Acta Aeronautica et Astronautica Sinica ›› 2026, Vol. 47 ›› Issue (15): 333092.doi: 10.7527/S1000-6893.2026.33092
• Electronics and Electrical Engineering and Control • Previous Articles
Yang LIU, Ni LI(
), Zexi YU, Shangjie JIA
Received:2025-11-17
Revised:2025-12-02
Accepted:2026-01-05
Online:2026-01-12
Published:2026-01-09
Contact:
Ni LI
E-mail:lini@buaa.edu.cn
Supported by:CLC Number:
Yang LIU, Ni LI, Zexi YU, Shangjie JIA. An intelligent modeling method for aircraft formation behaviors integrating LLM technology[J]. Acta Aeronautica et Astronautica Sinica, 2026, 47(15): 333092.
Table 1
Comparison of behavior modeling methods
| 建模方法 | 对抗性能 | 泛化能力 | 可解释性 | 建模成本 | 实时性 |
|---|---|---|---|---|---|
| 传统知识工程行为建模 | 中:规则驱动,稳定性较好,但复杂对抗适应性有限 | 低:跨场景通常需人工重构规则 | 高:决策逻辑清晰 | 高:规则设计与维护成本较高 | 高:决策计算开销小 |
| 基于优化算法的行为建模 | 中:可针对特定目标函数获得较优解 | 低:目标函数或约束变化需重新建模 | 中:目标与约束可解释,求解过程隐式 | 中:建模与求解成本随问题规模上升 | 低:在线优化计算开销较大 |
| 基于DRL的行为建模 | 高:在固定场景下可学习复杂对抗策略 | 中:对状态分布变化敏感,泛化能力受限 | 低:策略以隐式参数形式存在 | 高:训练样本与计算资源需求较大 | 中:推理阶段计算开销相对较小 |
| 基于LLM的行为建模 | 中:具备高层决策与协同推理能力 | 高:具备跨任务与场景迁移潜力 | 高:可通过语言与结构化推理进行解释 | 中:规则与目标函数设计减少,但需考虑模型适配与约束 | 低:需通过推理频率与模型规模进行控制 |
Table 3
Some single-aircraft tactical intentions and procedures
| 战术号 | 战术名称 | 战术意图 | 战术过程 |
|---|---|---|---|
| 1 | 搜索战术 | 在雷达未发现敌机时,对敌机进行搜索,确保远距时雷达可以锁定敌方 | 按照固定巡逻路径进行匀速平飞,在雷达成功锁定敌机后转向敌机并调整我机姿态稳定 |
| 2 | 正面强攻战术 | 朝向敌机飞行,建立进攻优势后发射导弹 | 在与敌机达到合适距离时,提高自身速度并转向敌机来袭方向,接着提升自身高度建立高度优势,进行发弹,而后立即俯冲,为规避敌方导弹做准备 |
| 3 | 回转脱离战术 | 回转偏离敌机方向,以提前规避来袭导弹 | 调整自身俯仰角,随后进行偏置机动,远离敌机方向 |
| 4 | 持续制导战术 | 转向敌机方向,持续为导弹提供导引信息 | 首先转弯至敌机方向,使敌机进入我机雷达扫描范围,接着保持朝着敌机方向飞行的情况下降低自身高度,为后续规避敌方导弹做准备 |
| 5 | 导弹规避战术 | 用于规避敌方已经锁定我机的导弹 | 立即偏离来袭导弹方向,转离后向远离导弹来袭方向进行大过载斜筋斗机动,通过自身海拔的升高消耗导弹最后阶段的动能 |
Table 4
‘State-action’ association for the Mopan tactic
| 状态名称 | 状态描述 | 粗粒度候选战术(Top-3) | 细粒度语义 相关性评分 |
|---|---|---|---|
| Patrol | 巡逻搜索 | 搜索战术 | 0.835 5 |
| 导弹规避战术 | 0.000 8 | ||
| 持续制导战术 | 0.001 7 | ||
| GuideMissile | 提供导弹制导 | 持续制导战术 | 0.919 2 |
| 导弹规避战术 | 0.000 2 | ||
| 正面强攻战术 | 0.000 5 | ||
| LaunchMissile | 发射导弹 | 正面强攻战术 | 0.828 5 |
| 导弹规避战术 | 0.168 2 | ||
| 回转脱离战术 | 0.033 2 | ||
| BreakAway | 回转脱离 | 回转脱离战术 | 0.744 0 |
| 导弹规避战术 | 0.003 5 | ||
| 持续制导战术 | 0.004 7 | ||
| EmergencyEvade | 紧急规避 | 导弹规避战术 | 0.408 8 |
| 正面强攻战术 | 0.125 8 | ||
| 搜索战术 | 0.014 9 |
Table 7
Results of different embedding models
| 序号 | 嵌入模型 | Hit@3 (粗筛) | MRR (粗筛) | MRR(引入细粒度匹配) |
|---|---|---|---|---|
| 1 | bge-small-zh | 0.94 | 0.781 7 | 0.878 3 |
| 2 | bge-base-zh | 0.96 | 0.808 3 | 0.881 7 |
| 3 | bge-large-zh | 0.96 | 0.825 0 | 0.881 7 |
| 4 | gte-small-zh | 0.96 | 0.808 3 | 0.888 3 |
