基于深度置信网络的近距空战态势评估

张彬超; 寇雅楠; 邬蒙; 左家亮

doi:10.13700/j.bh.1001-5965.2016.0956

基于深度置信网络的近距空战态势评估

doi: 10.13700/j.bh.1001-5965.2016.0956

空军工程大学航空航天工程学院, 西安 710038

基金项目:

航空科学基金 20155896026

详细信息

作者简介:
张彬超男, 硕士研究生。主要研究方向:机器学习与智能空战

寇雅楠女, 博士, 副教授, 硕士生导师。主要研究方向:航空兵作战训练与效能评估

通讯作者:
寇雅楠, E-mail:49841256@qq.com

中图分类号: V271;TP391.4
计量
- 文章访问数: 1024
- HTML全文浏览量: 200
- PDF下载量: 526
- 被引次数: 0
出版历程
- 收稿日期: 2016-12-20
- 录用日期: 2017-02-15
- 网络出版日期: 2017-07-20

Close-range air combat situation assessment using deep belief network

College of Aeronautics and Astronautics Engineering, Air Force Engineering University, Xi'an 710038, China

Funds:

Aeronautical Science Foundation of China 20155896026

More Information

Corresponding author: KOU Yanan, E-mail: 49841256@qq.com

摘要

摘要:
针对传统态势评估方法权值确定困难、大规模数据处理和特征提取能力不足的问题，结合当前空战数据特征，将深度置信网络（DBN）应用于近距空战态势评估。通过密度峰值算法对空战特征数据进行聚类分析，并结合态势函数和专家判读进行修正，建立标准空战态势样本库；以重构误差和测试错误率为基础，建立网络拓扑结构和最优参数确定方法，提高模型的训练效率，并通过样本数据，对模型进行训练和验证。实验表明，模型态势分类正确率达到92.7%，模型运行时间满足应用需求，实例评估结果与客观态势一致性强。
- 深度置信网络（DBN） /
- 态势评估 /
- 半监督学习 /
- 网络拓扑结构 /
- 密度峰值聚类
Abstract:
Considering the difficulty in parameter setting, weakness of traditional situation assessment methods in processing and feature extraction of big data, feature of air combat data, applications of deep belief network (DBN) to close-range air combat situation assessment are discussed. A sample library of combat situation was constructed. The data were clustered using density peaks algorithm, and the results were revised by specialists of air combat and traditional functions. Then the model of deep belief network was constructed. According to the standard of test and reconstruction error, the network topology structure and optimal parameters were determined. The model was trained by the data from the sample library. Experimental results show that the model's situation classification accuracy reaches to 92.7%, and its running time meets the application requirements. Analysis of the practical example verified the feasibility of the DBN model.
- deep belief network (DBN) /
- situation assessment /
- semi-supervised learning /
- network topology structure /
- density peaks clustering

