DARPA’s AIR Program: AI Redefining the Future of Air Combat
DARPA’s AIR initiative is creating AI co-pilots that can adapt, scale, and team with humans — transforming how the skies will be fought and won.
Introduction: The Autonomous Combat Imperative
The future of aerial warfare is being rewritten by artificial intelligence, with DARPA’s Artificial Intelligence Reinforcements (AIR) program emerging as a pivotal initiative in this transformation. Modern combat environments present unprecedented challenges – increasing complexity, unpredictable threats, and the need for split-second decision-making in contested battlespaces. These operational realities demand a new paradigm in autonomous systems development, one that moves beyond controlled test environments to address the messy uncertainties of real-world combat.
The AIR program represents a strategic response to these challenges, focusing on developing AI systems capable of handling the dynamic complexities of actual combat scenarios. At its core, this initiative recognizes that maintaining air superiority against technologically advanced adversaries requires more than incremental improvements – it demands a fundamental rethinking of how autonomy integrates with tactical operations.
A key focus area involves beyond visual range (BVR) combat dominance. In these engagements where pilots cannot visually identify targets, AI-powered decision-making becomes paramount. The AIR program will train specialized AI agents to process vast streams of sensor data, anticipate enemy maneuvers, and recommend optimal weapon deployment strategies – all occurring in real-time combat scenarios.
The Core Challenges of Tactical Autonomy
Traditional approaches to autonomous systems have consistently encountered several fundamental limitations that the AIR program specifically targets for resolution. The most significant of these is the challenge of real-world operational complexity. Unlike controlled test environments with predefined parameters, actual combat scenarios involve a constantly evolving mix of known and unknown variables – rapidly changing conditions, incomplete information, and adversaries actively working to deceive and disrupt system operations. The AIR program’s emphasis on handling “open-world problems” represents a critical advancement, as it acknowledges and addresses situations where not all variables can be anticipated or predefined.
Another major hurdle lies in sensor integration and data processing. Effective tactical autonomy in modern combat requires the seamless fusion of data from multiple sensor systems operating in highly contested electromagnetic environments. These systems must maintain functionality despite sophisticated electronic warfare countermeasures while providing accurate, actionable intelligence. AIR’s focus on fully-integrated sensor systems aims to create AI architectures capable of synthesizing information from diverse sources while maintaining operational effectiveness amid these challenging conditions.
The program also confronts the critical issue of scalability. While current autonomous systems might demonstrate effectiveness in simple one-on-one engagements, real-world operations typically involve complex, multi-ship formations coordinating across expansive battlespaces. AIR specifically addresses the challenge of scaling autonomous behaviors to larger, more dynamic engagements while preserving the speed and accuracy of decision-making processes. This scalability factor becomes increasingly vital as the Department of Defense pursues concepts like manned-unmanned teaming and collaborative combat aircraft operations.
Technical Breakthroughs: The AIR Program’s Innovative Approach
The AIR program’s technical architecture represents a significant leap forward in military AI development through two interconnected research thrusts. The first focuses on predictive modeling with uncertainty quantification, developing advanced systems that go beyond simple outcome prediction to incorporate sophisticated certainty measurements. These models employ cutting-edge machine learning techniques to achieve continuous accuracy improvement as they process operational data, with built-in capabilities to explicitly represent and reason about uncertainty in their predictions. This includes the ability to adapt to new information about both friendly and adversary capabilities while accounting for potential deceptive effects and countermeasures – critical factors in real-world combat scenarios.
The second technical thrust develops the algorithmic foundation for real-time distributed autonomous execution in complex environments. This research area creates AI systems capable of making coordinated decisions across multiple autonomous platforms while maintaining effectiveness in communications-degraded or jammed environments. These systems demonstrate the flexibility to dynamically adapt to changing mission parameters and threats, with the capacity for rapid replanning as new information becomes available. The true innovation lies in how these technical areas interact – the predictive models inform autonomous behaviors, while operational feedback continuously improves the models, creating a virtuous cycle of enhancement and adaptation.
Human-Machine Collaboration: The AIR Advantage
The AIR program distinguishes itself through its sophisticated approach to human-machine collaboration, recognizing that optimal combat effectiveness comes from combining the strengths of both human operators and AI systems. Human warfighters bring irreplaceable domain knowledge, tactical intuition, and contextual understanding that current AI cannot replicate. The program’s structure deliberately incorporates this human expertise directly into the development process, ensuring the resulting systems align with actual operational needs and constraints rather than theoretical ideals.
This collaborative approach manifests in a rapid iteration cycle that contrasts sharply with traditional defense development timelines. Human experts systematically evaluate autonomous behaviors in simulated engagements, providing targeted feedback that drives focused algorithmic improvements. This creates an accelerated enhancement cycle that yields operational capabilities far faster than conventional approaches. Importantly, the program’s philosophy emphasizes that the end goal is not full autonomy but optimized human-machine teaming. The AI systems under development are designed to serve as intelligent aids that enhance human operators’ situational awareness and decision-making, not replace their critical judgment and command functions.
