基于交替循环神经网络的水下防御态势预测方法

doi:10.19678/j.issn.1000-3428.0065183

摘要/Abstract

摘要：

时间序列在军事战术对抗等领域应用广泛，通过战场观测到的时间序列态势信息预测对抗目标的趋势是制定决策方案的重要前提。以潜艇防御声自导鱼雷攻击为背景，针对环境不明、目标不明、解算要素不明等导致水下态势获取困难的问题，提出一种基于循环神经网络的交替不完全时间序列预测方法。融合单变量自回归横向趋势与外部趋势，对观测变量进行趋势预测。针对多变量间存在非线性复杂关系的特点，将纵向多变量间关系特征提取与横向趋势预测相结合，采用插补数据集训练与趋势输出预测的方式实现纵向预测。最后，使用加权分配器对横向趋势预测和纵向预测结果进行融合，提高模型对不完全时序态势的学习能力，实现对未来态势的预测。实验结果表明，在不完全信息条件下，所提方法在缺失30%和60%数据、3种不同预测时间窗口长度条件下，在仿真数据集和电力公测数据集上的平均均方误差、平均绝对误差均达到了最优或次优结果，能够实现态势评估，为充分利用并融合多特征态势数据进行决策提供科学有效的参考。

关键词: 交替循环神经网络, 缺失数据, 水下防御, 注意力机制, 自回归

Abstract:

Time series are widely used in military tactics confrontation and other fields. Predicting the trend of confrontation targets through the time series situation information observed in the battlefield is an important prerequisite for decision-making plans. Based on the background of submarine defense against acoustic homing torpedo attack, an alternative incomplete time series prediction method based on a recurrent neural network is proposed to solve the problems of unknown environment, unknown target, unknown solution elements, etc. The proposed method integrates univariate autoregressive lateral trend prediction with external trends to perform trend prediction on observed variables. In response to the characteristics of nonlinear and complex relationships between multiple variables, the feature extraction of vertical multivariate relationships is combined with horizontal trend prediction, and longitudinal prediction is achieved through interpolation dataset training and trend output prediction. Finally, a weighted allocator is used to fuse the horizontal and vertical trend prediction results, improving the learning ability of incomplete time series situations and enabling prediction of future situations. The experimental results show that under the conditions of 30% and 60% missing data and three different lengths of prediction time window, the proposed method achieves optimal or near-optimal Mean Square Error(MSE)and Mean Absolute Error(MAE) on the simulation data set and electric power open test data set, can realize situation assessment, and can provide more effective decision-making reference for making full use of and fusing multi-feature situation data.

Key words: alternating recurrent neural network, missing data, underwater defense, attention mechanism, auto regression

杨静, 陆铭华, 马洁琼, 吴金平, 刘星璇. 基于交替循环神经网络的水下防御态势预测方法[J]. 计算机工程, 2023, 49(9): 69-78.

Jing YANG, Minghua LU, Jieqiong MA, Jinping WU, Xingxuan LIU. Underwater Defense Posture Prediction Method Based on Alternating Recurrent Neural Network[J]. Computer Engineering, 2023, 49(9): 69-78.

http://www.ecice06.com/CN/Y2023/V49/I9/69

图/表 11

图1 潜艇防御鱼雷示意图

Fig.1 Schematic diagram of submarine defense torpedo

图2 交替循环神经网络的结构

Fig.2 Structure of alternating recurrent neural network

图3 基于自回归的横向预测

Fig.3 Horizontal prediction based on auto regression

图4 基于横向填充的纵向长序列预测流程

Fig.4 Procedure of vertical long sequence prediction based on horizontal filling

图5 60%数据缺失条件下ETT-m数据集的预测结果对比

Fig.5 Comparison of prediction results of ETT-m date set under 60% data missing conditions

图6 60%数据缺失条件下仿真数据集的预测结果对比

Fig.6 Comparison of prediction results of simulation data set under 60% data missing conditions

