International Defense Security & Technology Your trusted Source for News, Research and Analysis
Related Articles
A dogfight, traditionally defined as an aerial battle between fighter aircraft at close range, remains a critical element of modern air combat despite the evolution of warfare technologies. While contemporary terminology favors Air Combat Maneuvering (ACM), which encompasses the tactical execution of Basic Fighter Maneuvers (BFM) to outmaneuver adversaries, the essence of dogfighting—quick decision-making under extreme pressure—persists. Pilots engaged in dogfights must process vast amounts of information in split seconds, balancing high-speed maneuvers, threat assessment, fuel management, terrain awareness, and mission objectives. Their survival often hinges on a combination of skill, situational awareness, and rapid, creative decision-making.
A vivid reminder of the dogfight’s enduring relevance occurred on February 27, 2019, when Indian and Pakistani Air Forces clashed in a high-stakes aerial battle over the Line of Control. The confrontation, the first between the two nations since the 1971 war, saw Indian Wing Commander Abhinandan Varthaman piloting a Soviet-era MiG-21 against Pakistan’s modern F-16s, JF-17s, and Mirage-5s. Despite the technological disparity, Varthaman successfully engaged and shot down an F-16 with a Vympel R-73 missile before his own aircraft was downed. This brief yet intense encounter underscored that even in an age dominated by beyond-visual-range (BVR) engagements and advanced missile systems, the fundamentals of dogfighting remain a critical component of aerial warfare.
Recognizing the challenges of high-speed, high-G aerial combat, the Defense Advanced Research Projects Agency (DARPA) has been spearheading the integration of Artificial Intelligence (AI) into air-to-air combat through its Air Combat Evolution (ACE) program. While AI has already surpassed human performance in strategic games like chess, Go, and even complex real-time strategy games such as Dota 2 and StarCraft II, dogfighting represents a unique challenge. The dynamic, high-stakes environment of aerial combat requires AI to process real-time data, adapt to unpredictable human behavior, and execute precise maneuvers—all under conditions that push the limits of both human and machine capabilities.
What is DARPA’s ACE Program?
The Air Combat Evolution (ACE) program is part of DARPA’s broader mission to advance AI and autonomy in military applications. Its primary goal is to develop AI-driven systems capable of performing complex air-to-air combat maneuvers, traditionally executed by human pilots, with speed, precision, and adaptability that surpasses human limitations.
Rooted in the foundational principles of the OODA Loop—Observe, Orient, Decide, Act—developed by U.S. Air Force Colonel John Boyd, ACE seeks to automate aerial engagements with machine-speed precision. The OODA Loop emphasizes rapid, iterative decision-making, where the ability to adapt faster than an opponent can determine victory. DARPA’s vision is to create AI systems capable of not just following this loop but executing it with superhuman speed and accuracy, thereby redefining the tempo of modern aerial warfare.
As warfare increasingly shifts toward human-machine teaming, DARPA’s ACE program focuses on automating the split-second decisions required in high-intensity dogfights. By leveraging advances in deep learning, AI can now excel in the “observe” phase, with remarkable capabilities in processing images, video, audio, and sensor data. The “decide” phase has also seen rapid progress through technologies like evolutionary computation and Bayesian optimization, enabling AI to make complex tactical decisions in dynamic environments. However, challenges remain in the “orientation” phase, where AI must interpret context, assess threats, and adapt strategies—a cognitive process traditionally dominated by human intuition. DARPA aims to overcome these hurdles using large-scale graphical models and real-time black-box optimization, allowing AI to learn, adapt, and optimize its decision-making under conditions of uncertainty.
ACE is not just about replacing human pilots; it’s about enhancing human-machine collaboration. The program envisions a future where pilots act as mission commanders, managing multiple autonomous aircraft (also known as loyal wingmen or unmanned combat aerial vehicles—UCAVs) while AI handles the split-second decisions required during dogfights.
