The discovery of gunpowder in the ninth century and the invention of the atomic bomb in the twentieth century may be considered the first two revolutions in warfare. The third revolution in warfare is Artificial Intelligence (AI), the branch of computer sciences that is engaged in the development of intelligence machines i.e. those that could think and function like human beings. AI has gained enough prominence in military spheres by way of autonomous weaponry on land, sea, air, space and cyber domains to be considered as a breakthrough that militaries around the world are scampering to exploit so as to dominate, or at least gain an advantage over, potential or existing adversaries.
Air power, from the days of Douhet, concerns air supremacy; that is to say, it aims at possessing the capability to use the medium of air to own advantage while denying its use to the adversary. However, concepts of air power thought have evolved remarkably since Douhet on account of technological innovations. From gladiatorial dogfights between knights of the air, the instruments of air power have progressed astoundingly with the advent of Beyond Visual Range (BVR) missiles, air-to-surface weapons launched from long distances without visually sighting the targets they are aimed at, stealth and speed enhancements and aircraft performance in terms of manoeuver ability and agility. AI is turning out to be an extraordinary additive to military aviation which is witnessing a steady and wondrous proliferation of AI.
SINGULARITY AND MILITARY AVIATION
In general terms, AI refers to the ability of machines to perform tasks that normally require human intelligence. From their very inception, computers have been executing tasks that the human brain can and have attained much higher speeds than the human brain. However, those essentially computational tasks could be generally clubbed under Common Artificial Intelligence (CAI) while another term, General Artificial Intelligence (GAI) pertains to cognitive functions i.e. thinking faculty of the brain. A nebulous AI concept is “singularity” which envisages computer programmes becoming so advanced that AI transcends human intelligence, potentially erasing the boundary between humanity and computers. The currently predicted time for that to happen is around 2045. Meanwhile, current research endeavours to elevate CAI to GAI by understanding how the human brain functions and then having AI machines mimicking the processes.
First generation aircraft were the seat-of-the-pant type but in subsequent generations, some human functions have been getting delegated to onboard avionics. A technological breakthrough was Fly-by-Wire (FBW), the term used for flight control systems that use computers to process the flight control inputs made by the pilot or autopilot and send corresponding electrical signals to the flight control surface actuators. This arrangement uses computers to ensure safe control of the aircraft and is widely used now. However, it still functions within an envelope defined beforehand and every action by FBW is a response to ‘If…Then….’ pre-set parameters. In other words, no cognitive intelligence is displayed by the machine. GAI aims at reacting to situations as they develop with a capability to learn from each situation or interaction; some examples are speech recognition, mastery of games such as chess and Go, autonomous systems, flight and combat simulations.
A typical military aviation challenge lies in moving up from say, carrying out guided missile attacks on designated targets to making decisions on the target to be attacked. While AI, replete with super-fast computers, can almost instantly access and analyse huge amounts of data to provide inputs for command and control decisions, subjective problem solving is still not possible by AI. GAI is the objective that will move beyond rules-engined software to software that learns and adapts itself, machine learning (statistical techniques that use algorithms to learn from data without explicit programming) and deep learning (machine learning that develops multiple hidden layers of analysis or deep artificial neural networks to make predictions). Deep learning makes for faster decision making and impels military aviation closer towards GAI, although we are at least two decades away from singularity. Nonetheless, the current and emerging contributions of AI to military aviation are fascinating and are changing military aviation significantly.
There is no universal agreement on airtight definitions of fighter generations; a coarse grouping starts with the appearance of jet fighters at the end of World War II (mid 1940s) which are referred to as the first generation while improvements in design and speeds led to the second generation being pronounced. These were the fighters operating during the Korean War, many with swept wing designs. Supersonic speeds and advanced engines characterised the third generation through the 1950s and 1960s. The fourth generation came in during the 1970s, and was differentiated by significant improvement in avionics and automation like FBW and Full Authority Digital Engine Controls (FADEC). Leading current day fighters are essentially fourth generation although they are being constantly retrofitted with features to enhance their original capabilities. The defining characteristic of the fifth generation is significant amount of stealth, some other features being Active Electronically Scanned Array (AESA) radars, super cruise, plug-and-play electronics, and automation permitting fair amount of autonomy in some areas of operations.