图7 30%数据缺失条件下不同数据集的预测结果

Fig.7 Prediction results of different data set under 30% missing data conditions

图8 不同损失函数对应的训练损失值

Fig.8 Training loss values corresponding to different loss functions

参考文献 26

1	陈颜辉, 张永. 来袭鱼雷弹道两种预测方法仿真分析. 水下无人系统学报, 2021, 29 (3): 357- 362. URL
	CHEN Y H, ZHANG Y. Simulation analysis on two methods for incoming torpedo trajectory prediction. Journal of Unmanned Undersea Systems, 2021, 29 (3): 357- 362. URL
2	黄文斌, 陈颜辉, 薛昌友. 来袭鱼雷类型识别指标提取与算法设计. 南京理工大学学报, 2011, 35 (2): 199- 203. doi: 10.3969/j.issn.1005-9830.2011.02.010
	HUANG W B, CHEN Y H, XUE C Y. Indices extraction and algorithm design for recognizing type of incoming torpedo. Journal of Nanjing University of Science and Technology, 2011, 35 (2): 199- 203. doi: 10.3969/j.issn.1005-9830.2011.02.010
3	程健, 张会. 潜艇使用声抗器材防御鱼雷方案优化模型及模型求解策略. 指挥控制与仿真, 2019, 41 (6): 48- 51. doi: 10.3969/j.issn.1673-3819.2019.06.009
	CHENG J, ZHANG H. Optimal model and model solving strategy of submarine torpedo defence using acoustic countermeasure equipment. Command Control & Simulation, 2019, 41 (6): 48- 51. doi: 10.3969/j.issn.1673-3819.2019.06.009
4	TAYLOR S J, LETHAM B. Forecasting at scale. The American Statistician, 2018, 72 (1): 37- 45. doi: 10.1080/00031305.2017.1380080
5	ARIYO A A, ADEWUMI A O, AYO C K. Stock price prediction using the ARIMA model[C]//Proceedings of the 16th International Conference on Computer Modelling and Simulation. Washington D. C., USA: IEEE Press, 2015: 106-112.
6	魏紫薇, 屈丹, 柳聪. 基于连接注意力的行人重识别特征提取方法. 计算机工程, 2022, 48 (7): 220- 226. URL
	WEI Z W, QU D, LIU C. Method for person re-identification feature extraction based on connected attention. Computer Engineering, 2022, 48 (7): 220- 226. URL
7	HOCHREITER S, SCHMIDHUBER J. Long short-term memory. Neural Computation, 1997, 9 (8): 1735- 1780. doi: 10.1162/neco.1997.9.8.1735
8	CHEN W R, MCCOOL E N, SUN L L, et al. Evaluation of machine learning models for proteoform retention and migration time prediction in top-down mass spectrometry. Journal of Proteome Research, 2022, 21 (7): 1736- 1747. doi: 10.1021/acs.jproteome.2c00124
9	QIN Y, SONG D, CHEN H, et al. A dual-stage attention-based recurrent neural network for time series prediction[EB/OL]. [2022-06-06]. https://arxiv.org/abs/1704.02971.
10	BELTAGY I, PETERS M E, COHAN A. Longformer: the long-document transformer[EB/OL]. [2022-06-06]. https://arxiv.org/abs/2004.05150.
11	郭奉琦, 孟凡荣, 王志晓. 基于变分自编码器的谣言立场分类算法. 计算机工程, 2022, 48 (2): 99- 105. URL
	GUO F Q, MENG F R, WANG Z X. Rumor stance classification algorithm based on variational auto-encoder. Computer Engineering, 2022, 48 (2): 99- 105. URL
12	WEN Q, ZHOU T, ZHANG C, et al. Transformers in time series: a survey[EB/OL]. [2022-06-06]. https://arxiv.org/abs/2202.07125.
13	LAI G, CHANG W, YANG Y, et al. Modeling long-and short-term temporal patterns with deep neural networks[EB/OL]. [2022-06-06]. https://arxiv.org/abs/1703.07015.
14	ZUO S, JIANG H, LI Z, et al. Transformer hawkes process[EB/OL]. [2022-06-06]. https://arxiv.org/abs/2002.09291.
15	KARIM F, MAJUMDAR S, DARABI H. Insights into LSTM fully convolutional networks for time series classification. IEEE Access, 2019, 7, 67718- 67725. doi: 10.1109/ACCESS.2019.2916828
16	SALINAS D, FLUNKERT V, GASTHAUS J, et al. DeepAR: probabilistic forecasting with autoregressive recurrent networks. International Journal of Forecasting, 2020, 36 (3): 1181- 1191. doi: 10.1016/j.ijforecast.2019.07.001
17	WU Y, LOBATO J M H, GHAHRAMANI Z. Dynamic covariance models for multivariate financial time series[EB/OL]. [2022-06-06]. https://www.mendeley.com/catalogue/d75844e4-f4db-3cc2-8af6-d5df1f1f23e7/.
18	CHENG J Z, HUANG K Z, ZHENG Z B. Towards better forecasting by fusing near and distant future visions. Proceedings of the AAAI Conference on Artificial Intelligence, 2020, 36 (11): 3593- 3600.
19	李晓, 卢先领. 基于双重注意力机制和GRU网络的短期负荷预测模型. 计算机工程, 2022, 48 (2): 291-296, 305. URL
	LI X, LU X L. Method for forecasting short-term power load based on dual-stage attention mechanism and gated recurrent unit network. Computer Engineering, 2022, 48 (2): 291-296, 305. URL
20	ZHU X F, YANG J Y, ZHANG C Y, et al. Efficient utilization of missing data in cost-sensitive learning. IEEE Transactions on Knowledge and Data Engineering, 2021, 33 (6): 2425- 2436. doi: 10.1109/TKDE.2019.2956530
21	SHUKLA S N, MARLIN B M. Heteroscedastic temporal variational autoencoder for irregularly sampled time series[EB/OL]. [2022-06-06]. https://arxiv.org/abs/2107.11350.
22	ZHOU H, ZHANG S, PENG J, et al. Informer: beyond efficient transformer for long sequence time-series forecasting[EB/OL]. [2022-06-06]. https://arxiv.org/abs/2012.07436.
23	李长文, 初磊. 理想声自导鱼雷运动要素的距离方位观测解算方法. 弹道学报, 2016, 28 (3): 43-47, 58. URL
	LI C W, CHU L. Method of solving motion factors of ideal acoustic homing torpedo by distance and bearing observations. Journal of Ballistics, 2016, 28 (3): 43-47, 58. URL
24	赵建昕, 笪良龙. 基于异方差伪线性模型的纯方位目标要素估计. 探测与控制学报, 2017, 39 (2): 7- 12. URL
	ZHAO J X, DA L L. Target bearings-only elements estimation based on heteroscedastic pseudo-linear model. Journal of Detection & Control, 2017, 39 (2): 7- 12. URL
25	李雯, 迟利华, 张会, 等. 潜艇使用自航式声诱饵防御鱼雷模型并行计算方法研究. 计算机工程与科学, 2020, 42 (12): 2163- 2168. URL
	LI W, CHI L H, ZHANG H, et al. Research on the parallel computing method of submarine's torpedo defense model with self-propelled acoustic decoy. Computer Engineering and Science, 2020, 42 (12): 2163- 2168. URL
26	CHUNG J, GULCEHRE C, CHO K, et al. Empirical evaluation of gated recurrent neural networks on sequence modeling[EB/OL]. [2022-06-06]. https://arxiv.org/abs/1412.3555.