| 5 | gte-base-zh | 0.90 | 0.826 7 | 0.840 0 |
| 6 | gte-large-zh | 0.88 | 0.716 7 | 0.815 0 |
| 7 | m3e-small | 0.92 | 0.736 7 | 0.845 0 |
| 8 | m3e-base | 0.96 | 0.805 0 | 0.900 0 |
| 9 | text2vec-chinese | 0.92 | 0.820 0 | 0.848 3 |
Table 8
Workflow and time comparison between intelligent and manual modeling methods
| 建模方法 | 建模阶段 | 所耗时间 | |
|---|---|---|---|
| 磨盘战术 | 诱敌战术 | ||
| 本文智能建模方法 | 1.清洗协同战术语义描述,编写为用户提示词 | 5.0 min | 3.0 min |
| 2.智能生成有人长机和无人僚机的状态机形式化语言 | 4.7 min | 2.9 min | |
| 3.将决策状态描述向量化并与单机战术行为库进行检索匹配 | 9.27 s | 7.55 s | |
| 4.行为模型仿真测试与人工微调 | 32.0 min | 13.0 min | |
| 总过程 | 41.7 min | 19.0 min | |
| 传统人工建模方法 | 1.分解协同战术语义描述,通过人工阅读/讨论方式完成任务分段 | 66.4 min | 27.8 min |
| 2.状态机概念模型构建,设计状态集与状态转移 | 114.2 min | 42.2 min | |
| 3.状态机形式化转化,将协同战术状态机模型编写为形式化语言 | 59.1 min | 18.7 min | |
| 4.决策状态与底层行为树逐条人工匹配与绑定 | 28.9 min | 9.9 min | |
| 5.行为模型仿真测试与人工微调 | 21.2 min | 15.3 min | |
| 总过程 | 289.8 min | 113.9 min | |
| [1] | 杜梓冰, 张立丰, 陈敬志, 等. 有人/无人机协同作战演示验证试飞关键技术[J]. 航空兵器, 2019, 26(4): 75-81. |
| DU Z B, ZHANG L F, CHEN J Z, et al. Critical technologies of demonstration flight test of cooperative operation for manned/unmanned aerial vehicles[J]. Aero Weaponry, 2019, 26(4): 75-81 (in Chinese). | |
| [2] | 张琪. 学习驱动的CGF决策行为建模方法研究[D]. 长沙: 国防科技大学, 2018. |
| ZHANG Q. Research on learning driven behavior modeling methods for decision making of computer generated forces(CGFs)[D]. Changsha: National University of Defense Technology, 2018 (in Chinese). | |
| [3] | BOURASSA N R. Modeling and simulation of fleet air defense systems using EADSIM[D]. Monterey: Naval Postgraduate School, 1993. |
| [4] | CLIVE P D, JOHNSON J A, MOSS M J, et al. Advanced framework for simulation, integration and modeling (AFSIM)[C]∥Proceedings of the International Conference on Scientific Computing (CSC), 2015: 73. |
| [5] | 王鹏, 刘昊雨, 李妮, 等. 基于改进ABC算法的有-无人协同空战行为建模[J]. 系统仿真学报, 2024, 36(12): 2871-2883. |
| WANG P, LIU H Y, LI N, et al. Behavioral modeling of manned-unmanned cooperative air combat based on improved ABC algorithm[J]. Journal of System Simulation, 2024, 36(12): 2871-2883 (in Chinese). | |
| [6] | 刘波, 魏潇龙, 屈虹, 等. 有人—无人机协同空战机动决策研究[J]. 航空工程进展, 2023, 14(6): 63-72. |
| LIU B, WEI X L, QU H, et al. Research on MAV-UAV cooperative air combat maneuver decision[J]. Advances in Aeronautical Science and Engineering, 2023, 14(6): 63-72 (in Chinese). | |
| [7] | MNIH V, KAVUKCUOGLU K, SILVER D, et al. Playing atari with deep reinforcement learning[DB/OL]. arXiv preprint: 1312.5602, 2013. |
| [8] | 单圣哲, 张伟伟. 基于自博弈深度强化学习的空战智能决策方法[J]. 航空学报, 2024, 45(4): 328723. |