The ACE program is not just about enhancing AI capabilities; it’s fundamentally about building trust between human pilots and autonomous systems. Just as new fighter pilots begin their training with basic flight maneuvers before advancing to complex aerial combat, AI must first demonstrate competence in controlled, one-on-one dogfighting scenarios. This iterative learning process mirrors human pilot development, ensuring AI systems can reliably execute basic combat tasks before scaling up to multi-aircraft engagements. According to Lieutenant Colonel Dan Javorsek, ACE program manager, “We envision a future in which AI handles the split-second maneuvering during within-visual-range dogfights, keeping pilots safer and more effective as they orchestrate large numbers of unmanned systems into a web of overwhelming combat effects.”
The Historic Virtual Dogfights: AI vs. Human Pilot
In August 2020, the world witnessed a groundbreaking event that redefined the future of aerial combat. In a virtual dogfight organized by the Defense Advanced Research Projects Agency (DARPA), an artificial intelligence (AI) system defeated a seasoned U.S. Air Force fighter pilot in five consecutive simulated engagements. This historic achievement was part of DARPA’s Air Combat Evolution (ACE) program, a cutting-edge initiative designed to integrate AI into air-to-air combat, setting the stage for a new era where autonomous systems handle close-range dogfights while human pilots oversee missions as strategic commanders.
In the ACE program’s most publicized event, an AI agent developed by Heron Systems, a small defense contractor specializing in machine learning, faced off against a highly experienced F-16 fighter pilot known by the call sign “Banger” in a series of simulated dogfights. The AI’s victory wasn’t just surprising—it was dominant, winning all five rounds without a single loss.
The Rise of AI in Aerial Combat: DARPA’s Breakthrough with the X-62A VISTA
In a landmark achievement for military aviation, the Pentagon’s Defense Advanced Research Projects Agency (DARPA), in collaboration with the U.S. Air Force (USAF), has successfully conducted the first-ever air combat drills pitting an artificial intelligence (AI) agent against a human fighter pilot. This breakthrough took place using a specially modified Lockheed Martin F-16D, known as the X-62A Variable In-Flight Stability Test Aircraft (VISTA), designed to support cutting-edge flight control technologies. The milestone, achieved in 2023 but disclosed in April 2024, marks a pivotal step in automating close-range aerial combat, signaling a transformative shift in the future of air warfare.
DARPA’s Air Combat Evolution (ACE) program spearheaded this initiative, conducting 21 test flights between December 2022 and September 2023. Throughout these trials, AI systems accumulated over 17 hours of flight time, demonstrating remarkable agility and tactical decision-making in within-visual-range (WVR) dogfights—scenarios traditionally dominated by human instinct and experience. The AI-controlled X-62A engaged human pilots in dynamic aerial maneuvers, reaching speeds of over 1,931 km/h and closing distances within 610 meters. Impressively, despite having human safety pilots onboard with override capability, the AI performed flawlessly without requiring intervention during any of the dogfights.
The rapid pace of development—refining over 100,000 lines of flight-critical software—highlights DARPA’s commitment to advancing machine-learning algorithms capable of handling complex aerial combat scenarios. These AI systems are designed to adapt in real-time, optimizing flight-envelope protection, collision avoidance, weapons engagement, and tactical positioning. The success of these tests not only showcases AI’s potential to outperform human pilots in certain combat situations but also builds the foundational trust necessary for future human-machine collaboration in military operations.
Looking ahead, the lessons learned from the X-62A’s dogfighting trials are expected to feed directly into the USAF’s Collaborative Combat Aircraft (CCA) program, aimed at deploying autonomous jets alongside manned platforms. With defense contractors like Boeing, Kratos, and General Atomics leading the development of autonomous aircraft such as the MQ-28 Ghost Bat, the future battlespace will likely feature seamless integration of AI-driven systems. This technological leap is not just about enhancing combat effectiveness—it’s a strategic imperative to counter emerging threats, particularly from near-peer adversaries like China, where numerical superiority in aircraft and missile systems poses significant challenges. DARPA’s achievement with the X-62A VISTA may well be the dawn of a new era in air combat dominance.
