| 185 | 0 | 131 |
| 下载次数 | 被引频次 | 阅读次数 |
围绕世界模型技术在智能化战争中的应用需求与典型场景展开系统评述，并对未来发展方向进行前瞻性探讨。首先，从战场态势理解与预测、作战方案仿真推演、有人/无人行动规划与协同三个维度，系统归纳了世界模型技术在军事作战中潜在的主要运用模式；随后，深入探讨了支撑世界模型军事运用的关键技术体系，涵盖场景理解与感知预测、动力学建模与仿真推演、行动规划与决策优化等核心维度；最后，对未来世界模型技术军事应用亟需突破的重点方向给出建议。研究表明，世界模型技术通过构建“感知-预测-决策”的完整认知闭环，有望将作战决策周期从传统的“周/天”量级压缩至“小时/分钟”量级，进而形成对传统作战节奏的“代差式”超越。这一技术范式的变革，正在推动作战体系从“数据驱动”的被动响应向“模型驱动”的前瞻预判跃升。
Abstract：This paper systematically reviews the application requirements and typical scenarios of world model technology in intelligent warfare and provides a forward-looking discussion of its future development directions. First， the potential and main operational application patterns of world model technology are systematically summarized from three dimensions: battlefield situation understanding and prediction， operational plan simulation and wargaming， and manned-unmanned operation planning and collaboration. Furthermore， the key technical systems underpinning military world model applications are explored in depth， encompassing core aspects such as scene understanding and perceptual prediction， dynamics modeling and simulation rollout， as well as action planning and decision optimization. Finally， recommendations are proposed for key directions demanding urgent breakthroughs in the future military applications of world model technology. It is demonstrated that world model technology， by constructing a complete cognitive closed loop of “perception-prediction-decision”， is expected to compress the operational decision-making cycle from the traditional magnitude of weeks or days to hours or minutes， thereby achieving a generational leap over traditional operational tempos. This paradigm shift in decision-making is driving operational systems to evolve from data-driven reactive responses toward model-driven proactive anticipation.
[1] Vincent C. The U.S. DoD’s AI acceleration strategy [EB/OL]. 2026-01-22. https://dsm.forecastinternational.com/2026/01/22/the-u-s-dods-ai-acceleration-strategy.
[2] Michael P. US Air Force wants AI to power high-speed wargaming [EB/OL]. 2025-12-10. https://www.defensenews.com/air/2025/12/09/us-air-force-wants-ai-to-power-high-speed-wargaming/.
[3] Cristina S. US Air Force debuts operational AI wargame system [EB/OL]. 2026-04-17. https://www.c4isrnet.com/industry/techwatch/2026/04/15/us-air-force-debuts-operational-ai-wargame-system/.
[4] Ben A. DARPA’s DRBE to revolutionize electronic warfare training [EB/OL]. 2026-08-13. https://defenceinnovationreview.com/2025/08/13/darpas-drbe-to-revolutionize-electronic-warfare-training/.
[5] Li X， He X， Zhang L， et al. A comprehensive survey on world models for embodied AI[J/OL]. 2026-06-25. https://doi.org/10.48550/arXiv.2510.16732.
[6] Hu Y， Wang L， Liu X， et al. Simulating the real world: A unified survey of multimodal generative models[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence， 2026.
[7] Zidan A H， Pan Y， Jiang H， et al. World models: A comprehensive survey of architectures， methodologies， reasoning paradigms， and applications [J/OL]. 2026-05-28. https://doi.org/10.48550/arXiv.2606.00133.
[8] 赵翔，黑梦哲，李家旭，等.基于大模型的世界模型研究综述（特邀）[J].计算机工程，2026，52(6):1-16.
[9] The Investor. NC AI wins defense AI project with Hyundai Rotem [EB/OL]. 2026-05-28. https://m.theinvestor.co.kr/article/10758105.
[10] DARPA. ERIS special topic area – DARPAVERSE [EB/OL]. 2025-11-13. https://sam.gov/workspace/contract/opp/e2b84cd5ac614ef0affe1bf2c57d0e35/view.