| SHAN S Z, ZHANG W W. Air combat intelligent decision-making method based on self-play and deep reinforcement learning[J]. Acta Aeronautica et Astronautica Sinica, 2024, 45(4): 328723 (in Chinese). | |
| [9] | 李佐龙, 朱纪洪, 匡敏驰, 等. 基于混合动作的空战分层强化学习决策算法[J]. 航空学报, 2024, 45(17): 530053. |
| LI Z L, ZHU J H, KUANG M C, et al. Hierarchical decision algorithm for air combat with hybrid action based on deep reinforcement learning[J]. Acta Aeronautica et Astronautica Sinica, 2024, 45(17): 530053 (in Chinese). | |
| [10] | 王宇琨, 王泽, 董力维, 等. 基于分层的智能建模方法的多机空战行为建模[J]. 系统仿真学报, 2023, 35(10): 2249-2261. |
| WANG Y K, WANG Z, DONG L W, et al. Research on multi-aircraft air combat behavior modeling based on hierarchical intelligent modeling methods[J]. Journal of System Simulation, 2023, 35(10): 2249-2261 (in Chinese). | |
| [11] | 刘玮, 张永亮, 程旭. 基于深度强化学习的人机智能对抗综述[J]. 指挥信息系统与技术, 2023, 14(2): 28-37. |
| LIU W, ZHANG Y L, CHENG X. Survey of human-computer intelligence gaming based on deep reinforcement learning[J]. Command Information System and Technology, 2023, 14(2): 28-37 (in Chinese). | |
| [12] | 罗欣悦, 刘同, 陈文龙, 等. 无人集群通感算协同仿真系统综述:现状、技术与展望[J/OL]. 无线电工程, (2025-12-05)[2025-12-10]. . |
| LUO X Y, LIU T, CHEN W L, al et, Review of UAV swarm synaesthesia computing cooperative simulation system: Status, technology and prospect[J/OL]. Radio Engineering, (2025-12-05)[2025-12-10]. (in Chinese). | |
| [13] | 孙宇祥, 赵俊杰, 解宇轩, 等. 自生成兵棋AI: 基于大语言模型的双层Agent任务规划[J]. 控制与决策, 2024, 39(12): 3927-3936. |
| SUN Y X, ZHAO J J, XIE Y X, et al. Self generated wargame AI: Double layer agent task planning based on large language model[J]. Control and Decision, 2024, 39(12): 3927-3936 (in Chinese). | |
| [14] | 王彤, 赵美静, 徐沛, 等. 基于大语言模型的兵棋推演智能决策技术[J]. 自动化学报, 2025, 51(6): 1205-1217. |
| WANG T, ZHAO M J, XU P, et al. Decision technology based on large language model for wargame[J]. Acta Automatica Sinica, 2025, 51(6): 1205-1217 (in Chinese). | |
| [15] | 刘大勇, 董志明, 郭齐胜, 等. LLM赋能的战术兵棋决策Agent构建方法[J]. 系统仿真学报, 2026, 38 (3) : 758-775. |
| LIU D Y, DONG Z M, GUO Q S, et al. Construction approach of LLM-empowered tactical wargame decision-making agents[J]. Journal of System Simulation, 2026,38(3):758-775 (in Chinese). | |
| [16] | SAHOO P, SINGH A K, SAHA S, et al. A systematic survey of prompt engineering in large language models: Techniques and applications[DB/OL]. arXiv preprint: 2402. 07927, 2024. |
| [17] | LEWIS P, PEREZ E, PIKTUS A, et al. Retrieval-augmented generation for knowledge-intensive NLP tasks[C]∥Proceedings of the 34th International Conference on Neural Information Processing Systems. New York: ACM, 2020: 9459-9474. |
