指控领域数据中台的大语言模型应用设计

doi:10.12305/j.issn.1001-506X.2026.03.18

Abstract

Abstract:

In the practical application of command and control domain, the data middle platform still faces the issue of limited support for decision-making efficiency. To address this, an intelligent data application based on large language models is designed to effectively integrate data into the command and control process, ultimately enabling command decision-making with fragmented datas. Firstly, a hierarchical application architecture is established, and the command information flow is clarified. Then, a command and control domain specific large language model construction method and a multi-model collaboration architecture are proposed for the model layer. Implementation methods for four types of applications, including event element generation, are elaborated for the software layer. Finally, the effectiveness of the proposed method is demonstrated using the task platform as an example, providing reference for the construction of data middle platform and intelligent command.

Key words: data middle platform, large language model, command and control, intelligent command

CLC Number:

TP 399

Xinyuan QIU, Honghua CHEN, Xiaolong CUI, Min ZHANG, Xuanzhu SHENG. Design of large language model application of data middle platform for command and control[J]. Systems Engineering and Electronics, 2026, 48(3): 918-931.

Figures/Tables 18

Fig.1

Fig.2

Fig.3

Table 1

Table 2

Fig.4

Fig.5

Fig.6

Fig.7

Fig.8

Fig.9

Table 3

Comparative analysis of the proposed method and traditional methods"

应用	传统方法			本文方法
应用	原理	优势	劣势	原理	优势	劣势
事件要素生成	基于卷积神经网络、图神经网络等深度神经网络进行重要特征提取	1）过程可解释性强	1）模型训练依赖于大量标注数据 2）推理分析弱（仅依赖事件抽取模型无法直接生成摘要）	大模型基于语义理解能力，按照提示模板将多源情报信息转化为统一的事件表示形式	1）能够较好处理多源数据 2）支持多模态输入 3）可扩展性强（可输出Json 等多种格式）	1）可能出现事实性幻觉^[36]
知识图谱生成	基于支持向量机等统计模型依赖规则模板抽取实体、关系	1）资源消耗小 2）可解释性强 3）准确率高	1）高度依赖人工 2）可扩展性差（新领域需要专家重新设计规则） 3）需要注入大量领域知识数据	大模型进行文本拓展和自动化抽取三元组，适用于构建特定领域知识图谱	1）自动化程度高 2）泛化能力强 3）无需注入大量领域知识	1）算力资源消耗大 2）可能出现知识冗余、知识冲突
决策方案生成	暂无能够有效处理非结构化、半结构化决策问题的指控领域决策生成模型，主要通过业务人员人工撰写	—	—	大模型基于知识库数据和历史案例进行内容生成	1）高效的知识融合、信息处理 2）节省人力成本，可快速生成多套可参考的决策方案 3）格式高度规范	1）方案的可执行性有限 2）可能出现事实性幻觉
智能要图标绘	目前公开的军事标图自动化方法主要采用基于规则^[37]或模式^[38]的方法抽取标绘信息，而后业务人员基于此通过标绘软件实现标绘	1）相较手工作业，有效提升业务人员工作效率	1）预置规则复杂，涉及句型分析、分词处理、语义角色标注等 2）需要集成标绘软件，且需要人工编辑输入	大模型结合相似度计算，抽取出目标位置、兵力部署等要素，用作支持标绘工具生成态势图	1）显著提升业务人员工作效率；相较文献[37−38]，预置规则更简单 2）自动化程度提升：大模型输出结果可按标绘软件要素按序输出	1）涉及标号库、标绘软件等第三方专业工具 2）算力资源消耗大

