The future is autonomous tactical air combat, but will Pilots become Weapon Systems Officers (WSO) trusting their autonomous wingmen of the future?

On August 20, 2020, an artificial intelligence (AI) algorithm, piloting a virtual F-16, beat a seasoned human US Air Force fighter pilot 5-0 in a series of air combat simulations, unveiling the coming of AI-controlled unmanned platforms and gathering widespread media attention. Organized by the Defense Advanced Research Projects Agency (DARPA) under the Air Combat Evolution (ACE) program, the AlphaDogfight Trials (ADT) demonstrated how AI could outperform human pilots reliably in a tournament-style contest between human and machine.

For several decades, technological advancements have enabled air warfare to evolve from dogfighting to beyond-visual-range air-to-air engagements. Due to the advancements in sensor and data fusion technologies along with network-enabled munitions and information systems that are already fielded on 5th generation fighters, the platforms that leveraged the increasingly sophisticated combinations of these technologies will prevail in the future of aerial combat. As aviation technology advanced, complex mechanical systems and analog switchboards in the cockpit were subsequently replaced by digitized cockpits with sensor readouts, autopilot, and automated navigation. Correspondingly, most of the pilot workload on advanced aircraft shifts from how to best fly the plane to using the aircraft's sensors and weapons in conjunction with other offboard assets to support friendly forces. This development crucially and fundamentally changed the importance of pilots from operating the fighter to performing a wider range of functions in more lethal and effective ways. AI is the critical enabler in each of these manifestations of human-machine symbiosis for air warfare. Further advancements in this field will allow the pilot to focus more on broader mission planning while entrusting the business of flying to the AI. The ultimate goal of using AI in air warfare is to provide a decisive advantage in an engagement to achieve victory. 

The AlphaDogfight Trials and the more ubiquitous DARPA ACE program are crucial catalysts to deliver advancements for pilot assist technologies. The USAF seeks to develop revolutionary technologies to field manned fighters teamed with sophisticated unmanned aerial systems, loyal wingmen, or fully autonomous multi-role fighters. Over the last decade, the USAF sought for novel technologies to field current and next-generation piloted fighter aircraft teamed with sophisticated "attritable" unmanned aerial systems or fully autonomous multi-role fighters. Although achieving this vision is not impossible despite the challenges, fully realizing this concept will not happen as quickly as desired. On the road to a significant evolution of air warfare, a more pragmatic conceptualization is required for how AI can be best used in conjunction with unmanned systems.

Regarding the importance of studies on AI in combat systems, the U.S. House Armed Services Committee released a bipartisan congressional report on September 23, 2020, assessing U.S. military capabilities and preparedness to meet current threats. The "Future of Defense Task Force Report" provided some insights about global AI developments and identified China and Russia as the top security threats to the U.S. The report emphasizes that the United States must undertake and win the artificial intelligence race by leading in the invention and deployment of AI using the "Manhattan Project" as a model. Following on the same topic, Bloomberg released an article on February 5, 2020, stating that the global Artificial Intelligence (AI) market size was valued at US$39.9 billion in 2019 and is projected to reach US$390.9 billion by 2027, with a compound annual growth rate (CAGR) of 46.2% from 2019 to 2025. AI is estimated to contribute up to US$15.7 trillion to the global economy in 2030, with US$6.6 trillion likely to come from increased productivity and US$9.1 trillion from consumption-side effects. According to Bloomberg, the unclassified U.S. Government spending on defense AI R&D in 2020 is around $4 billion, while the White House announced an FY2020 non-defense AI R&D budget request of nearly $1 billion. In contrast, the level of Chinese government spending on AI R&D is not clear. Complete annualized figures for Chinese government spending are not publicly available. Instead, only announcements of planned, multi-year spending offer a window into the scale of overall government R&D spending at the national, provincial, and local levels. 

In light of these events, the recent DARPA AlphaDogfight Trials (ADT) were an important display of both technology and AI development-related advancements for future combat environments. As part of the broader DARPA Air Combat Evolution (ACE) technology and experimentation effort, the ADT has pushed for state-of-the-art agent-based modeling and artificial intelligence (AI) applications to air warfare in just over a year. To develop new concepts for employing current and emerging Unmanned Aerial Systems (UAS) capabilities alongside manned aircraft, the USAF and DARPA initiated vanguard programs to provide attritable UAS solutions with the aim of increasing combat capacity and reduce risk during future air operations in permissive and contested environments. Attritable unmanned systems have significantly lower unit procurement costs relative to manned aircraft if manufactured in significant quantities. These are the Low-Cost Attritable Aircraft Technology (LCAAT) program, Low-Cost Attritable Strike Demonstrator (LCASD), Skyborg, Low-Cost Attritable Aircraft Platform Sharing (LCAAPS), Golden Horde, Gremlins, Air Combat Evolution (ACE), and Collaborative Operations in Denied Environments (CODE).

