计算机科学 ›› 2026, Vol. 53 ›› Issue (1): 39-50.doi: 10.11896/jsjkx.250400064
刘大勇1,2, 董志明1, 郭齐胜1, 高昂3, 邱雪欢1
LIU Dayong1,2, DONG Zhiming1, GUO Qisheng1, GAO Ang3, QIU Xuehuan1
摘要： 战术对抗仿真实验是作战分析、模拟训练和基于仿真的装备活动的核心手段,其智能化、自动化水平直接影响实验效能和战斗力的生成。针对传统仿真实验存在的实验设计、模型构建、导调控制和人机交互效率低等问题,参考MCP协议提出大模型赋能战术对抗仿真实验的体系架构。该架构包含基础层、工具资源层、AI Agent层、赋能层、应用层,这5层架构自顶向下牵引,自底向上逐层整合,可实现大小模型与数据资源和传统小模型的耦合聚合,并赋能基于仿真的各项军事活动。在此基础上,重点研究讨论了大模型赋能战术对抗仿真实验的具体路径:大模型赋能仿真实验设计,大模型赋能决策模型构建,大模型赋能导调控制。最后,分析了大模型赋能战术对抗仿真实验面临的挑战,并给出了相应的应对措施。
中图分类号:
| [1]LEI K H,LI Y L,SHEN X H.Research on the Application of Large Models in Military Action Planning[J].Military Digest,2025(5):32-35. [2]XU R M.Combat Modeling and Simulation[M].Beijing:Mili-tary Science Press,2012. [3]LI C,YANG B,WANG Y F,et al.Research on Army Combined Operations Experimentation Systems[M].Beijing:Ordnance Industry Press,2022. [4]SI G,WU X,WANG Y,et al.Introduction to Joint Operations Experimentation[M].Beijing:National Defense Industry Press,2024. [5]CAO Y H,GUAN Q B,BAI H B.Theory and Technology of Combat Experimentation[M].Beijing:National Defense Industry Press,2013. [6]HU JW.Combat Simulation Experimental Design and Analysis[M].Beijing:National Defense Industry Press,2010 [7]Palantir Technologies.AIP:AI Platform for Defense[EB/OL].(2023).https://www.palantir.com/solutions/defense/. [8]XIN X.Development Status of Generative Artificial Intelligence in the U.S.Military Command and Control Domain[N].China Aviation News,2024-06-18(3961). [9]Scale AI.Force-Multiply with Donovan in Action[EB/OL].[2025-04-07].https://scale.com/donovan. [10]GOECKS V G,WAYTOWICH N R.COA-GPT:GenerativePre-trained Transformers for Accelerated Course of Action Development in Military Operations[C]//Proceedings of the NATO Science and Technology Organization Symposium(ICMCIS).Koblenz,Germany:NATO Science and Technology Organization,2024. [11]SUN Y X,ZHAO J J,XIE Y X,et al.Self-generating war game AI:Task planning of double-layer agent based on large language model[J/OL].Control and Decision.https://doi.org/10.13195/j.kzyjc.2023.1497. [12]WENG L.LLM-powered Autonomous Agents[DB/OL].(2023-06-23)[2024-06-30].https://lilianweng.github.io/posts/2023-06-23-agent/. [13]SUN Y F,WU J,LI Z,et al.Research on Large Sequence Model for Wargame Simulation[C]//Proceedings of the 10th “Complexity of War and War Simulation” High-Level Academic Seminar.National Defense University Joint Warfare College,2023:30-37. [14]Yuanting Technology launches decision optimization big model and intelligent agent suite[EB/OL].(2025-04-03)[2025-04-07].https://mp.weixin.qq.com/s/_nzl_QNMOk6usdB9LplpHA. [15]YOUNGJOON L,TAEHYUN P,YUNHO L,et al.Exploring Potential Prompt Injection Attacks in Federated Military LLMs and Their Mitigation[J].arXiv:2501.18416,2025. [16]REN B C,QIAO S F,YU W H,et al.KnowRL:ExploringKnowledgeable Reinforcement Learning for Factuality[J].ar-Xiv:2506.19807v1,2025. [17]ALBERT T, SUN Z F,CHRISTOPHER D S.Model-Preserving Adaptive Rounding[J].arXiv:2505.22988v1,2025. [18] KIM J H,KIM J,KWON S,et al.KVzip:Query-Agnostic KV Cache Compression with Context Reconstruction[J].arXiv:2505.23416,2025. [19]DING X, SUN R, ZHANG Y J,et al.DipSVD:Dual-importance Protected SVD for Efficient LLM Compression[J].arXiv:2506.20353,2025. [20]WANG Z R,LAN T F,SU Z Y,et al.Towards Efficient LLM Storage Reduction via Tensor Deduplication and Delta Compression[J].arXiv:2505.06252,2025. [21]LIU X Y,WEN Z C,WANG S B,et al.Shifting AI Efficiency From Model-Centric to Data-Centric Compression[J].arXiv:2505.19147v1,2025. |
|
||
首页