| [18] | 郜洪奎, 马瑞祥, 包骐豪, 等. 基于混合检索增强的双塔模型研究[J]. 计算机科学, 2025, 52(6): 324-329. |
| GAO H K, MA R X, BAO Q H, et al. Research on hybrid retrieval-augmented dual-tower model[J]. Computer Science, 2025, 52(6): 324-329 (in Chinese). | |
| [19] | DEEPSEEK-AI. DeepSeek-V3 technical report[DB/OL]. arXiv preprint: 2412.19437v2, 2025. |
| [20] | WANG Y X, SUN Q X, HE S C. M3 E: Moka massive mixed embedding model[EB/OL]. (2023-06-07)[2025-10-30]. . |
| [21] | CHEN J L, XIAO S T, ZHANG P T, et al. BGE M3-embedding: Multi-lingual, multi-functionality, multi-granularity text embeddings through self-knowledge distillation [DB/OL]. arXiv preprint: 2402.03216, 2024. |
| [22] | KIMI TEAM. Kimi K2: Open agentic intelligence[DB/OL]. arXiv preprint: 2507.20534, 2025. |
| [23] | QWEN TEAM. Qwen3 technical report[DB/OL]. arXiv preprint: 2505.09388, 2025. |
| [24] | 字节跳动Seed团队. 豆包Seed1.6模型技术介绍文档[EB/OL]. (2025-06-25)[2025-10-30]. . |
| BYTEDANCE SEED. Doubao Seed 1.6 large language model[EB/OL]. (2025-06-25)[2025-10-30] (in Chinese). | |
| [25] | XIAO S T, LIU Z, ZHANG P T, et al. C-pack: Packed resources for general Chinese embeddings[C]∥ Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval. New York: ACM, 2024: 641-649. |
| [26] | LI Z H, ZHANG X, ZHANG Y Z, et al. Towards general text embeddings with multi-stage contrastive learning [DB/OL]. arXiv preprint: 2308.03281, 2023. |
| [27] | XU M. Text2vec: A tool for text to vector[DB/OL]. [2025-10-30]. . |
| [1] | Jinli GUO, Mou CHEN. Finite-time inverse optimization composite control of high-speed aircraft formation [J]. Acta Aeronautica et Astronautica Sinica, 2026, 47(S1): 732952-732952. |
| [2] | Qiaoqiao ZHENG, Mou CHEN. Formation switching control for high speed aircraft based on performance function [J]. Acta Aeronautica et Astronautica Sinica, 2026, 47(S1): 732951-732951. |
| [3] | WU Huisong, LIN Qi, PENG Miaojiao, LIU Ting, JI Yangfeng, WANG Xiaoguang. Wire-driven parallel suspension mechanism for aircraft model of formation flight in low-speed wind tunnel [J]. ACTA AERONAUTICAET ASTRONAUTICA SINICA, 2019, 40(11): 123144-123144. |
| Viewed | ||||||
|
Full text |
|
|||||
|
Abstract |
|
|||||
Address: No.238, Baiyan Buiding, Beisihuan Zhonglu Road, Haidian District, Beijing, China
Postal code : 100083
E-mail:hkxb@buaa.edu.cn
Total visits: 6658907 Today visits: 1341All copyright © editorial office of Chinese Journal of Aeronautics
All copyright © editorial office of Chinese Journal of Aeronautics
Total visits: 6658907 Today visits: 1341