Key Takeaways from the Dogfight
The recent dogfight between DARPA’s AI-powered X-62A VISTA and a human fighter pilot marked a pivotal moment in military aviation, revealing the transformative potential of artificial intelligence in aerial combat. One of the standout observations was the AI’s superhuman reaction time. Unlike human pilots, who are limited by biological response speeds, the AI executed precise maneuvers within milliseconds. This lightning-fast responsiveness allowed it to react to threats, adjust trajectories, and seize tactical advantages with a level of precision and speed unattainable by even the most skilled human aviators.
In addition to its rapid reflexes, the AI demonstrated unconventional tactics that defied traditional dogfighting norms. Free from the cognitive biases and risk-averse tendencies ingrained through human training, the AI employed aggressive, unpredictable strategies. It wasn’t hesitant to execute risky head-on attacks—maneuvers that a human pilot would typically avoid due to the high probability of failure or self-endangerment. This boldness added an element of surprise, giving the AI an edge in dynamic combat scenarios.
Another remarkable capability was the AI’s adaptive learning. Rather than relying on static programming, the system continuously refined its tactics through each engagement. With every dogfight, the AI absorbed new data, adjusted its strategies, and improved its performance—adapting at a pace far beyond what human pilots could achieve through traditional training. While the controlled simulation environment differs from real-world combat conditions, the demonstration proved that AI could outperform human pilots in high-stakes, fast-evolving scenarios where split-second decisions are the difference between victory and defeat.
While dogfighting may become rare in future warfare, the skills and trust developed through advanced combat autonomy programs are critical for broader applications. These initiatives lay the groundwork for leveraging autonomous systems beyond individual aircraft, evolving toward the coordination and management of large-scale drone swarms and complex autonomous combat networks. By building confidence in AI’s ability to handle high-stakes, dynamic scenarios, these programs pave the way for more sophisticated human-machine collaboration in diverse military operations.
However, the goal extends beyond outperforming human pilots in isolated scenarios. ACE envisions a future where human pilots act as mission commanders, orchestrating multiple autonomous unmanned systems while AI handles the complexities of individual engagements. This approach aligns with DARPA’s broader “mosaic warfare” strategy, which emphasizes adaptable, resilient force structures composed of both manned and unmanned assets, capable of responding dynamically to evolving threats.
How the ACE AI Works: Beyond Simple Automation
The AI developed under DARPA’s Air Combat Evolution (ACE) program represents a significant leap beyond basic automation. Rather than following a rigid set of pre-programmed instructions, ACE’s AI is built on advanced reinforcement learning algorithms, enabling it to learn through trial and error, much like a human gaining expertise through experience. This learning model allows the AI to adapt dynamically, improving with every encounter and evolving its strategies to meet new challenges.
At the core of ACE’s success is reinforcement learning, where the AI receives virtual “rewards” for successful maneuvers—such as gaining a positional advantage or evading an attack—and “penalties” for errors like losing situational awareness or exposing vulnerabilities. Through thousands of simulated dogfights, this reward-based system enables the AI to identify and optimize the most effective combat strategies. This continuous feedback loop accelerates the learning process far beyond what traditional pilot training can achieve.
Complementing this learning approach is the use of high-fidelity simulation training. These advanced virtual environments replicate the complexities of real-world aerial combat with remarkable detail, allowing the AI to experience countless combat scenarios in a compressed timeframe. This intensive training accelerates the AI’s development, providing exposure to a wide range of tactical situations, environmental conditions, and adversary behaviors.