Table 3

Table 4

Fig.10

Table 5

Fig.11

Fig.12

Table 6

References 38

1	阿里巴巴. 大数据之路-阿里巴巴大数据实践[M]. 北京: 电子工业出版社, 2017: 5−25.
	ALIBABA. The road to big data-Alibaba’s big data practice[M]. Beijing: Electronics Industry Press, 2017: 5−25.
2	吴信东, 应泽宇, 盛绍静, 等. 数据中台框架与实践[J]. 大数据, 2023, 9 (6): 137- 159.
	WU X D, YING Z Y, SHENG S J, et al. Data central-platform: architecture and practice[J]. Big data, 2023, 9 (6): 137- 159.
3	付登坡, 江敏, 赵东辉, 等. 数据中台: 让数据用起来[M]. 北京: 机械工业出版社, 2024: 5−25.
	FU D P, JIANG M, ZHAO D H, et al. Data middle office: make data valuable[M]. Beijing: China Machine Press, 2024: 5−25.
4	郭建胜, 顾涛勇, 张晓丰. 军事信息系统中台建设[J]. 国防科技, 2021, 42 (2): 2- 7. doi: 10.13943/j.issn1671-4547.2021.02.01
	GUO J S, GU T Y, ZHANG X F. Middle platform in military information systems[J]. National Defense Technology, 2021, 42 (2): 2- 7. doi: 10.13943/j.issn1671-4547.2021.02.01
5	李文俊, 杨学强, 杜家兴. 基于数据中台的装备保障数据集成[J]. 系统工程与电子技术, 2020, 42 (6): 1317- 1323. doi: 10.3969/j.issn.1001-506X.2020.06.15
	LI W J, YANG X Q, DU J X. Equipment support data integration based on ODPS[J]. Systems Engineering and Electronics, 2020, 42 (6): 1317- 1323. doi: 10.3969/j.issn.1001-506X.2020.06.15
6	刘东波, 罗常伟. 基于软件定义的任务规划中台架构设计[J]. 火力与指挥控制, 2024, 49 (5): 111- 116，123.
	LIU D B, LUO C W. Architecture design for mission planning middle platform based on software defined technology[J]. Fire Control & Command Control, 2024, 49 (5): 111- 116，123.
7	BROWN T B, MANN B, RYDER N, et al. Language models are few-shot learners[C]//Proc. of the 34th International Conference on Neural Information Processing Systems, 2020: 1877−1901.
8	蒲云强, 唐川, 徐婧, 等. 基于大语言模型的科技动态情报感知研究[J]. 情报理论与实践, 2025, 48 (2): 11- 20.
	PU Y Q, TANG C, XU J, et al. Research on technology trends intelligence perception using large language models[J]. Information Studies:Theory & Application, 2025, 48 (2): 11- 20.
9	张华平, 李春锦, 魏顺平, 等. 大语言模型驱动的开源情报认知[J]. 国防科技, 2024, 45 (3): 51- 57. doi: 10.13943/j.issn1671-4547.2024.03.08
	ZHANG H P, LI C J, WEI S P, et al. Large language model-driven open-source intelligence cognition[J]. National Defense Technology, 2024, 45 (3): 51- 57. doi: 10.13943/j.issn1671-4547.2024.03.08
10	赵浜, 曹树金. 国内外生成式AI大模型执行情报领域典型任务的测试分析[J]. 情报资料工作, 2023, 44 (5): 6- 17. doi: 10.12154/j.qbzlgz.2023.05.001
	ZHAO B, CAO S J. Test analysis of typical tasks in the information field performed by generative large models at home and abroad[J]. Information and Documentation Services, 2023, 44 (5): 6- 17. doi: 10.12154/j.qbzlgz.2023.05.001
11	孙怡峰, 廖树范, 吴疆, 等. 基于大模型的态势认知智能体[J]. 指挥控制与仿真, 2024, 46(2): 1−7.
	SUN Y F, LIAO S F, WU J, et al. Research on situation awareness agent based on large models[J]. Command and Control & Simulation, 46(2): 1−7.
12	张洪广, 杨林, 杨雄军, 等. 大模型驱动的智能辅助决策原理与典型应用[J]. 指挥与控制学报, 2024, 10 (6): 661- 668.
	ZHANG H G, YANG L, YANG X J, et al. Principle and typical implementation of foundation model-based intelligent aided decision making systems[J]. Journal of Command and Control, 2024, 10 (6): 661- 668.
13	丁滟, 王怀民, 史佩昌, 等. 可信云服务[J]. 计算机学报, 2015, 38 (1): 133- 149.
	DING Y, WANG H M, SHI P C, et al. Trusted cloud service[J]. Chinese Journal of Computers, 2015, 38 (1): 133- 149.
14	郑少秋, 韩立斌, 王静, 等. 军事大数据概念内涵、发展挑战与技术实践[J]. 指挥与控制学报, 2023, 9 (5): 508- 516. doi: 10.3969/j.issn.2096-0204.2023.05.0508
	ZHENG S Q, HAN L B, WANG J, et al. Conceptual connotation, development challenges and technical practices of military big data[J]. Journal of Command and Control, 2023, 9 (5): 508- 516. doi: 10.3969/j.issn.2096-0204.2023.05.0508
15	MOHAMMED S, STEPHEN W. Domain driven design vs soft-domain driven design frameworks[J]. International Journal of Computer, Electrical, Automation, Control and Information Engineering, 2016, 10 (7): 1364- 1370.
16	ZENG A H, LIU X, DU Z X, et al. GLM-130B: an open bilingual pre-trained model[EB/OL]. [2024-10-20]. https: //arxiv.org/abs/2210.02414.
17	BAI J Z, BAI S, CHU Y F, et al. Qwen technical report[EB/OL]. [2024-10-20]. https: //arxiv.org/abs/2309.16609.