AlphaDogfight Trials (ADT) and the ACE Program

The DARPA AlphaDogfight was a program that pitted several Artificial Intelligence (AI) algorithms using virtual F-16 flight simulators against one another. The AIs were managed by eight teams and competed in a single-round elimination. The participating artificial intelligence (AI) algorithms control simulated F-16 fighters in aerial combat, culminating in a matchup on August 20, 2020, between the top AI and an experienced Air Force fighter pilot flying a virtual reality F-16 simulator. Eight participating companies spent less than a year developing and teaching their AI agents on how to fly and excel in simulated aerial combat. The teams were Aurora Flight Sciences, EpiSys Science, Georgia Tech Research Institute, Heron Systems, Lockheed Martin, Perspecta Labs, PhysicsAI, and SoarTech. 

The first trial, which was held in November 2019, featured algorithms in early development. The agency held the second trial in January 2020 and pitted the further improved algorithms against other AI adversaries developed by the Johns Hopkins University Applied Physics Laboratory (APL). The third and final trial started on August 18, with eight teams flying against the APL's AI adversaries on the first day and then against each other in a round-robin tournament on the second. It was the first time the participants pitted their AIs against one another in a public event. On the last day, the top four teams competed in a single-elimination tournament for the championship title. The last team standing faced an Air Force fighter pilot controlling a virtual plane to test the AI's abilities against a human. Due to the COVID pandemic, the third and final trial could not be held in person at AFWERX (USAF technology accelerator program) in Las Vegas as originally planned. Instead, the Johns Hopkins Applied Physics Laboratory (APL) in Laurel, Maryland, virtually hosted the finals August 18-20 and streamed the event live on YouTube. 

The small (about 30 employees) Maryland company Heron Systems, Inc. took first place in the AlphaDogfight Trials Final event. The company wrote a deep learning software tool that beat the human pilot. Heron's F-16 AI agent, dubbed "Falco," defeated seven other companies' F-16 AI agents and dominated the main event – a series of simulated basic fighter maneuvers (BFM) also known as "dogfights" against an experienced Air Force F-16 pilot – winning 5-0 through aggressive and precise maneuvers the human pilot couldn't outmatch. AI agents developed by Lockheed Martin, Aurora Flight Sciences, and PhysicsAI finished in the top four teams.

Heron Systems' "Falco" was quite aggressive through the trials, with its AI pilot consistently outturning its opponents and scored hits on the virtual F-16 piloted by the unnamed Air Force pilot, call sign "Banger." He is a graduate of the USAF's highly selective Weapons School at Nellis AFB, with more than 2,000 hours in the F-16. Banger and Falco fought in five different scenarios with the simulated fight only using the F-16's guns. The algorithm operated within the limits of the F-16 (it did not pull Gs beyond what the real-world aircraft could do), and each round, the AI managed to outmaneuver and take-out Banger. However, the jet was not limited by the training and thinking of an Air Force pilot, as Banger stated. Having lost four rounds to the AI, he said: "The kind of standard things we do as fighter pilots are not working. For this last one, I'll try to change it up a little bit and do something different." DARPA's Strategic Technology Office (STO) program manager Air Force Col. Dan Javorsek (call sign "Animal") outlined that an F-16 pilot performs basic fighter maneuvers (BFM) while establishing some limits such as not passing within 500 feet (152 meters) or limiting the angle-of-attack (AOA) when firing the gun. The AI did not need to follow these instructions, which helped it gain an advantage. Banger started the match following the basic rules, and in the following rounds, he tried to learn the algorithm's methods, flying more aggressively. The AI can also make adjustments on a "nanosecond level," giving the algorithm "superhuman aiming ability." Compared to AI, the average reaction time for humans is 0.25 seconds for a visual stimulus, 0.17 for an audio stimulus, and 0.15 seconds for a touch stimulus, and the human "OODA loop" (observe, orient, decide, and act) takes longer.