HTML全文

图 1 空战双方位置关系

Figure 1. Position relationship between two sides in air combat

下载: 全尺寸图片幻灯片

图 2 4种典型空战态势

Figure 2. Four kinds of typical air combat situations

下载: 全尺寸图片幻灯片

图 3 样本库构建流程图

Figure 3. Flowchart for constructing sample library

下载: 全尺寸图片幻灯片

图 4 密度峰值算法结果

Figure 4. Results of density peaks algorithm

下载: 全尺寸图片幻灯片

图 5 4类空战态势聚类结果

Figure 5. Clustering results of four kinds of air combat situations

下载: 全尺寸图片幻灯片

图 6 2种方法结果对比

Figure 6. Result comparison of two algorithms

下载: 全尺寸图片幻灯片

图 7 DBN结构

Figure 7. Structure of DBN

下载: 全尺寸图片幻灯片

图 8 RBM网络结构图

Figure 8. Structure of RBM network

下载: 全尺寸图片幻灯片

图 9 DBN深度确定流程图

Figure 9. Flowchart for determining depth of DBN

下载: 全尺寸图片幻灯片

图 10 不同深度DBN重构误差

Figure 10. Reconstruction error of different depth of DBN

下载: 全尺寸图片幻灯片

图 11 最后一层隐含层训练重构误差

Figure 11. Training reconstruction error of last hidden layer

下载: 全尺寸图片幻灯片

图 12 最后一层隐含层测试错误率

Figure 12. Test error rate of last hidden layer

下载: 全尺寸图片幻灯片

图 13 微调次数与测试错误率

Figure 13. Fine-tuning times and test error rate

下载: 全尺寸图片幻灯片

图 14 4类空战态势逐层特征提取线

Figure 14. Feature extraction lines per layer of four kinds of air combat situations

下载: 全尺寸图片幻灯片

图 15 两机空战对抗实例

Figure 15. Example of 1 versus 1 in air combat

下载: 全尺寸图片幻灯片

表 1 空战参数

Table 1. Parameters of air combat

参数	值域
目标方位角λ/rad	[0, π]
目标进入角ψ/rad	[0, π]
速度矢量夹角η/rad	[0, π]
两机距离r/km	[0, 10]
两机速度平方差Δv²
我机速度v_R
两机高度差Δh/m	[0, 18 000]
我机高度h/m	[0, 18 000]

下载: 导出CSV

表 2 不同深度训练结果

Table 2. Training result of different depth

网络深度	测试错误率/%	训练时间/s
1	10.12	266.758 7
2	9.46	431.971 7
3	8.72	578.341 0
4	8.40	933.181 1
5	9.88	1 269.745 0

下载: 导出CSV

表 3 3种算法效果对比

Table 3. Effect comparison of three algorithms

算法	训练集正确率/%(正确样本/训练样本)	测试集正确率/%(正确样本/测试样本)	运行总时间/s	单组数据运行时间/s	内存占用率峰值/%
BP神经网络	87.88	86.5	21.553 7	0.010 4	1.62
支持向量机	89.50	89.1	15.482 9	0.004 1	0.82
DBN	93.01	92.7	23.105 9	0.011 5	1.99

下载: 导出CSV

参考文献(20)