18	YANG A Y, XIAO B, WANG B N, et al. Baichuan 2: open large-scale language models[EB/OL]. [2024-10-20]. https: //arxiv.org/abs/2309.10305.
19	LUO J P, YU H Y, TAN C, et al. Enhanced Qwen-VL 7b model via instruction finetuning on chinese medical dataset[C]//Proc. of the 5th International Conference on Computer Engineering and Application, 2024: 526−530.
20	HUANG Y, BAI Y, ZHU Z, et al. C-EVAL: a multilevel multi-discipline Chinese evaluation suite for foundation models[C]//Proc. of the Advances in Neural Information Processing System, 2023.
21	XU L, LI A Q, ZHU L, et al. SuperCLUE: a comprehensive Chinese large language model benchmark[EB/OL]. [2024-10-20]. https: //arxiv.org/abs/2307.15020.
22	TOUVRON H, MARTIN L, STONE K, et al. Llama 2: open foundation and fine-tuned chat models[EB/OL]. [2024-10-20]. https: //arxiv.org/abs/2307.09288.
23	EDWARD J H, YELONG S, PHILLIP W, et al. LoRA: low-rank adaptation of large language models[C]//Proc. of the International Conference on Learning Representations, 2022.
24	BIDERMAN D, PORTES J, ORTIZ J J G, et al. LoRA learns less and forgets less[EB/OL]. [2024-10-20]. https: //arxiv.org/abs/2405.09673.
25	RASCHKA S. Finetuning LLMs with LoRA and QLoRA: insights from hundreds of experiments[EB/OL]. [2024-10-22]. https: //lightning.ai/pages/community/lora-insights/.
26	ZENG A H, XU B, WANG B, et al. ChatGLM: a family of large language models from GLM-130B to GLM-4 all tools[EB/OL]. [2024-10-22]. https: //arxiv.org/abs/2406.12793.
27	WANG P, BAI S, TAN S N, et al. Qwen2-VL: enhancing vision-language model’s perception of the world at any resolution[EB/OL]. [2024-10-22]. https: //arxiv.org/abs/2409.12191.
28	WANG L, MA C, FENG X Y, et al. A survey on large language model based autonomous agents[J]. Front, 2024, 18 (6): 186- 345.
29	QU C L, DAI S H, WEI X C, et al. Tool learning with large language models: a survey[J]. Frontiers of Computer Science, 2025, 19, 198343. doi: 10.1007/s11704-024-40678-2
30	李特丽, 康轶文. LangChain入门指南[M]. 北京: 电子工业出版社, 2024: 12−64.
	LI T L, KANG Y W. LangChain: getting started guide[M]. Beijing: Electronics Industry Press, 2024: 12−64.
31	REYNOLDS L, MCDONELL K. Prompt programming for large language models: beyond the few-shot paradigm[C]//Proc. of the Computer-Human Interaction Conference on Human Factors in Computing Systems, 2021.
32	WEI J, WANG X, SCHUURMANS D, et al. Chain of thought prompting elicits reasoning in large language models[C]//Proc. of the Advances in Neural Information Processing Systems 35, 2022.
33	ZHANG Z, GAO J, DHALIWAL R S, et al. VISAR: a human-AI argumentative writing assistant with visual programming and rapid draft prototyping[C]//Proc. of the 36th Annual ACM Symposium on User Interface Software and Technology, 2023: 5−16.
34	姚奕, 陈朝阳, 杜晓明, 等. 多模态知识图谱构建技术及其在军事领域的应用综述[J]. 计算机工程与应用, 2024, 60 (22): 18- 37. doi: 10.3778/j.issn.1002-8331.2404-0285
	YAO Y, CHEN C Y, DU X M, et al. Survey of multimodal knowledge graph construction technology and its application in military field[J]. Computer Engineering and Applications, 2024, 60 (22): 18- 37. doi: 10.3778/j.issn.1002-8331.2404-0285
35	杨露菁, 陈志刚, 李煜, 等. 作战辅助决策理论及应用[M]. 3版. 北京: 国防工业出版社, 2023: 15−16.
	YANG L J, CHEN Z G, LI Y, et al. Theory and application of operational auxiliary decision-making [M]. 3rd ed. Beijing: National Defense Industry Press, 2023: 15−16.
36	FILIPPOVA K. Controlled hallucinations: learning to generate faithfully from noisy data[EB/OL]. [2024-10-20]. https: //arxiv.org/abs/2010.05873 2020.
37	杨健, 高文逸, 王衍波. 一种作战文书军事标图自动化方法[J]. 解放军理工大学学报(自然科学版), 2006(6): 543−547.
	YANG J, GAO W Y, WANG Y B. Documents based automated military plotting method[J] Journal of PLA University of Science and Technology, 2006(6): 543−547.
38	孙宏纲, 姚景顺, 闫国玉. 基于自然语言理解的军事自动标图系统[J]. 火力与指挥控制, 2005(S1): 27−29.
	SUN H G, YAO J S, YAN G Y. Research on military auto-plot system based on natural language understanding[J] Fire Control and Command Control, 2005(S1): 27−29.