A standout feature of ACE’s AI is its capability for decision-making under uncertainty. In the chaotic and unpredictable environment of air combat, pilots often operate with incomplete or rapidly changing information. The AI excels in these conditions, processing vast data streams in real time to make swift, informed decisions—a critical advantage in high-pressure engagements.
Importantly, the ACE program isn’t solely focused on the AI’s technical prowess. A key objective is building trust between human pilots and autonomous systems. The development of intuitive human-machine interfaces and transparent AI behaviors ensures that pilots feel confident relying on the AI during critical missions. Establishing this trust is vital for the seamless integration of autonomous systems into future combat operations.
The Future of Air Combat: Human Pilots as Mission Commanders
DARPA’s ACE program is poised to fundamentally reshape the role of fighter pilots in the battlespace of the future. Rather than being directly involved in every dogfight maneuver, future pilots will serve as mission commanders, overseeing autonomous drones that handle close-range engagements. This paradigm shift allows human operators to step back from the intense demands of real-time aerial combat and focus on broader strategic objectives.
In this new operational model, pilots will manage swarms of AI-powered aircraft, coordinating missions, optimizing resource allocation, and adapting strategies in real time based on evolving battlefield conditions. While the AI excels in executing tactical maneuvers with superhuman speed and precision, human operators will provide the critical elements of strategic judgment, ethical oversight, and adaptability that machines cannot replicate.
This evolution represents more than just a technological advancement—it’s a transformation in how air combat will be conducted in the 21st century. As AI takes on an increasingly central role in tactical decision-making, human pilots will become the architects of battle, leveraging the strengths of autonomous systems to achieve mission success in complex, multidomain warfare environments.
ACE’s technological roadmap spans three phases:
- Simulation: Testing AI algorithms in virtual environments to refine tactics and decision-making models.
- Unmanned Flight: Implementing AI in drones to validate performance in controlled, real-world conditions.
- Manned-Unmanned Teaming: Conducting full-scale, live dogfights with human pilots and autonomous systems working in unison.
“Our vision is a future where AI not only augments human capability but becomes an indispensable partner in complex combat scenarios,” Javorsek asserts.
Key Benefits of Human-AI Teaming
The integration of artificial intelligence (AI) into aerial combat through programs like DARPA’s Air Combat Evolution (ACE) offers numerous strategic advantages. One of the primary benefits is cognitive offloading, where AI takes on the intense mental demands of dogfighting, allowing human pilots to focus on broader mission objectives such as strategic coordination and situational awareness. This dynamic shift enables pilots to manage complex scenarios more effectively without being overwhelmed by the fast-paced, split-second decisions required in close-range engagements.
Another significant advantage is force multiplication. AI-driven autonomous systems allow a single pilot to command and coordinate multiple unmanned aircraft simultaneously, exponentially increasing combat capabilities without the need for additional personnel. This not only boosts operational efficiency but also offers greater flexibility in executing large-scale missions. Additionally, risk reduction becomes a key factor, as autonomous aircraft can undertake high-risk maneuvers in contested or hostile environments, minimizing the exposure of human pilots to life-threatening situations.
Perhaps most transformative is the potential for enhanced decision-making. The synergy between human intuition and AI’s data-driven precision creates a formidable partnership. While AI excels at processing vast amounts of data and executing optimal tactical responses, human pilots bring adaptability, creativity, and moral judgment to the equation. This combination can lead to superior outcomes in dynamic combat scenarios, where both rapid calculations and nuanced decision-making are critical.
Recent Developments and Partnerships Under DARPA’s ACE Program
DARPA’s Air Combat Evolution (ACE) program has seen rapid advancements through strategic partnerships with key industry players and research institutions, each contributing to the evolution of AI-driven air combat capabilities. These collaborations are shaping the future of aerial warfare, moving from simulated environments to real-world applications, and expanding the potential of human-machine teaming.