DARPA started the AlphaDogfight Trials (ADT) to expand its AI developer base under the Air Combat Evolution (ACE) program. ACE seeks to automate air-to-air combat and increase trust in combat autonomy and artificial intelligence by using collaborative dogfighting as an entry point into improved human-machine teaming. The technologies developed within the ACE program will ultimately enable future pilots to confidently offload some high workload tasks like air-to-air engagements so they can better focus on managing the larger battlespace. ACE aims to apply existing artificial intelligence technologies to the dogfight problem and, in parallel, implement methods to measure, calibrate, increase, and predict human trust in combat autonomy performance. Ultimately, the program intends to scale the tactical application of autonomous dogfighting to more complex operational-level simulated scenarios based on live data to lay the groundwork for future campaign-level Mosaic Warfare experimentation.

In contested airspace, a single human pilot can increase lethality by effectively orchestrating multiple autonomous unmanned platforms from a piloted aircraft, shifting the human role drastically from a platform operator to a mission commander. In particular, ACE aims to deliver a capability that enables pilots to execute broader global air command missions while their aircraft and teamed unmanned systems are engaged in individual tactics. According to DARPA's statement, ACE creates a hierarchical framework for autonomy in which higher-level cognitive functions (e.g., developing an overall engagement strategy, selecting and prioritizing targets, determining best weapon or effect, etc.) may be performed by a human, while lower-level functions (i.e., details of aircraft maneuver and engagement tactics) is left to the autonomous system. To achieve this framework, pilots must be able to trust the autonomy of their "loyal wingmen" to conduct complex combat behaviors in scenarios such as within-visual-range (WVR) dogfights before progressing to beyond visual range (BVR) engagements. In a statement released by DARPA on August 26, 2020, Tim Grayson, director of DARPA's Strategic Technology Office (STO), said: "The AlphaDogfight Trials outcome shows great promise for future airborne combat systems and concepts involving human-machine symbiosis. As part of STO's Mosaic Warfare vision of distributed manned and unmanned systems, the trials laid a strong foundation for further algorithm development in the ACE program as it moves now from a simulation environment to testing algorithms and measuring pilot trust on actual aircraft." DARPA intends to take the simulator and the simulations used in the AlphaDogfight trials to Nellis Air Force Base to allow other Air Force pilots to compete against AI agents. DARPA's next step will be testing AI capabilities to perform different types of aerial combat missions.

In the future, ACE intends to bridge the gap between the currently used simple physics-based automated systems and the complex systems capable of effective autonomy within highly dynamic and uncertain environments. In this context, DARPA recently awarded contracts to five companies to develop algorithms enabling mixed teams of manned and unmanned combat aircraft to conduct aerial dogfighting autonomously. Boeing, EpiSci, Georgia Tech Research Institute, Heron Systems, and physicsAI were chosen to develop air combat maneuvering algorithms for individual and team tactical behaviors under Technical Area (TA) 1 of DARPA's Air Combat Evolution (ACE) program. Each team is tasked with developing artificial intelligence agents that expand one-on-one engagements to two-on-one and two-on-two within-visual-range (WVR) aerial battles. The algorithms will be tested in three phases: modeling and simulation, sub-scale unmanned aircraft, and full-scale combat representative aircraft scheduled in 2023.

Regarding the contracts, DARPA's Strategic Technology Office program manager Air Force Col. Dan "Animal" JAVORSEK said: "The TA1 performers include a large defense contractor, a university research institute, and boutique AI firms, who will build upon the first-gen autonomous dogfighting algorithms demonstrated in ADT. We will be evaluating how well each performer can advance their algorithms to handle individual and team tactical aircraft behaviors, in addition to how well they can scale the capability from a local within-visual-range environment to the broader, more complex battlespace." After the selection of the TA1 algorithm developers and performers, the four technical areas of the program are all now on contract. 

