数量32
	Dropout	dropout率0.2
LSTM	LSTM层	神经元数量32
	Dropout	dropout率0.2
	全连接层	神经元数量3
	1D Conv	卷积核数量16,长度2
	1D Conv	卷积核数量64,长度2
	LSTM层	神经元数量32
CNN-LSTM	Dropout	dropout率0.2
	LSTM层	神经元数量32
	Dropout	dropout率0.2
	全连接层	神经元数量3
	1D Conv	卷积核数量16,长度2
	1D Conv	卷积核数量64,长度2
	LSTM层	神经元数量32
CNN-LSTM-ATT	Dropout	dropout率0.2
	LSTM层	神经元数量32
	Dropout	dropout率0.2
	Attention	-
	全连接层	神经元数量3