Artificial Intelligence is evolving as an important constituent of air power
Gp Capt A.K. Sachdev (retd)
In the summer of 2016, ALPHA, an Artificial Intelligence (AI) system developed by a doctoral graduate at the University of Cincinnati was pitched against retired US Air Force Colonel Gene Lee in an air combat simulator.
Colonel Lee is an instructor who has trained thousands of US Air Force pilots and is an Air Battle Manager with experience in air combat simulations since the Eighties; ALPHA runs on Raspberry Pi — a low cost, credit-card sized computer that plugs into a computer monitor or TV, uses a standard keyboard and mouse, and costs just USD35. The contest turned out to be unevenly matched; ALPHA consistently attained an advantage over the Colonel in simulated air combat.
Considering the velocity acquired by AI evolution in recent years, it is easy to imagine what ALPHA’s successors would be capable of achieving in aerial combat when liberated from the control of (comparatively) sluggish manned control limited in swiftness by the human brain. The tactical munificence of AI — like in the example above — is but one example of how AI is transforming the core processes in military aviation acumen. In a convoluted manner, the tactical accomplishments of AI are impinging on operational, strategic and doctrinal aspects of air power. This article looks at how AI is evolving as a constituent of air power.
Empowering the Manned Combat Aircraft
While there are no universally accepted benchmarks to define levels of sophistication in combat aircraft, the current leading-edge aircraft types are called the ‘fifth generation’ and are generally distinguishable from the previous or fourth generation by improved performance, enhanced stealth, advanced avionics and integrated digital flight systems.
According to Military Factory, a website cataloguing military aircraft, there are 14 fifth generation aircraft in various stages of development; out of these the Lockheed Martin F22 Raptor has been in service since 2005 (no longer being produced), the Lockheed Martin F35 Lightning entered service in 2015, and the Chengdu J20 Black Eagle in 2017. All others in that list are yet to reach operational status but a notable fact is that already there is a flurry of activity and interest around a ‘sixth generation’ of aircraft; unsurprisingly, AI is this generation’s most significant facet.
These aircraft are expected to be capable of optionally manned missions (with the same airframe capable of conducting AI-controlled missions), controlling a swarm of drones acting in both a defensive and reconnaissance role for the controlling aircraft, and providing battlefield data fusion with the aircraft acting as a network node capable of receiving and relaying data to multiple other platforms such as other aircraft, ground vehicles or satellites and processing that data onboard to dynamically generate new target lists or update mission parameters on the fly. To achieve these attributes, the focus is on AI.
The continued R&D going into sixth generation aircraft (nine nations are reportedly working on sixth generation fighters either individually or jointly) corroborates the inevitability of the manned aircraft being at the centre of air combat in the near future. However, manned aircraft are hugely expensive machines and carry precious human lives on board whose training costs make them valuable assets, difficult to lose and replace.
On the other hand, unmanned options can be less demanding in terms of design, can perform manoeuvres not limited by human physiological limits, and are comparatively far more expendable. This has been the driving rationale behind many R&D programmes to produce Unmanned Aerial Vehicles (UAVs) and Unmanned Combat Aerial Vehicles (UCAVs) that can act as ‘loyal wingmen’ to manned combat aircraft and support them in air-to-air and air-to-surface missions especially in high risk situations entailing aerial or ground based threats. At the leading edge of this research is Kratos Defence & Security Solutions’ XQ-58 Valkyrie, which first flew in March 2019 and is now undergoing tests as part of the Low Cost Attritable Strike Demonstrator (LCASD) programme being run for the USAF. USAF’s Skyborg programme (a developing software tool that allows engineers and researchers to develop autonomous capabilities) is aiding the quest for solutions in this direction by empowering the XQ-58 Valkyrie drone to be the unmanned wingman for the F-35.
A prototype autonomous UCAV is expected to be operational by 2023. The allure of the idea is self-evident: each XQ-58A Valkyrie costs just around USD2 million (as compared to the cost of a modern combat aircraft which approaches USD 150 million apiece). Moreover, its use does not risk a pilot whose training is horrendously expensive — an F-35 pilot’s cost of operationalisation runs to nearly USD11 million according to one estimate.
Australia, the UK, France, Germany, South Africa and China are also working on similar programmes; in addition, many other air forces aspire to include ‘loyal wingmen’ in their inventories in the future. The Boeing Airpower Teaming System under development (and presented at the Australian International Ari Show this year) will support manned Royal Australian Air Force aircraft; closer home, at the MAKS Air Show in August 2019, a Chinese exhibitor identified as CCKW revealed a stealth-endowed high speed drone that could fill the role of a ‘loyal wingman’. AI thus promises to considerably reduce the cost of aerial combat in terms of both — man and machine.
Unmanned strike drones already exist and in August this year, the Russian defence ministry released the first footage of its autonomous drone, named ‘Okhotnik-B’ or Hunter. The drone, based on the B-2 Bomber is designed to support manned aircraft by striking at targets on the ground. The craft was designed to weigh 20 tonnes and has a range of nearly 5,000 km. While the drone is largely autonomous, weapons use will require human approval, maintaining a ‘man in the loop’ who can critically analyse a combat situation and, if necessary, abandon an attack. Okhotnik’s combat AI system will eventually go into Russia’s sixth generation fighter.