Implementation and Future Trajectory
The AIR program’s implementation strategy reflects its innovative technical approach, beginning with the November 2022 Proposers Day event that outlined DARPA’s vision to potential research partners. The program follows a phased development approach that emphasizes rapid prototyping and testing, establishing streamlined processes to quickly design, evaluate, and implement successive iterations of AIR software. This agile methodology stands in stark contrast to traditional defense acquisition timelines, enabling faster capability delivery without sacrificing rigor or operational relevance.
Transition planning forms a critical component of the program’s structure, with clear pathways established to move successful technologies into operational use. This includes specific plans for integration with next-generation platforms like the Air Force’s Next Generation Air Dominance (NGAD) system and Collaborative Combat Aircraft (CCA) program. The program deliberately designs its outputs to be compatible with both existing and planned defense systems, ensuring seamless incorporation into the broader military ecosystem. Looking beyond its initial aerial combat focus, the core technologies developed under AIR possess significant potential for expansion across multiple domains including naval warfare systems, ground combat operations, multi-domain command and control architectures, and advanced electronic warfare systems.
Lockheed Martin’s Pivotal Role in AI Combat Development
Lockheed Martin has been selected to spearhead a crucial 18-month effort under the AIR program, focusing on AI-driven surrogate modeling of aircraft systems, sensors, electronic warfare capabilities, and weapon systems. This collaboration represents a strategic effort to bridge the critical gap between simulation environments and real-world performance, ensuring AI models can accurately predict and enhance combat effectiveness. The program stands as a vital component in maintaining U.S. air superiority amid rapidly evolving global threats and increasingly sophisticated adversary capabilities.
As the prime contractor for this critical DARPA initiative, Lockheed Martin’s Skunk Works® division brings its unparalleled expertise in autonomous systems, electronic warfare, and artificial intelligence to the forefront. The company’s work program encompasses several groundbreaking efforts that will shape the future of aerial combat.
The development of high-speed, high-fidelity surrogate models represents a core technical challenge. These digital replicas of aircraft and sensor systems must overcome traditional computational limitations to provide realistic performance predictions. Lockheed’s team will train advanced AI agents through intensive simulated engagements, including both dogfighting scenarios and complex BVR combat situations, to enhance tactical decision-making capabilities. During the 18-month period of performance, Lockheed Martin will apply AI and machine learning (ML) techniques to create surrogate models of aircraft, sensors, electronic warfare systems and weapons within dynamic and operationally representative environments.
A particularly innovative aspect involves the integration of electronic warfare dynamics into the AI training environment. This ensures the developed systems can effectively adapt to and counter sophisticated jamming techniques and other deception tactics employed by modern adversaries. The ultimate objective is to create intelligent AI co-pilots and mission planning systems that can seamlessly assist human operators during high-intensity combat operations, simultaneously reducing cognitive workload while dramatically improving mission success probabilities.
“In complex airborne missions, our customers need access to advanced technologies that connect critical systems quickly across all domains. The DARPA AIR programme will use state-of-the-art scientific ML technology and Lockheed Martin’s ARISE infrastructure to deliver unprecedented amounts of data that service members can use to make faster and more informed decisions,” Gaylia Campbell, vice president of engineering and technology for Lockheed Martin Missiles and Fire Control, was quoted as saying in a company press release. “This will provide significant cost savings opportunities for the Department of Defense and serve as a foundation for future AI defence solutions, ensuring the US and its allies maintain their competitive advantage no matter the circumstances.”
ARISE is a family of integrated toolkits used to build a system-level weapon simulation – or “digital twin” – tool. As the mission simulation standard for programmes at Missiles and Fire Control, ARISE is used to reduce product development time, decrease cost, maximise anomaly detection prior to live flight tests, and further integrate model-based engineering into Lockheed Martin’s processes.
DARPA’s AIR Program and BAE Systems’ Role
The DARPA Artificial Intelligence Reinforcements (AIR) program marks a significant leap in the evolution of aerial warfare by targeting one of its most demanding domains—beyond-visual-range (BVR) combat. These engagements unfold at speeds and levels of complexity that outpace human cognitive limits, requiring split-second threat assessments, precise maneuvering, and synchronized control of sensors, weapons, and electronic countermeasures. AIR is designed to produce AI agents capable of autonomously managing this battlespace, dynamically adapting to emerging threats while maintaining tactical dominance in contested environments. As part of Phase 1, DARPA has awarded BAE Systems a $4 million contract to develop and demonstrate these capabilities on F-16 testbeds, advancing the groundwork laid by earlier AI-human dogfight trials such as the X-62A VISTA program—where AI agents showed the ability to challenge human pilots in simulated engagements.