DARPA’s partnerships with industry leaders such as Dynetics, Boeing, EpiSci, Georgia Tech Research Institute, and physicsAI have been instrumental in advancing ACE’s objectives. These collaborations focus on developing AI algorithms that excel not only in individual dogfighting tactics but also in team-based aerial combat, where coordination and adaptability are crucial. The program’s ultimate goal is to create an autonomous combat ecosystem capable of functioning effectively in Mosaic Warfare—a military strategy emphasizing the rapid assembly of diverse capabilities to overwhelm adversaries through agility and unpredictability.
One of the pivotal developments occurred in May 2020, when Dynetics, a subsidiary of Leidos, secured a $12.3 million contract for Phase 1 of the ACE program. Their focus is on Alpha Mosaic (Technical Area 3), a critical component aimed at scaling automated dogfighting technologies to large-scale, multi-aircraft scenarios. During this 18-month phase, Dynetics will advance algorithms for automated dogfighting, preparing them for operational scenarios involving large, heterogeneous aircraft fleets. The focus is on improving within-visual-range, air-to-air engagements and expanding to campaign-level experiments with manned and unmanned vehicles. This phase is foundational for transitioning from individual engagements to complex, coordinated missions involving numerous autonomous and manned aircraft. According to Tim Keeter, Dynetics ACE Program Manager, “Our team brings novel solutions scalable to ACE’s objectives. These efforts will help expand the U.S. military’s advantage in the evolution of Mosaic warfare.” Dynetics’ work not only advances AI combat capabilities but also lays the groundwork for manned-unmanned team experiments, a key element in future air operations.
Progressing from virtual simulations to live flight environments, Calspan Corporation entered the ACE initiative in July 2020 with a $14.1 million contract. Their mission is to modify Aero Vodochody L-39 Albatros jets, integrating autonomous fly-by-wire systems to create advanced experimental testbeds. These modified jets serve as platforms for evaluating cutting-edge AI algorithms and human-machine interfaces under real-world flight conditions. Test operations are conducted from Calspan’s facility in Niagara Falls, NY, with flight activities over Lake Ontario. This represents a significant leap from controlled digital simulations to the dynamic, unpredictable conditions of live aerial combat, providing invaluable data on AI performance and pilot-AI interactions.
By November 2020, DARPA expanded the ACE program’s AI ecosystem by awarding contracts to a diverse group of industry leaders and academic institutions. This cohort includes Boeing, EpiSci, the Georgia Tech Research Institute, Heron Systems, and physicsAI. Each organization brings specialized expertise to the program, focusing on the development of sophisticated algorithms designed to enhance both individual and team-based aerial tactics. Their work goes beyond traditional dogfighting scenarios, exploring how mixed teams of manned and unmanned aircraft can operate seamlessly in complex, contested environments. This approach not only strengthens AI’s tactical versatility but also ensures adaptability across a range of military applications, from air superiority missions to multidomain operations.
In JUly 2024, Defense Advanced Research Projects Agency (DARPA) selected Cubic Defense, a leading provider of technology solutions including advanced air combat training, as a partner in its Air Combat Evolution (ACE) program. In 2023, ACE successfully tested the X-62A, or variable in-flight simulator test aircraft (VISTA), which is an AI-controlled F-16 used to test the AI algorithm capability against human-operated F-16s. The X-62A became the first autonomous fighter jet to engage in visual-range combat scenarios against a manned aircraft.
The X-62A engine used high-fidelity time space position information from Cubic Defense’s secure live virtual and constructive advanced training environment (SLATE) system. According to Paul Averna, vice president and general manager of advanced training solutions for Cubic Defense, SLATE offers the realistic pacing of multidomain and high-threat environments to the live cockpits and operator consoles through computer-generated forces.