The sixth generation’s distinctive feature is the use of advanced digital technologies of which AI is the most prominent. While some fourth generation fighters carried a Weapons System Operator (WSO) to help the pilot, fifth generation fighters are all single-seat platforms with AI sharing the pilot workload. In the sixth generation, AI is expected to move from sharing the cockpit with a human pilot to manning it independently. An “optionally manned cockpit” design is thus slated to be a key attribute of the sixth generation fighter.
The leader in sixth generation programmes is the United States (US) with two sixth generation programmes namely F-X, also known as Next Generation Air Dominance (NGAD) or Penetrating Counter Air (PCA) for the US Air Force (USAF) and F/A-XX, which the US Navy also calls NGAD, related to a replacement for US Navy’s F/A-18E/F. The projected date of induction for both NGAD is 2030.
Europe, too, has two sixth generation programmes afoot. The first is the Future Combat Air System (FCAS) being developed by a consortium comprising France, Germany and Spain. It aims at producing a Next Generation Fighter (NGF) which will connect through cloud to a variety of UAVs for offensive and surveillance roles. This programme is slated to undertake its first test flight in 2027, and be operational by 2045. The second is the Tempest being developed by the United Kingdom (UK), Sweden and Italy, none of which have a fifth generation programme. Japan has its own F-X sixth generation programme, but has been partnering on some parts of the Tempest programme since 2020, In December last year, it was announced that the UK and Japan are jointly going to produce an engine for Tempest and F-X. Besides the stealth and data fusion features of fifth generation, the Tempest is expected to have significant AI to assist the pilot and for management of teamed unmanned platforms. According to the UK Ministry of Defence (MoD), it is expected to enter service in mid-2030s.
Russia is working on Perspektivnyj Aviatsionnyj Kompleks Dal’nego Perekhvata or PAK DP (which translates to Prospective Air Complex of Long-Range Interception) MiG-41, a long range sixth generation interceptor optimised for air-to-air combat, with a projected first flight in mid-2020s and entry into service after 2030. Very few details have been officially revealed about the MiG-41, but it is expected to have substantial AI, be hypersonic and capable of operating in near space in the satellite hunter role. Informed conjecture has it that an unmanned variant may also emerge from the programme. However, the Ukraine war has put a huge question mark on its future.
In December 2021, a model of this Chinese fighter was exhibited in Shanghai at the First Science and Technology Conference of the airborne Cockpit System Division. A cockpit simulator was also on display. Earlier, in June 2018, a Chinese model called ‘Dark Sword’ was unveiled as the future sixth generation fighter.
US Defence Advanced Research Projects Agency (DARPA) proclaimed in 2019 that, “No AI currently exists, however, that can outduel a human strapped into a fighter jet in a high-speed, high-G dogfight.” A year later, under DARPA’s famous Alpha Dogfight trials, a simulated F-16 fighter with AI in the cockpit defeated a top-gun from US Air Force (USAF) in five sessions of mock air combat. The demonstration highlighted the possibility of AI replacing a pilot in the cockpit for something as fast moving as a dogfight.
The USAF has already flown a single seat U-2 on a simulated missile strike mission with an AI algorithm called ARTUμ as a working crew member. A human pilot flew the aircraft and coordinated with ARTUμ which was responsible for sensor employment and tactical navigation. The system is transferable to another type of aircraft with ease and could well be a cockpit occupant in modern fighters as well as future ones. DARPA’s Aircrew Labour In-Cockpit Automation System (ALIAS) is aimed at developing a modular kit that would permit appending high levels of AI driven autonomy into an existing aircraft. The ultimate aim is to execute an entire mission from take-off to landing without a pilot. There are reports of Japan developing an unmanned fighter to be operational by 2035, and AI in the cockpit appears to have caught the attention of military aviation. As mentioned earlier, all the sixth-generation designs under development envisage optionally manned cockpits.
AI ON UAVS
Unmanned Aerial Vehicles (UAVs) have been used in combat during military conflicts for long; notably, a US Predator first fired a missile in 2001. Azerbaijan and Armenia saw decisive use of UAVs in the latter half of 2020, and the ongoing Ukraine war has seen widespread and efficacious use of UAVs. Autonomous UAVs can now execute accurate strikes against ground targets carrying out target selection and firing premised on AI which is either onboard or on a platform they are in data communication with. The MQ-9 Reaper has been used to test Agile Condor, an AI pod designed to detect, categorise and track potential objects of interest. This AI pod has the potential to identify targets and determine priorities for engagement. This is applicable to loitering munitions as well, AI enabling them to carry out their suicidal missions without an operator in the loop, but permitting them enough autonomy to recover safely if no worthwhile target is unearthed.