超参数	取值
批量大小	1
梯度累积步数	16
日志记录步数	10
预热步数	20
存档步数	100
学习率	3e-4
训练回合	3
随机失活	0.05
秩	16
缩放更新系数	32

应用	传统方法			本文方法
应用	原理	优势	劣势	原理	优势	劣势
事件要素生成	基于卷积神经网络、图神经网络等深度神经网络进行重要特征提取	1）过程可解释性强	1）模型训练依赖于大量标注数据 2）推理分析弱（仅依赖事件抽取模型无法直接生成摘要）	大模型基于语义理解能力，按照提示模板将多源情报信息转化为统一的事件表示形式	1）能够较好处理多源数据 2）支持多模态输入 3）可扩展性强（可输出Json 等多种格式）	1）可能出现事实性幻觉^[36]
知识图谱生成	基于支持向量机等统计模型依赖规则模板抽取实体、关系	1）资源消耗小 2）可解释性强 3）准确率高	1）高度依赖人工 2）可扩展性差（新领域需要专家重新设计规则） 3）需要注入大量领域知识数据	大模型进行文本拓展和自动化抽取三元组，适用于构建特定领域知识图谱	1）自动化程度高 2）泛化能力强 3）无需注入大量领域知识	1）算力资源消耗大 2）可能出现知识冗余、知识冲突
决策方案生成	暂无能够有效处理非结构化、半结构化决策问题的指控领域决策生成模型，主要通过业务人员人工撰写	—	—	大模型基于知识库数据和历史案例进行内容生成	1）高效的知识融合、信息处理 2）节省人力成本，可快速生成多套可参考的决策方案 3）格式高度规范	1）方案的可执行性有限 2）可能出现事实性幻觉
智能要图标绘	目前公开的军事标图自动化方法主要采用基于规则^[37]或模式^[38]的方法抽取标绘信息，而后业务人员基于此通过标绘软件实现标绘	1）相较手工作业，有效提升业务人员工作效率	1）预置规则复杂，涉及句型分析、分词处理、语义角色标注等 2）需要集成标绘软件，且需要人工编辑输入	大模型结合相似度计算，抽取出目标位置、兵力部署等要素，用作支持标绘工具生成态势图	1）显著提升业务人员工作效率；相较文献[37−38]，预置规则更简单 2）自动化程度提升：大模型输出结果可按标绘软件要素按序输出	1）涉及标号库、标绘软件等第三方专业工具 2）算力资源消耗大

实验环境	GPU	CPU	内存/G	CUDA
服务端环境	V100S-PCIE-32G 双卡	Intel（R）Xeon（R） Platinum 8360H CPU @3.00GHz	377	12.4
用户端环境	NVDIA T1000 8G	11th Gen Intel（R） Core（TM） i7-11700 @2.50GHz	32	—

评价对象	评价因素	大模型能力类型	评价指标
事件要素生成应用	所抽取事件要素	总结	准确性、完整性
事件要素生成应用	所生成事件摘要	总结	准确性、简明性
知识图谱生成应用	所生成文本内容	拓展	准确性、全面性
	所抽取三元组要素	总结	准确性、完整性、规范性
	所呈现知识图谱	转换	准确性、完备性、一致性
决策方案生成应用	所生成方案内容	拓展	决策要素齐全性、术语专业性
决策方案生成应用	所生成方案内容	推理	数据可靠性、可参考性
智能要图标绘应用	所提取标绘要素	总结	队标号识别准确性、位置识别准确性
智能要图标绘应用	所呈现标图结果	转换	可用性、完整性