[1]	孙永芹, 世文荣, 彭海军, 等.基于改进非参量法的现代空战威胁评估研究[J].火控雷达技术, 2013, 42(3):28-33. http://www.cnki.com.cn/Article/CJFDTOTAL-HKLD201303007.htm SUN Y Q, SHI W R, PENG H J, et al.Study on improved non-parameter model based threat assessment in modern air combat [J].Fire Control Radar Technology, 2013, 42(3):28-33(in Chinese). http://www.cnki.com.cn/Article/CJFDTOTAL-HKLD201303007.htm
[2]	李望西, 黄长强, 王勇, 等.三维空间空战态势评估角度优势建模与仿真[J].电光与控制, 2012, 19(2):21-25. http://www.cnki.com.cn/Article/CJFDTOTAL-DGKQ201202008.htm LI W X, HUANG C Q, WANG Y, et al.Modeling and simulation of air combat situation assessment's angle superiority in three dimensional space[J].Electronics Optics and Control, 2012, 19(2):21-25(in Chinese). http://www.cnki.com.cn/Article/CJFDTOTAL-DGKQ201202008.htm
[3]	吴文海, 周思羽, 高丽, 等.基于导弹攻击区的超视距空战态势评估改进[J].系统工程与电子技术, 2011, 33(12):2679-2685. doi: 10.3969/j.issn.1001-506X.2011.12.20 WU W H, ZHOU S Y, GAO L, et al.Improvements of situation assessment for beyond-visual-range air combat based on missile launching envelope analysis[J].Systems Engineering and Electronics, 2011, 33(12):2679-2685(in Chinese). doi: 10.3969/j.issn.1001-506X.2011.12.20
[4]	顾佼佼, 刘卫华, 姜文志.基于攻击区和杀伤概率的视距内空战态势评估[J].系统工程与电子技术, 2015, 37(6):1306-1312. doi: 10.3969/j.issn.1001-506X.2015.06.13 GU J J, LIU W H, JIANG W Z.WVR air combat situation assessment model based on weapon engagement zone and kill probability[J].Systems Engineering and Electronics, 2015, 37(6):1306-1312(in Chinese). doi: 10.3969/j.issn.1001-506X.2015.06.13
[5]	胡云安, 刘振, 史建国.态势评估的变结构区间概率动态贝叶斯网络方法[J].系统工程与电子技术, 2013, 35(9):1891-1897. http://www.cnki.com.cn/Article/CJFDTOTAL-XTYD201309016.htm HU Y A, LIU Z, SHI J G.Situation assessment using variable structure interval probability dynamic Bayesian network[J].Systems Engineering and Electronics, 2013, 35(9):1891-1897(in Chinese). http://www.cnki.com.cn/Article/CJFDTOTAL-XTYD201309016.htm
[6]	史建国, 高晓光, 李相民.基于离散模糊动态贝叶斯网络的空战态势评估及仿真[J].系统仿真学报, 2006, 18(5):1093-1097. http://www.cnki.com.cn/Article/CJFDTOTAL-XTFZ200605002.htm SHI J G, GAO X G, LI X M.Modeling air combat situation assessment by using fuzzy dynamic Bayesian network [J].Journal of System Simulation, 2006, 18(5):1093-1097(in Chinese). http://www.cnki.com.cn/Article/CJFDTOTAL-XTFZ200605002.htm
[7]	王向华, 覃征, 刘宇, 等.径向基神经网络解决威胁排序问题[J].系统仿真学报, 2004, 16(7):1576-1579. http://www.cnki.com.cn/Article/CJFDTOTAL-XTFZ200407056.htm WANG X H, QIN Z, LIU Y, et al.RBF neural network for threat sequencing[J].Journal of System Simulation, 2004, 16(7):1576-1579(in Chinese). http://www.cnki.com.cn/Article/CJFDTOTAL-XTFZ200407056.htm
[8]	郭辉, 徐浩军, 刘凌.基于回归型支持向量机的空战目标威胁评估[J].北京航空航天大学学报, 2010, 36(1):123-126. http://bhxb.buaa.edu.cn/CN/abstract/abstract8578.shtml GUO H, XU H J, LIU L.Target threat assessment of air combat based on support vector machines for regression [J].Journal of Beijing University of Aeronautics and Astronautics, 2010, 36(1):123-126(in Chinese). http://bhxb.buaa.edu.cn/CN/abstract/abstract8578.shtml
[9]	张文忠, 孙永芹, 杨洪立, 等.基于Rough集和回归型SVM的超视距空战威胁评估[J].四川兵工学报, 2013, 34(7):14-18. http://www.cnki.com.cn/Article/CJFDTOTAL-CUXI201307005.htm ZHANG W Z, SUN Y Q, YANG H L, et al.Threat assessment based on rough set and support vector machines for regression in beyond-visual-range air combat [J].Journal of Sichuan Ordnance, 2013, 34(7):14-18(in Chinese). http://www.cnki.com.cn/Article/CJFDTOTAL-CUXI201307005.htm