AI has been surreptitiously making its way into UAV operations and each episode of success for UAVs strengthens the motivation to embed AI in diverse forms to enhance their capabilities. The US Navy X-47B has demonstrated not only deck launches, landings and go-around, but even an aerial refuelling, all carried out autonomously. The QF-16, an unmanned platform derived by converting old F-16s, has the capability to fly autonomously on predetermined routes for decoy roles or used as aerial targets for testing air-to-air missiles and guns. Looking ahead, it is easy to imagine UAVs that will be enabled by AI to execute fully autonomous operations.
Drone swarms have caught military attention in the recent past. Their small size and small Radar Cross Section (RCS), low cost that permits proliferate numbers to be used together and AI to connect them together to seemingly think like one entity makes them objects of military interest. Dramatic swarm attacks like the one on Saudi Aramco oil facilities at Abqaiq and Khurais in September 2019, have helped keep that interest alive. DARPA’s Offensive Swarm-Enabled Tactics (OFFSET) programme has contracted with nine companies to develop AI technology that will enable 250 small air and/or ground units to collaborate. Aimed primarily at facilitating operations in dense urban environments, the programme showcases the promise of swarming UAVs.
Edge computing, already being talked about as the next major revolution in information technology, is the emerging solution to the problem of real time autonomous control of high speed, highly manoeuvrable swarm UAVs collaborating with each other tactically in close proximity to each other and to hostile aerial platforms, as cloud computing is too slow for their cooperative communication. Edge computing eliminates the time taken for devices/UAVs to communicate through cloud computing which connects through centralised data centres or servers and processes data on the spot at the “edge” of the network i.e. at or near the source of the significant data. This munificence of the information technology is quintessential to the progression of swarm UAVs but it is predicated to the employment of AI.
A spin-off of optionally manned cockpits is the AI-driven capability of manned fighters to team with UAVs and Unmanned Combat Aerial Vehicles (UCAVs) for the performance of a variety of offensive and defensive roles. There are already programmes which have demonstrated and are constantly refining this Manned-Unmanned Team (MUM-T) concept. AI and machine learning are enabling the concept of one manned aircraft leading a team of unmanned aircraft capable of autonomous operations.
US Air Force Research Laboratory (AFRL) has been working on a concept of an unmanned aircraft serving as an adjunct flying platform to a manned command aircraft, scouting for threats and taking them on, if required, hence the term “loyal wingman”. Live demonstrations with one manned and one unmanned F-16 have been impressive and promising enough for all sixth generation aspirations to envision MUM-T as part of the design requirements. A demonstration has already been executed with three F/A-18 Super Hornets releasing a swarm of 103 autonomous micro drones which then proceeded to exhibit “collective decision-making, adaptive formation flying, and self-healing”. The swarming drone concept is especially alluring as it represents myriad possibilities. Some UAVs could have offensive roles, some could be sheer decoys, some could carry out Intelligence, Surveillance and Reconnaissance (ISR) tasks, some could carry EW payloads, some others could aid in target recognition and tracking and yet others could be loitering munitions with possibly air-to-air capability so that they can sacrifice themselves in defence of their command aircraft. The possibilities are indeed boundless.
The crucial AI technologies that are lodged in these UAVs enable them to be part of MUM-T programmes like the US Skyborg, one of three Vanguard programmes launched by AFRL. Skyborg was launched in 2018, by AFRL’s Strategic Development Planning and Experimentation (SDPE) and AFRL calls it “an autonomy-focused capability that will enable the Air Force to operate and sustain low-cost, teamed aircraft that can thwart adversaries with quick, decisive actions in contested environments.” Accompanying UAV programmes include the Kratos-built XQ-58 Valkyrie which completed a successful flight in March 2019 under AFRL’s Low Cost Attritable Aircraft Technology (LCAAT) project. Skyborg related contracts have also been awarded to General Atomics, Northrop Grumman and Boeing. Skyborg’s eventual loyal wingman could well be an unmanned, AI-piloted fighter.
In May 2019, DARPA launched Air Combat Evolution (ACE) Programme to adopt AI in individual and team aerial combat tactics and to develop Air Combat Manoeuvering (ACM) algorithms for visual 1-versus-1, 2-versus-1 and 2-versus-2 engagements with a broad spectrum of performance. The Alpha Dog Fight trials mentioned earlier were a part of ACE which embraces manned and unmanned aerial combat. While ACE is aimed at developing AI software capable of close combat by unmanned platforms autonomously, Skyborg has modern capabilities such as Beyond Visual Range (BVR) missiles and long range sensors in mind for loyal wingmen. However, as both programmes look at unmanned wingmen, they may be merged at some future point in time as the programmes mature. That will remove the dividing line between close aerial combat and BVR engagements by AI-enabled UAVs toiling as wingmen.
