洛馬「AI格鬥俱樂部」驗證自主空戰　模擬114年高強度作戰

照片

洛馬「AI格鬥俱樂部」驗證自主空戰　模擬114年高強度作戰(圖片取自官方網站)

建立虛擬戰場環境　加速AI自主作戰系統發展

美國軍工巨擘洛克希德馬丁（Lockheed Martin）近日公布「AI格鬥俱樂部」（AI Fight Club）大型模擬測試成果，透過高擬真虛擬空戰環境，完成相當於114年的人工智慧（AI）戰爭測試，加速自主作戰系統與AI空戰能力發展。

US defense giant Lockheed Martin has revealed results from its large-scale “AI Fight Club” simulation exercise, using a highly realistic virtual air combat environment to conduct the equivalent of 114 years of artificial intelligence warfare testing and accelerate the development of autonomous combat systems and AI-driven aerial warfare capabilities.

模擬540兆美元作戰成本　虛擬損失1800萬架戰機

洛馬表示，相關測試在高強度空戰條件下，模擬超過540兆美元作戰成本，以及多達1800萬架軍機的虛擬損失，用以驗證AI在極端戰場環境中的決策能力與生存效率。

Lockheed stated that the simulations modeled more than 540 trillion US dollars in operational costs and the virtual loss of up to 18 million aircraft during high-intensity air combat conditions, aiming to validate AI decision-making and survivability in extreme battlefield environments.

高擬真模擬框架　專注壓力測試AI決策能力

報導指出，上述數據並非實際軍事成本或損失估算，而是來自高擬真模擬框架，用於在複雜空戰情境中，持續對AI自主決策系統進行壓力測試。

Reports noted that the figures were not intended to represent real-world military costs or losses, but instead came from a high-fidelity simulation framework designed to continuously stress-test autonomous AI decision-making systems in complex aerial combat scenarios.

　洛馬強調，該系統可提供具規模化能力的測試環境，藉此觀察AI系統在長時間、高壓作戰下的行為模式與任務表現。

Lockheed emphasized that the system provides a scalable testing environment capable of observing AI behavior patterns and mission performance under prolonged, high-pressure combat conditions.

4對4空戰驗證　AI操控第4代戰機對抗

此次「AI格鬥俱樂部」演習由洛馬攜手Ansys Government Initiatives與ATG公司共同執行，安排多組AI代理系統，操作模擬中的第4代戰機進行4對4空戰。

The “AI Fight Club” exercise was jointly conducted by Lockheed Martin, Ansys Government Initiatives, and ATG, with multiple AI agent systems controlling simulated fourth-generation fighter aircraft in 4-versus-4 aerial combat engagements.

　測試期間，系統主要驗證AI自主代理在作戰環境中的任務規劃、交戰決策、損失控制與協同行動能力。

During the exercise, the systems primarily evaluated AI autonomous agents in mission planning, engagement decision-making, loss mitigation, and coordinated operational behavior within contested combat environments.

「Cogniverse」模擬空戰環境　重現真實戰場複雜性

所有測試情境均在洛馬開發的「Cogniverse」模擬環境中進行；該系統專門設計用來重現現代戰術空戰的複雜性與動態變化。

All simulation scenarios were conducted within Lockheed Martin’s “Cogniverse” synthetic environment, which was specifically designed to replicate the complexity and dynamic nature of modern tactical aerial combat.

　Ansys Government Initiatives副總裁佩達奇（Mike Pedaci）表示，結合高可信度任務級模擬與AI分析能力，能讓軍方在部隊實際部署前，先行探索作戰情境、預測結果並驗證作戰方案。

Mike Pedaci, vice president of Ansys Government Initiatives, stated that combining trusted mission-level simulations with AI-driven analytics allows military planners to explore combat scenarios, anticipate outcomes, and validate operational concepts before actual force deployment.

未來擴展海空太空領域　朝聯合作戰發展

洛馬表示，「AI格鬥俱樂部」計畫最早於2025年提出，定位為符合美國國防部標準的AI自主系統驗證平臺，讓各家企業能在接近真實作戰條件下測試自主作戰能力。

Lockheed noted that the “AI Fight Club” initiative was first introduced in 2025 as a Department of Defense-aligned proving ground for autonomous AI systems, enabling companies to test their technologies under realistic operational conditions.

　ATG作戰總監麥克唐納（Scott MacDonald）指出，此次演習展現未來國防自主系統發展方向，即「大規模、以模擬為核心、且高度協同化」。

ATG Director of Operations Scott MacDonald stated that the exercise demonstrated the future direction of defense autonomy development: “large-scale, simulation-first, and highly collaborative.”

　未來版本預計將擴展至本土防衛、海上作戰，以及太空情監偵等領域，逐步邁向「聯合全域作戰」（Joint All-Domain Operations）架構。

Future iterations are expected to expand into homeland defense, maritime warfare, and space-based ISR missions, gradually evolving toward a Joint All-Domain Operations framework.