BAE Systems’ FAST Labs division is engineering a machine learning architecture specifically optimized for air combat’s harsh and unpredictable conditions. This architecture is built on three integrated technical pillars. First, advanced physics-based modeling captures the aerodynamic and system-level behaviors experienced during high-G maneuvers, ensuring that AI agents train in simulations grounded in real-world flight dynamics. Second, sophisticated sensor and electronic warfare emulation creates operationally representative threat environments, enabling AI to detect, classify, and counter hostile radar signatures, jamming attempts, and missile launches with precision. Third, a rapid iteration and validation pipeline allows AIR algorithms to evolve continuously, with each software cycle rigorously tested in high-fidelity combat scenarios that mimic real-world engagements.
A critical component of the AIR initiative is the establishment of a rapid iteration pipeline, enabling continuous evolution of the AI software stack. BAE’s approach streamlines the development, testing, and deployment of AI updates using high-fidelity training data. Platforms like Cubic Defense’s SLATE system, which replicates multidomain threat environments, are integral to this process—helping AIR prototypes train in scenarios that mirror real combat conditions. This closed-loop feedback cycle not only improves AI decision-making accuracy but also boosts pilot trust in machine autonomy by embedding transparency and reliability into the development lifecycle. By leveraging platforms like Cubic Defense’s SLATE for multidomain threat simulation, BAE ensures the AI is not just tactically competent but also explainable and trustworthy to human pilots.
Operating from its Arlington, VA, and Burlington, MA facilities, BAE is aligning its AIR program efforts with DARPA’s human-machine teaming vision, in which AI assumes real-time tactical execution roles while human operators retain strategic oversight. This shift not only promises to increase combat effectiveness in BVR scenarios but also lays the foundation for a doctrinal transformation in aerial warfare—where trust in autonomous systems becomes as critical as the systems themselves.
Strategic Implications for National Security
The AIR program carries profound implications that extend far beyond its specific technical achievements, influencing broader U.S. defense capabilities and strategic posture. In an era where adversaries are developing increasingly sophisticated anti-access/area denial (A2/AD) systems, AIR-derived technologies provide critical tools to maintain operational freedom of action. The program’s focus on uncertainty management and adaptability directly counters adversary efforts to create complex, unpredictable battlespaces that traditionally favor defensive strategies.
At the force structure level, successful AIR technologies could enable transformative concepts of operation, particularly in enabling effective manned-unmanned teaming at previously unattainable scales. This capability could dramatically increase combat capacity without requiring proportional increases in personnel or traditional platforms, offering a force multiplier effect of strategic significance. The program also addresses the growing challenge of operator cognitive overload in information-dense combat environments. By automating certain tactical decisions and sensor management tasks, AIR systems allow human operators to focus their attention on higher-level command functions and strategic decision-making.
Maintaining air superiority against technologically advanced adversaries stands as a primary concern. With nations like China and Russia making substantial investments in AI-driven warfare systems, the United States must maintain a decisive edge in autonomous air combat technologies. The AIR program directly addresses this need by ensuring American pilots and autonomous systems can consistently outthink and outmaneuver opponents in highly contested operational environments.
The program also tackles the persistent challenge of simulation-to-reality gaps. By refining AI models to achieve unprecedented alignment with real-world performance, the Department of Defense can conduct more effective training programs while minimizing costly operational surprises during actual missions. This advancement carries particular significance as military systems grow increasingly complex and interdependent.
Perhaps most importantly, the AIR program lays essential groundwork for future autonomous systems integration. Alongside parallel efforts like the Air Force’s Collaborative Combat Aircraft (CCA) initiative, this research paves the way for AI-controlled drone wingmen that can operate alongside manned fighters. Such systems promise to exponentially expand mission capabilities while mitigating risk to human pilots in high-threat environments.
Perhaps most importantly, the AIR program establishes a new model for defense innovation cycles. Its emphasis on rapid iteration, continuous improvement, and direct operational feedback creates a template for faster capability development across the defense enterprise. This approach proves particularly valuable in the AI domain, where commercial sector advancements threaten to outpace traditional defense development timelines without deliberate intervention.
Conclusion: Redefining the Future of Autonomous Air Combat
DARPA’s AIR program represents a watershed moment in military AI development, marking the transition from theoretical capabilities to practical, operational solutions for real-world combat challenges. By directly addressing the most difficult aspects of tactical autonomy – uncertainty management, distributed execution, and human-machine collaboration – the program moves beyond incremental improvements to deliver transformative capabilities to the warfighter.
The program’s dual focus on advanced predictive modeling and distributed autonomous execution, combined with its innovative human-in-the-loop development approach, creates a unique value proposition in defense AI development. As the program progresses through its initial 18-month phase and beyond, it promises to redefine the boundaries of possible in aerial combat while establishing foundational technologies for autonomous systems across all military domains.
In an era of great power competition, initiatives like AIR serve a vital strategic function by ensuring the U.S. maintains its technological edge while adapting to the evolving character of warfare. The technologies emerging from this program will not only enhance immediate combat effectiveness but also establish the framework for next-generation autonomous systems that can operate effectively in the complex, contested environments of future conflicts. As such, AIR represents both a specific solution to current challenges and a prototype for the future of defense innovation in an age of artificial intelligence