[1]	孙龙, 张荣芬, 刘宇红, 饶庭漓. 监控视角下密集人群口罩佩戴检测算法[J]. 计算机工程, 2023, 49(9): 313-320.
[2]	苏晓东, 李世洲, 赵佳圆, 亮洪宇, 张玉荣, 徐红岩. 基于多级叠加和注意力机制的图像语义分割[J]. 计算机工程, 2023, 49(9): 265-271, 278.
[3]	邵良杉, 赵松泽. 基于多模型融合的不完整数据分数插补算法[J]. 计算机工程, 2023, 49(9): 79-88,98.
[4]	韩璐, 霍纬纲, 张永会, 刘涛. 基于多尺度特征融合与双注意力机制的多元时间序列预测[J]. 计算机工程, 2023, 49(9): 99-108.
[5]	龙玉江, 卫薇, 舒彧, 张正刚, 王道累, 李峰. 基于自适应关键点的破损旋转绝缘子检测方法[J]. 计算机工程, 2023, 49(9): 272-278.
[6]	丰芳宇, 罗晓曙, 蒙志明, 王广宇. 基于抗混叠残差注意力网络的人脸表情识别[J]. 计算机工程, 2023, 49(8): 190-198.
[7]	王书朋, 何引弟. 融合特征注意力机制的非均匀光照图像增强算法[J]. 计算机工程, 2023, 49(8): 232-239.
[8]	刘昊鑫, 董超, 勾智楠, 高凯. 融合混合表征的小样本关系抽取方法[J]. 计算机工程, 2023, 49(8): 63-68.
[9]	杨长沛, 廖列法. 基于门控空洞卷积特征融合的中文命名实体识别[J]. 计算机工程, 2023, 49(8): 85-95.
[10]	刘俊豪, 王美林, 谢兴, 宋烨兴, 许莉花. 基于改进YOLOv5的皮革瑕疵检测算法[J]. 计算机工程, 2023, 49(8): 240-249.
[11]	马娜, 温廷新, 贾旭, 李晓会. 复杂光照条件下自适应的车脸重识别模型[J]. 计算机工程, 2023, 49(8): 275-282, 290.
[12]	陈露萌, 曹彦彦, 黄民, 谢鑫钢. 基于改进YOLOv5的火焰检测方法[J]. 计算机工程, 2023, 49(8): 291-301, 309.
[13]	李强龙, 周新文, 位梦恩, 甘阳洲. 基于条形池化和注意力机制的街道场景红外目标检测算法[J]. 计算机工程, 2023, 49(8): 310-320.
[14]	张家熔, 苑津莎, 许珈宁, 罗志宏. 基于多元信息嵌入与协同神经网络的力学实体识别算法[J]. 计算机工程, 2023, 49(7): 125-134.
[15]	白明昌. 基于折叠路径聚合的属性网络节点嵌入方法[J]. 计算机工程, 2023, 49(7): 76-84.

选择文件类型/文献管理软件名称

选择包含的内容