According to DARPA's statement on November 11, performers for TAs 2-4 were selected earlier this year. TA2 performer SoarTech is developing an experimental methodology for modeling and measuring pilot trust in BFM autonomy as well as novel human-machine interfaces (HMI). TA3 performers Dynetics and Lockheed Martin are developing a data set and model for large force exercise data analytics to evaluate how well TA1 algorithms scale to larger, more complex multi-aircraft scenarios. TA4 performer Calspan Flight Research will supply full-scale L-39 aircraft for Phase 3. Under this contract, Calspan will modify up to four Aero Vodochody L-39 Albatros jet trainers with Calspan's own autonomous fly-by-wire flight control system technology to demonstrate the capabilities of TA1 autonomous dogfighting algorithms and TA2 advanced Human Machine Interfaces (HMI) and AI algorithms on full-scale combat aircraft with an onboard human pilot. Noting that much of the excitement surrounding ACE is based upon the success of ADT and the promise of autonomous tactical air combat, Tim Grayson, director of DARPA's Strategic Technology Office, emphasizes that the program's ultimate goal is to develop a protocol for teaching humans to trust autonomy and underlines that the new contracts represent the first step toward developing the AI side of that partnership.

Skyborg and DARPA’s Vanguard Programs 

The Skyborg project is a United States Air Force Vanguard program initiated by the Air Force Research Lab (AFRL) to develop unmanned combat aerial vehicles to accompany a manned fighter aircraft. Skyborg aims to produce and field low-cost Unmanned Combat Aerial Vehicles (UCAVs) that can be networked for Manned-Unmanned Teaming (MUM-T) to support piloted combat aircraft while learning directly from the behavior of their human peers by using artificial intelligence (AI). The MUM-T will allow the drone to use its sensors and payloads to protect the piloted fighter while providing reconnaissance data to enhance the fighter's capabilities. The shared network enables manned-unmanned teaming (MUM-T), allowing UCAVs and piloted aircraft to work together and complete missions more effectively. The system extends the fleet's reach while keeping the piloted aircraft and personnel out of harm's way. It will allow the UAVs to serve as the eyes and ears for pilots, collecting and sending data from the battlespace to a manned fighter. The autonomy focused Skyborg initiative will enable the Air Force to operate and sustain low-cost, teamed aircraft that can thwart adversaries with quick and decisive actions in contested environments. 

As a follow-up to the Air Combat Evolution (ACE) program, Skyborg will integrate Artificial Intelligence (AI) with autonomous Unmanned Air Vehicles (UAVs) to enable manned-unmanned teaming. It will incorporate the U.S. Air Force AI Accelerator developed in partnership with the Massachusetts Institute of Technology, which will provide fully autonomous flight controls. If successful, Skyborg will finalize the AI-enabled Low-Cost Attritable Aircraft Technology (LCAAT) under the Next Generation Air Dominance (NGAD) program. According to the unclassified U.S. Department of Defense Fiscal Year (FY) 2021 Budget Documentation on Air Force Research, Development, Test & Evaluation (vol 1, pg. 260, February 2020), the Skyborg program's official description is: "Develop and demonstrate an autonomous, attributable vehicle architecture suite which will enable the Air Force to posture, generate, and sustain multi-mission sorties at sufficient tempo to thwart adversary attempts at quick, decisive action in contested and highly contested environments. Develops and demonstrates autonomy architecture and software, enabling machine-machine and manned-unmanned teaming, while ensuring openness, modularity, and expandability. Develops, demonstrates, and integrates hardware components and Open Architecture standards needed to allow modular sensor, communication, and other payload integration into the autonomy and vehicle architectures in systems integration laboratories, platforms, and vehicles. Demonstrates low-cost attributable vehicle concepts and technologies for expeditionary mass generation, including sortie generation employment concepts. Conducts analysis and tests on concepts of operations and concepts of employment for attributable, autonomous, unmanned systems." 

To fast-track capability development capability, the U.S. Air Force Office of Strategic Development Planning and Experimentation (SDPE) at the Air Force Research Laboratory designated Skyborg as one of three Vanguard programs (Navigation Technology Satellite 3/NTS-3, Skyborg, and Golden Horde) in October 2018. These priority initiatives integrate several technology components across multiple domains to create complex, multidisciplinary solutions. 