Unmanned roles for rotary wing machines in the battlefield are also in an advanced stage of progression. The epitome is the KMAX unmanned helicopter which has served the US Marines admirably in Afghanistan and earned the admiration of the personnel it sustained and supported in difficult terrain. In a related development, in August this year, Japan’s NEC Corp demonstrated its autonomous flying passenger vehicle. With three wheels and four rotors, the machine resembles a quadcopter but is much bigger in size and can carry a passenger and deliver him autonomously. The possibilities are endless and AI drives them inexorably towards more extraordinary accomplishments.
Drones and Swarms
AI is also at the heart of swarming drones as military systems. US Defence Advanced Research Projects Agency’s (DARPA’s) OFFSET (OFFensive Swarm-Enabled Tactics) programmes involves quadcopter drones flying as a swarm to present an expansive aerial view to humans fighting on the ground. The swarm maps the neighbourhood below in an intricate pattern to create a real-time mosaic of surveillance over a selected objective area.
Project MAVEN is a Pentagon project intended to use machine learning and AI in order to differentiate people and objects in footage of thousands of hours collected and transmitted by a large number of drones. The resulting appreciation of ground situations in real time is game-changing and that too without the expensive and risky use of manned aircraft. AI is also being harnessed to enable the control of drone swarms by a manned fighter or its loyal wingman, in conjunction with manned/unmanned fighter formations if required, to achieve offensive and defensive missions under hostile conditions.
In September this year, DARPA threw open the Context Reasoning for Autonomous Teaming (CREATE) project for investigating new approaches to autonomous teaming of physically distributed groups of military autonomous machines in situations where centralised coordination is not possible or difficult. The project aims to explore the utility of AI on the autonomous formation of scalable machine-to-machine teams, and develop the theoretical foundations of autonomous AI teaming to pursue several simultaneous and unplanned mission goals — especially in degraded communications conditions.
An AI driven autonomous machine working in a team will be able to understand how its own sensing and actuation modes relate to common mission goals to help machine team members balance their resources autonomously. This is a little different from some of the remotely directed and controlled systems in use currently, which use far reaching communications networks to give humans fine-tuned control over those systems. Naturally, in case of a communications breakdown, machines are programmed to move autonomously towards mission accomplishment, albeit at reduced efficiency.
CREATE’s objectives are a bit different from those of swarming drones inasmuch as DARPA wants to find a framework that can communicate with a heterogeneous group of machines and — with the right measure of AI — get the machine-machine combinations to distinguish their role and capability in the context of a mission and act independently. In addition, they can meet multiple spontaneous goals that arise over the course of a mission. The end goal is a system that can learn and, especially, a system that can learn from mistakes and can improvise as circumstances require. The AI at the heart of CREATE renders it revolutionary in concept.
The Cold War may have reached moderate temperatures but the nuclear warheads at its heart have not disappeared and the US and Russia still hang on to 93 per cent of the world’s nuclear warheads, with the race between the two showing no signs of coming to an end despite overt iterations and symbolic gestures indicating otherwise. The deterring factors over the last seven decades have been premised on human cognition: a disincentive to launch nuclear weapons on an adversary for fear of the risk of unacceptable retaliation. Thus, mutually assured destruction and associated concepts have kept the world safe from an all-out nuclear war — but just.
The possibility of a nuclear exchange has never been ruled out; nor has been the probability of a significant section of one protagonist’s military decision making been compromised as a result of such an exchange. With the levels of speed and near human acumen that AI has displayed so far and with possibilities of even more brilliant and outstanding ability evident in the acceleration that AI development is revealing, there are experts who are proposing the placement of at least some of US nuclear assets under the control of an AI system. The underlying string appears to be the accelerated wartime decision making that AI would offer in contrast to human intervention.
While AI can be expected to process vast amount of information much faster than humans and also do it impassively and without emotion, the underlying principle of mutually assured destruction can still be sustained with the bonus that an AI system could still take a decision to launch even if there was no human decision forthcoming due to a massive first strike — the concept of an AI-enabled ‘dead hand’. Space constraints restrict us from discussing numerous other aspects of how AI would impinge on nuclear deterrence and nuclear warfare theory. Suffice it to say that AI is increasingly the focus of a debate on nuclear strategy.
Air power is all about air supremacy and air dominance — the capability to use the medium of air to own advantage while denying its use to the adversary. However, technological evolution has constantly tempered the manner in which air power is exercised. From the gladiatorial dogfights between knights of the air in their magnificent machines, flying them almost as extensions of their own bodies, we have progressed through many stages to Beyond Visual Range (BVR) missiles and air-to-surface weapons launched from long distances without sighting the targets they are aimed at. Stealth and speed have altered the mechanics of air power and now AI promises to rework the very thought processes behind air power use.
In the foreseeable future, sixth generation fighters working in tandem with loyal wingmen and swarms of drones could be the phantom apparitions that fight and aim to win tactical wars, dictate operational level deployments, and steer strategy and doctrine. Ethical and legal issues continue to nag military cogitative processes annoyingly but, as history has evidenced before, they will be brushed aside by military and strategic imperatives. Meanwhile, use of AI in military aviation proliferates steadily and inexorably and is making itself into an inalienable part of air power.