[11] DARPA. Critical orientation of mathematics to produce advancements in science and security (COMPASS) workshop [EB/OL]. 2025-03-05. https://sam.gov/opp/66657ec7d1d945bb801f2d54f04192ad/view.
[12] NATO M&S COE. Modelling and Simulation for Autonomous System Conference - MESAS 2025 [EB/OL]. 2025-10-23. https://www.mscoe.org/modelling-and-simulation-for-autonomous-system-conference-mesas-2025/.
[13] Danielle M. Dimitris Giovanis receives DARPA grant to advance mathematical approaches for military decision-making using wargaming simulations [EB/OL]. 2025-09-15. https://engineering.jhu.edu/case/news/dimitris-giovanis-receives-darpa-grant-to-advance-mathematical-approaches-for-military-decision-making-using-wargaming-simulations/.
[14] 刘阳，秦刘震，张洋.美以伊冲突无人机运用情况及分析[J].战术导弹技术，2026，(3):18-25+158.DOI:10.16358/j.issn.1009-1300.20260601.
[15] Assran M， Duval Q， Misra I， et al. Self-supervised learning from images with a joint-embedding predictive architecture[C].Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. Vancouver， BC， Canada: IEEE， 2023: 15619-15629.
[16] He K， Chen X， Xie S， et al. Masked autoencoders are scalable vision learners[C].Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). New Orleans， LA， USA: IEEE， 2022: 16000-16009.
[17] Zhou G， Pan H， LeCun Y， et al. DINO-WM: World models on pre-trained visual features enable zero-shot planning[C].Proceedings of the 42nd International Conference on Machine Learning (ICML). Vancouver， BC， Canada: 2025.
[18] Karypidis E， Kakogeorgiou I， Gidaris S， et al. DINO-Foresight: Looking into the future with DINO[C].Advances in Neural Information Processing Systems (NeurIPS). San Diego， CA， USA / Mexico City， Mexico: NeurIPS Foundation， 2025.
[19] Wu T， Zhu B， Zhang R， et al. Pairwise proximal policy optimization: Language model alignment with comparative RL[C].First Conference on Language Modeling. Philadelphia， PA， USA: COLM， 2024.
[20] Wang G， Tang P， Ren X， et al. Learning vision-language-action world models for autonomous driving[C].Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Denver， CO， USA: IEEE， 2026: 1073-1084.
[21] Chen D， Moutakanni T， Chung W， et al. Planning with reasoning using vision language world model [J/OL]. 2025-09-06. https://doi.org/10.48550/arXiv.2509.02722.
[22] Jianan X， Yi G， Zihan L， et al. PAN: A world model for general， interactable， and long-horizon world simulation [J/OL]. 2026-02-03. https://doi.org/10.48550/arXiv.2511.01467.
[23] Wu T， Yang S， Po R， et al. Video world models with long-term spatial memory[C].Advances in Neural Information Processing Systems (NeurIPS). San Diego， CA， USA/Mexico City， Mexico: NeurIPS Foundation， 2025.
[24] Missal T， Domingues L， Guler B， et al. RopeDreamer: A kinematic recurrent state space model for dynamics of flexible deformable linear objects [J/OL]. 2026-04-30. https://doi.org/10.48550/arXiv.2604.28161.
[25] Gu A， Dao T. Mamba: Linear-time sequence modeling with selective state spaces [J/OL]. 2024-05-31. https://doi.org/10.48550/arXiv.2312.00752.
[26] Wu J， Yin S， Feng N， et al. Rlvr-world: Training world models with reinforcement learning[J]. Advances in Neural Information Processing Systems， 2025， 38: 125312-125350.
[27] Lanier J， Kim K， Karamzade A， et al. Adapting world models with latent-state dynamics residuals [J/OL]. 2025-04-03. https://doi.org/10.48550/arXiv.2504.02252.
[28] 赵云龙，齐亚莉，陆利坤.扩散模型研究综述：理论、改进、应用与展望[J/OL].计算机工程与应用，1-21[2026-07-21].https://link.cnki.net/urlid/11.2127.TP.20260610.1845.012.