Boeing, which has a stake in US Skyborg programme, is also designing and developing in collaboration with Royal Australian Air Force (RAAF) the Airpower Teaming System (ATS), its first unmanned system in Australia. According to Boeing, the loyal wingman UAV developed by it for RAAF has similar target detection, prioritisation and engagement capabilities. Flight tests have already been carried out.
The European Next Generation Weapon System (NGWS) aims at teaming sixth generation manned fighters (possibly NGFs) with UAVs (called Remote Carriers (RCs) under the programme) which will be appendable in a scalable and flexible manner.
Russia has also been working on developing new UAVs to function as autonomous loyal wingmen. Reportedly ‘Grom’, meaning thunder, developed by Kronstadt Group will be able to control a swarm of ten smaller drones called ‘Molniya’ or lightning. Another UAV, the S-70 Okhotnik-B (Hunter) designed by Sukhoi Design Bureau and Russian Aircraft Corporation MiG first flew in 2019, and is also designed to be a loyal wingman capable of carrying two tons of internal payload including missiles and bombs or be installed with electro-optic targeting, communication and reconnaissance equipment.
AI IN TRAINING AND SIMULATION
AI has been used by USAF for accelerating pilot training through its Pilot Training Next (PTN) programme since 2018, with noteworthy success in speeding up the training and in the assessed quality of programme graduates. When Virtual Reality (VR) was added on to AI, the training became even better and meaningful. Another development enabled by AI in pilot training is Live, Virtual and Constructive (LVC) training. Live refers to a pilot in an actual cockpit, Virtual relates to a pilot in an aircraft simulator and Constructive is a purely software simulated training asset. This capability is critical to training for, say the fifth generation F-35, which does not have a dual seat version at all, all three versions being single-seat.
The Chinese have also been planning to achieve similar arrangements using L-15 jet trainer aircraft. Reportedly, the aircraft would be fitted with an AI-driven virtual tactical training system which could be linked to other aircraft or a ground simulator with the two being interconnected. Akin to AlphaDogfight and indeed based on it, there are projects afoot to provide pilots in real aircraft, but with Augmented Reality (AR) headsets to pit their combat skills against AI driven virtual fighters mimicking enemy aircraft. Clearly, AI, in tandem with VR and AR, is moving at an impressive pace towards providing speedier and superior training and simulation solutions for military aviation.
AI AND AIRCRAFT MAINTENANCE
Aircraft maintenance is hugely expensive and any reduction in maintenance repairs or replacements can save large amounts of money. Lapses in maintenance or material/component failures before prescribed schedules can have air safety implications, not to mention expensive hull losses and catastrophic pilot casualties. AI is proving to be a useful tool in predictive and preventive maintenance. Preventive maintenance is regular and routine maintenance so as to keep an aircraft flying and prevent any costly unplanned downtime from unexpected equipment failure. Predictive maintenance uses AI and data analytical tools to predict how long an aircraft component will last and can be seen as a sub-set of preventive maintenance. Although maintenance activities are not as visible as the operations, the deployment of AI into the maintenance domain is certainly a largesse that military aviation can benefit from and be grateful for.
AI in the cockpit promises to be AI’s single most significant benediction. The competence to autonomously fly a fighter aircraft in its full spectrum of roles, is not yet within reach of current day technology despite the sensational Alpha Dogfight demonstration, but it is not hard to visualise it happening over the next few years. Besides, cockpit AI and connectivity between manned and unmanned elements of the envisioned teams is going to be a problem with edge computing still not ready to deliver at speeds needed for aerial combat in teams. When that becomes possible, solo missions by AI in military aircraft could well become the norm. Meanwhile, ethical and operational apprehensions about AI’s credentials for occupying a cockpit are being assuaged by projecting the near future as MUM-T with a human in command of a mission and all its elements in contrast to an autonomous decision making AI in any cockpit. Despite a lot of noise about restraining the use of AI for military purposes, its permeation into military aviation is inexorable, irreversible and predictable. It is unlikely that any international consensus or treaty will bridle AI’s rapid inroads into military aviation.