Cubic Defense joins Calspan Corporation, EpiSci, Lockheed Martin Skunk Works and Shield AI as an industry partner with ACE. The program is a collaboration between academia, government and private industry. ACE’s government partners include the Air Force Test Center, Air Force Research Laboratory, DARPA and the Air Force Test Pilot School. Academic partners include Johns Hopkins University and the MIT Lincoln Laboratory. All X-62A tests and demonstrations of autonomous combat maneuvers have taken place at the Air Force Test Pilot School at Edwards Air Force Base, California, and will continue throughout 2024.
Challenges and Ethical Considerations
Despite the promising benefits, the integration of AI into lethal combat systems introduces a host of technical, operational, and ethical challenges. One of the foremost issues is trust and reliability. Pilots must have complete confidence in AI’s ability to make life-or-death decisions in the unpredictable chaos of real-world combat. Achieving this level of trust requires rigorous testing, validation, and transparency in AI behavior under diverse conditions.
Moreover, ethical concerns arise regarding accountability in warfare. When an autonomous system is responsible for lethal actions, questions emerge about who bears responsibility for those decisions—designers, operators, commanders, or the AI itself. Establishing clear ethical frameworks and rules of engagement will be essential to address these dilemmas.
Another critical challenge is real-world testing. While AI may perform exceptionally in controlled simulations and test environments, the unpredictability of actual combat—factors like electronic warfare interference, extreme weather, and the erratic behavior of human adversaries—poses unique obstacles. AI systems must demonstrate resilience and adaptability beyond laboratory conditions to prove their effectiveness in active combat zones.
Lastly, adversarial threats represent a significant security risk. AI-driven platforms could become targets for cyberattacks or adversarial manipulation, potentially compromising mission integrity or even turning autonomous systems against their operators. Robust cybersecurity measures and fail-safes are imperative to mitigate these vulnerabilities.
Beyond Dogfights: ACE’s Broader Impact on Military Aviation
The impact of the ACE program extends well beyond the realm of traditional dogfighting. The AI algorithms, human-machine interfaces, and decision-making architectures developed through ACE have broad applications across the military landscape. In Unmanned Aerial Vehicle (UAV) operations, AI can enhance the autonomy of drones, enabling more sophisticated reconnaissance missions, precision strikes, and electronic warfare capabilities without direct human control.
Furthermore, the concept of autonomous wingmen is gaining traction through initiatives like the U.S. Air Force’s Skyborg project. This program aims to develop AI-powered Unmanned Combat Aerial Vehicles (UCAVs) designed to operate alongside manned aircraft, providing tactical support, acting as decoys, or even engaging in offensive operations independently. This human-machine teaming model is expected to redefine air combat strategies, offering greater flexibility and survivability in contested environments.
Beyond aerial operations, the ACE program’s innovations are pivotal for multidomain operations, where AI-driven platforms are integrated across air, land, sea, space, and cyber domains. This interconnected approach enables coordinated military campaigns, leveraging AI’s ability to process real-time data and optimize strategic decisions across diverse operational theaters. As warfare becomes increasingly complex and data-centric, the technologies emerging from ACE will play a crucial role in shaping the future of military dominance.
Conclusion: The Inevitable Evolution of Aerial Combat
DARPA’s ACE program has ushered in a new era where artificial intelligence is no longer just an assistant—it’s a formidable combatant. The virtual dogfight that saw AI defeat a human pilot wasn’t just a milestone; it was a glimpse into the future of warfare, where machines and humans operate as cohesive teams.
As ACE progresses, it will serve as a blueprint for integrating AI across military operations, paving the way for Mosaic Warfare—a future where manned and unmanned systems form a dynamic, cohesive force capable of overwhelming adversaries through superior speed, coordination, and adaptability.
As AI continues to evolve, the question is no longer whether autonomous systems will play a role in future conflicts—but how nations will balance the power of machine intelligence with the irreplaceable judgment of human commanders. The sky is no longer the limit; it’s the battlefield where the future is being written, one algorithm at a time.
References and resources also include:
https://www.intelligent-aerospace.com/military/article/14179147/darpa-air-combat-evolution