数据智能应用	综合标准分	大模型能力维度	大模型能力分数
事件要素生成应用	9.14	总结	9.14
知识图谱生成应用	7.96	拓展	7.81
		总结	8.16
		转换	8.11
决策方案生成应用	7.55	拓展	9.19
决策方案生成应用	7.55	推理	6.93
智能要图标绘应用	9.51	总结	9.31
智能要图标绘应用	9.51	转换	9.63

Design of large language model application of data middle platform for command and control

RichHTML

PDF (PC)

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 18

References 38

Related Articles 15

Recommended Articles

Metrics

Comments

[1]	Zhixin CHEN, Lingyun LU, Ling YE, Shaoqiu ZHENG, Fang WANG, Xin GAO. Research on digital twin for command and control system [J]. Systems Engineering and Electronics, 2025, 47(3): 912-922.
[2]	Jinzhi LU, Guoxin WANG, Xijin TANG, Junjie TANG, Yuejie WEN, Jian TANG, Yangyang ZHANG, Xiaoping LAN, Qi LIU, Junlin LI, Junda MA, Shouxuan WU, Xiaodu HU. MBSE approach for spatial intelligence [J]. Systems Engineering and Electronics, 2025, 47(12): 3877-3889.
[3]	Rujiang PAN, Jiongyi CHEN, Jianbo WANG, Zhemei FANG. Intelligent modeling method of system architecture view based on large language model [J]. Systems Engineering and Electronics, 2025, 47(12): 3912-3923.
[4]	Yun ZHONG, Lujun WAN, Peiyang YAO, Jieyong ZHANG. Measurement and optimization of generation and evolution of agile C2 organization [J]. Systems Engineering and Electronics, 2023, 45(4): 1090-1097.
[5]	Yongqi PAN, Wei WEI, Yi LIU, Cheng ZHU. Blockchain based method of dynamic command and control [J]. Systems Engineering and Electronics, 2022, 44(9): 2817-2825.
[6]	Yifan LI, Huaming QIAN, Hongzhong HUANG, Tingyu ZHANG, Tudi HUANG. Reliability analysis of command and control network system based on generalized continuous time Bayesian network [J]. Systems Engineering and Electronics, 2022, 44(12): 3880-3886.
[7]	Meigen HUANG, Weiping WANG, Tao WANG, Xiaobo LI, Hua HE, Song YANG. EC2-based architecture design method of cloud-flow C2 for unmanned combat SoS [J]. Systems Engineering and Electronics, 2022, 44(11): 3413-3422.
[8]	Jiayi LIU, Shaohua YUE, Gang WANG, Xiaoqiang YAO, Jie ZHANG. Cooperative evolution algorithm of multi-agent system under complex tasks [J]. Systems Engineering and Electronics, 2021, 43(4): 991-1002.
[9]	Liyao WU, Wei HAN, Yong ZHANG, Yao XIONG. Structure design of command and control system for manned/unmanned aerial vehicles formation [J]. Systems Engineering and Electronics, 2020, 42(8): 1826-1834.
[10]	Bingfeng GE, Boyuan XIA, Zhiwei YANG, Qingsong ZHAO, Hechuan WEI. Modeling and analysis of ExtendSim model and data-driven command and control processes [J]. Systems Engineering and Electronics, 2020, 42(5): 1063-1072.
[11]	SUN Peng, WU Junsheng, WANG Xun, JIAO Zhiqiang, ZHANG Jieyong. Dynamic resources scheduling method based on multi-objective optimization in C2 organization [J]. Systems Engineering and Electronics, 2019, 41(4): 793-800.
[12]	ZHANG Zhuang, LI Linlin, WEI Zhenhua, YU Hongfeng. Dynamic effectiveness evaluation of command and control system based on variable weight projection gray target [J]. Systems Engineering and Electronics, 2019, 41(4): 801-809.
[13]	WANG Xun, YAO Peiyang, ZHANG Jieyong, WAN Lujun, JIAO Zhiqiang, CHEN Jieyu. Research on problem of Holonic-C2 organization resource scheduling [J]. Systems Engineering and Electronics, 2019, 41(3): 564-570.
[14]	LI Linlin, LU Yunfei, ZHANG Zhuang, HE He. System construction and modeling of command and control system index based on information superiority [J]. Systems Engineering and Electronics, 2018, 40(3): 577-582.
[15]	SUN Peng, WU Junsheng, ZHANG Jieyong, DU Huahua. Modeling and solving method for the collocating problem of decision-making entity based on knowledge constraint [J]. Systems Engineering and Electronics, 2017, 39(9): 2039-2045.