[10]	尹宝才, 王文通, 王立春.深度学习研究综述[J].北京工业大学学报, 2015, 41(1):48-59. http://youxian.cnki.com.cn/yxdetail.aspx?filename=DSWZ201706161&dbname=CJFDPREN YIN B C, WANG W T, WANG L C.Review of deep learning[J].Journal of Beijing University of Technology, 2015, 41(1):48-59(in Chinese). http://youxian.cnki.com.cn/yxdetail.aspx?filename=DSWZ201706161&dbname=CJFDPREN
[11]	左家亮, 杨任农, 张滢, 等.基于模糊聚类的近距空战决策过程重构与评估[J].航空学报, 2015, 36(5):1650-1660. http://www.cnki.com.cn/Article/CJFDTOTAL-HKXB201505028.htm ZUO J L, YANG R N, ZHANG Y, et al.Reconstruction and evaluation of close air combat decision making process based on fuzzy clustering [J].Acta Aeronautica et Astronautica Sinica, 2015, 36(5):1650-1660(in Chinese). http://www.cnki.com.cn/Article/CJFDTOTAL-HKXB201505028.htm
[12]	马耀飞, 龚光红, 彭晓媛.基于强化学习的航空兵认知行为模型[J].北京航空航天大学学报, 2010, 36(4):379-383. http://bhxb.buaa.edu.cn/CN/abstract/abstract8461.shtml MA Y F, GONG G H, PENG X Y.Cognition behavior model for air combat based on reinforcement learning[J].Journal of Beijing University of Aeronautics and Astronautics, 2010, 36(4):379-383(in Chinese). http://bhxb.buaa.edu.cn/CN/abstract/abstract8461.shtml
[13]	朱建益. 空战中的威胁评估与态势评估研究[D]. 西安: 西安电子科技大学, 2013: 1-22. http: //cdmd. cnki. com. cn/Article/CDMD-10701-1013295653. htm ZHU J Y.A study of threat and situation assessment in air combat[D].Xi'an:Xidian University, 2013:1-22(in Chinese). http: //cdmd. cnki. com. cn/Article/CDMD-10701-1013295653. htm
[14]	ALEX R, ALESSANDRO L.Clustering by fast search and find of density peaks[J].Science, 2014, 344(6191):1492-1496. doi: 10.1126/science.1242072
[15]	HINTON G E, SALAKHUDINOV R R.Reducing the dimensionality of data with neural network[J].Science, 2006, 313(5786):504-507. doi: 10.1126/science.1127647
[16]	张春霞, 姬楠楠, 王冠伟.受限波尔兹曼机[J].工程数学学报, 2015, 32(2):159-173. http://www.cnki.com.cn/Article/CJFDTOTAL-GCSX201502001.htm ZHANG C X, JI N N, WANG G W.Restricted Boltzmann machines [J].Chinese Journal of Engineering Mathematics, 2015, 32(2):159-173(in Chinese). http://www.cnki.com.cn/Article/CJFDTOTAL-GCSX201502001.htm
[17]	黄海波, 李人宪, 杨琪, 等.基于DBN的车辆悬架减振器异响鉴别方法[J].西南交通大学学报, 2015, 50(5):776-782. http://www.cnki.com.cn/Article/CJFDTOTAL-XNJT201505002.htm HUANG H B, LI R X, YANG Q, et al.Identifying abnormal noise of vehicle suspension shock absorber based on deep belief networks[J].Journal of Southwest Jiaotong University, 2015, 50(5):776-782(in Chinese). http://www.cnki.com.cn/Article/CJFDTOTAL-XNJT201505002.htm
[18]	潘广源, 柴伟, 乔俊飞.DBN网络的深度确定方法[J].控制与决策, 2015, 30(2):256-260. http://www.cnki.com.cn/Article/CJFDTOTAL-KZYC201502010.htm PAN G Y, CHAI W, QIAO J F.Calculation for depth of deep belief network[J].Control and Decision, 2015, 30(2):256-260(in Chinese). http://www.cnki.com.cn/Article/CJFDTOTAL-KZYC201502010.htm
[19]	BENGIO Y. Learning deep architectures for AI [J].Foundations and Trends in Machine Learning, 2009, 2(1):1-127. doi: 10.1561/2200000006
[20]	周树森. 基于深度置信网络的分类方法[D]. 哈尔滨: 哈尔滨工业大学, 2012: 5-25. http: //d. wanfangdata. com. cn/Thesis/D419897 ZHOU S S.Deep belief networks based classification methods[D].Harbin:Harbin Institute of Technology, 2012:5-25(in Chinese). http: //d. wanfangdata. com. cn/Thesis/D419897