On July 23, 2020, the Air Force Life Cycle Management Center awarded indefinite-delivery, indefinite-quantity contracts to Boeing Co., General Atomics Aeronautical Systems Inc., Kratos Unmanned Aerial Systems, Inc., and Northrop Grumman Systems Corp., authorizing the companies to compete for up to $400 million for the Skyborg Vanguard Program over the next five years. No money was allocated to the companies, as they will compete against each other for future orders. Later, on September 28, 2020, the U.S. Air Force added nine vendors to the companies that will compete. Under the second phase, the USAF awarded each firm an indefinite-delivery, indefinite-quantity contract worth up to $400 million. The nine companies were AeroVironment Inc., Autodyne LLC, BAE System Controls Inc., Blue Force Technologies Inc., Fregata Systems Inc., Lockheed Martin Aeronautics Company, NextGen Aeronautics Inc., Sierra Technical Services, and Wichita State University. Similar to phase one awards, no money was allotted to vendors as the 13 companies will compete against each other for future delivery orders. Under the contract, Boeing, General Atomics, Northrop Grumman, and Kratos Unmanned Aerial Systems will carry out the Skyborg Prototyping, Experimentation, and Autonomy Development, overseen by the Air Force Life Cycle Management Center and Air Force Research Laboratory. 

The first Low-Cost Attritable Aircraft Technology (LCAAT) demonstrator, the XQ-58A Val”kyrie, is an experimental stealthy unmanned combat aerial vehicle (UCAV) designed and built by Kratos Defense & Security Solutions for the Low-Cost Attritable Strike Demonstrator (LCASD) program. The role of the LCAAT was defined as to escort the 5th generation F-22 or F-35 aircraft during combat missions via the gatewayONE radio link, which is developed by Northrop Grumman as part of the next Advanced Battle Management System (ABMS). The XQ-58A was initially designated XQ-222. The Valkyrie successfully completed its first flight on March 5, 2019, at Yuma Proving Ground, Arizona. The XQ-58 was designed to act as a "loyal wingman" that is controlled by a parent aircraft to accomplish tasks such as scouting or absorb enemy fire if attacked. 

Valkyrie will also be deployed as part of a swarm of drones, with or without direct pilot control. It features stealth technology with a trapezoidal fuselage with a chined edge, V-tail, and an S-shaped air intake. The XQ-58A Valkyrie is the first instantiation of a class of attritable aircraft, which opens the door to new manned-unmanned concepts being explored in the Skyborg program. Through the program, the USAF will assess the military utility of various LCAAT through an experimentation campaign focused on reliability, maintainability, sustainment, and life cycle cost.

Another candidate for the program is the Boeing Airpower Teaming System (ATS), also known as the Boeing Loyal Wingman project, which is a stealth unmanned aerial vehicle in development by Boeing Australia to perform autonomous missions using artificial intelligence. Being developed for the global defense market as well as primary customer, the Royal Australian Air Force, the Loyal Wingman is the first of three class prototypes to be built under Australia's Advanced Development Program, which aims to produce jet-powered, autonomous, artificial-intelligence-powered teaming aircraft that can fly alongside UAVs and piloted combat aircraft. The Loyal Wingman has an interchangeable nose cone, which can be quickly interchanged with other modules for a new mission. The UAV will be designed to act as a "loyal wingman" that is controlled by a parent aircraft to accomplish tasks such as scouting or absorbing enemy fire if attacked. The drone will be the first combat aircraft designed and developed in Australia in over half a century. The Royal Australian Air Force plans to initially buy three Airpower Teaming System (ATS) systems as part of the Loyal Wingman Advanced Development Program (LWADP). Skyborg-equipped Loyal Wingman aircraft could fly alongside fifth-generation fighter platforms like the F-35, participating in combat operations and even making decisions about critical situations where it may be better equipped than human operators.

If the ACE, Skyborg, and DARPA's other Vanguard programs successfully demonstrate that certain old-school air combat capabilities represent a failing strategy, it will prove that the novel and more flexible unmanned systems using modular architecture and additive manufacturing techniques are truly the way forward. Considering the experiments and development programs that aim for manned-unmanned teaming and the emerging role for unmanned systems as loyal wingmen to support piloted platforms, human pilots can become cockpit-based commanders, orchestrating the larger battlespace easily and affordably to present greater complications to the enemy while trusting AI to pilot their own plane or networking unmanned platforms together as expendable assets. AI can become the virtual crew that can navigate and perform complex flying functions, while the human pilot retains more of a Weapon Systems Officer (WSO) focused role. Furthermore, using AI's strength to assess the aircraft's capabilities and field performance will allow pilots to have a higher probability of kill in a wider range of conditions, including extreme situations requiring split second reactions. Ultimately, automated systems with advanced AI algorithms will perform lower-level tasks allowing humans to command and control multiple situations/operations/tasks effectively and efficiently in the complex and dynamic/chaotic battlespace of the future