[29] Fuest M， Ma P， Gui M， et al. Diffusion models and representation learning: A survey[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence， 2026 ， 48(7): 7209-7228.
[30] Huang S， Wu J， Zhou Q， et al. Vid2world: Crafting video diffusion models to interactive world models [J/OL]. 2026-03-08. https://doi.org/10.48550/arXiv.2505.14357.
[31] Xie K， Yang I， Gunerli J， et al. Making large language models into world models with precondition and effect knowledge[C].Proceedings of the 31st International Conference on Computational Linguistics. Abu Dhabi， UAE: Association for Computational Linguistics， 2025: 7532-7545.
[32] 牛莅原，游光荣，朱晓敏，等.基于蒙特卡洛树搜索与大语言模型智能体的战略博弈策略优化研究[J].军事运筹与评估，2025，40(4):28-34.
[33] Zhao Z， Fu T， Wang Y， et al. From forecasting to planning: Policy world model for collaborative state-action prediction[J]. Advances in Neural Information Processing Systems， 2026， 38: 134585-134611.
[34] Chen D， Moutakanni T， Chung W， et al. Planning with reasoning using vision language world model [J/OL]. 2025-09-06. https://doi.org/10.48550/arXiv.2509.02722.
[35] Ye S， Ge Y， Zheng K， et al. World action models are zero-shot policies [J/OL]. 2026-02-17. https://doi.org/10.48550/arXiv.2602.15922.
[36] Feng Q， Yu J， Liu J， et al. Harmowam: Harmonizing generalizable and precise manipulation via adaptive world action models [J/OL]. 2026-05-11. https://doi.org/10.48550/arXiv.2605.10942.
[37] Zhang X， Zeng X， Zhang W. From world models to world action models: A concise tutorial for robotics [J/OL]. 2026-08-12. https://doi.org/10.48550/arXiv.2607.00836.
[38] Cen J， Yu C， Yuan H， et al. Worldvla: Towards autoregressive action world model. [J/OL]. 2025-06-26. https://doi.org/10.48550/arXiv.2506.21539.
[39] Sanchez-Martin P， Rateike M， Valera I. Vaca: Design of variational graph autoencoders for interventional and counterfactual queries [J/OL]. 2024-02-04. https://doi.org/10.48550/arXiv.2110.14690.
[40] Nishimoto Y， Matsubara T. Object-centric world models for causality-aware reinforcement learning[C].Proceedings of the AAAI Conference on Artificial Intelligence. Singapore: AAAI Press， 2026， 40(29): 24585-24593.
[41] 王亚珅，张尚志，刘馨竹，等.2024年军事智能技术主要发展分析[J].战术导弹技术，2025，(2):44-63.
[42] Zhang W， Terver B， Zholus A， et al. Hierarchical planning with latent world models， [J/OL]. 2026-06-16. https://doi.org/10.48550/arXiv.2604.03208.
[43] Wang J， Li Y， Gao R X， et al. Hybrid physics-based and data-driven models for smart manufacturing: Modelling， simulation， and explainability[J]. Journal of Manufacturing Systems， 2022， 63: 381-391.
[44] Fang L， Yu X， Cai J， et al. Knowledge distillation and dataset distillation of large language models: Emerging trends， challenges， and future directions[J]. Artificial Intelligence Review， 2026， 59(1): 17.
[45] 李彩军，王亚珅，江昊，等.美以伊冲突中人工智能技术的应用分析及启示[J].战术导弹技术，2026，(3):1-9.
[46] Gu J， Jiang X， Shi Z， et al. A survey on llm-as-a-judge[J]. The Innovation， 2026， 7(6): 101253:1-101253:30.
基本信息:
DOI：10.16358/j.issn.1009-1300.20260142
中图分类号:E91
引用信息:
[1]相虎生,王亚珅,韩龙,等.智能化战争中世界模型技术运用的相关思考[J].战术导弹技术().DOI:10.16358/j.issn.1009-1300.20260142.
基金信息:
国家自然科学基金（U22B2061）; 重点实验室自主科研项目（2025-JCJQ-LC-007）
2026-09-21
2026-09-21
2026-09-21