The views and opinions expressed or implied in WBY are those of the authors and should not be construed as carrying the official sanction of the Department of Defense, Air Force, Air Education and Training Command, Air University, or other agencies or departments of the US government or their international equivalents.
By Dr. Peter Layton
/ Published March 26, 2021
It is a time of rapid disruptive technological change and none more so than in the field of artificial intelligence (AI). While developed by and for the commercial sector, AI’s apparent potential for military use is now pushing armed forces globally to begin experimenting with embryonic AI-enabled defense systems. There may be a considerable first-mover advantage to the country that first understands AI adequately enough to change its existing human-centered force structures and embrace AI warfighting.
In a new Joint Studies Paper published by the Australian Defence College, I explore sea, land and air operational concepts appropriate to fighting near-to-medium term future AI-enabled wars.1 With much of the underlying narrow AI technology already developed in the commercial sector, this is less of a speculative exercise than might be assumed. Moreover, the contemporary AI’s general-purpose nature means it’s initial employment will be within existing operational level constructs, not wholly new ones.
Here, the focus is the air domain. To make the discussion manageable, it’s tightly constrained to air defense and avoids broadening into joint and combined operations. Even so, this gives scope to explore operational concepts that might stimulate thinking about the future and preparing for it. In all this, it is important to remember that AI enlivens other technologies. AI is not a stand-alone actor, rather it works in the combination with numerous other digital technologies. It provides a form of cognition to these.
For the near-to-medium term, AI’s principal attraction will be its ability to quickly identify patterns and detect items hidden within very large data troves. While also giving mobile systems a new autonomy, AI will transform sensing, localizing, and identifying objects across the battlespace. Hiding will become increasingly difficult. AI is not, however, faultless. Its well-known problems include being able to be fooled, brittleness, an inability to transfer knowledge gained in one task to another, and high data dependence.2
AI’s principal warfighting utility then becomes ‘find and fool’. With its machine learning, AI can be exceptional at finding items hidden within a high clutter background, certainly better than humans and tremendously faster. On the other hand, AI can be fooled through various means. Its great finding capabilities lack robustness.
Sensor Fields and C2
The ‘find’ starting point is placing a large number of low-cost Internet of Things (IoT) sensors in the optimum land, sea, air, space, and cyber locations in the areas across which hostile forces may transit.3 In some respects, this idea is already in place in integrated air defense system (IADS) concepts that include chains of surface-based radar stations complemented by airborne early warning and control aircraft to detect high and low flying aircraft.4 The war-in-the-air AI-enabled defense concept suggests a massive supplementation of this existing high-cost, limited-number sensor deployment through using large numbers of AI-enabled small, low-cost surface and airborne sensors.
The smaller elements of the expanded IoT sensor field can use AI edge-computing with partly processed data sent through the cloud to a fusion center and then into the command and control system.5 These smaller IoT sensors could be active, short range radar emitters, however, power constraints may limit such usage. More likely are passive IoT sensors detecting emissions across the electromagnetic spectrum including the acoustic, ultraviolet, infrared, radio, and radar bands. Each sensor individually may be relatively low performance, but when its outputs are combined with potentially several hundred others, air traffic may be able to be tracked and identified, perhaps in three dimensions.
The surface air defense IoT sensors might be fixed and persistent whereas sensor equipped uncrewed air vehicles (UAV) could have endurance varying from some hours up to a day. There are some emerging IoT applications that might considerably increase this endurance including high-altitude balloons, smallsats, and pseudo-satellites, all potentially incorporating AI.6
Having a large IoT sensor field that uses passive detection means that penetrating aircraft must avoid using transmitting systems such as radars, data links, and communications so as to try to avoid detection. Even so, normal aircraft emissions such as noise, temperature, and the visual signature may still reveal the aircrafts’ presence. In this, having a deep IoT sensor field is important. As they approach known sensors, aircraft may maneuver to limit their emissions, particularly those emanating from the aircraft’s forward sector. A deep field means that a penetrating aircraft may be detected on its flanks and in its rear sector even if it is not picked up when directly approaching.
The very large IoT sensor field made possible by AI would feed partly processed data through the cloud into a fusion facility where AI would undertake further processing. In considering these steps, the Observe-Orient-Decide-Act model is useful.7 In ‘Observe’, AI would be involved as noted in each IoT’s edge computing and then again in the fusion center. In ‘Orient’, AI would play an important part in the battle management system.8 AI would not only produce a comprehensive near-real time air picture but also predict the enemy air courses of action and movements.
The next AI layer handling ‘Decide’ and aware of friendly air defense units availability, would pass the human commander for approval a prioritized list of approaching air targets to engage, the optimum types of cross-domain attack to employ, the timings involved, and any deconfliction considerations. In this, humans would remain deeply engaged through in-the-loop or on-the-loop control.
After human approval, the next AI layer would assign the preferred weapons to each target passing the requisite targeting data automatically, ensure deconfliction with friendly forces, confirm when the target was engaged, and potentially order weapon munition resupply. This final step, ‘Act’, would be primarily an AI function.
With several high-performance UAVs already flying, developing a within-visual-range, air-to-air combat UAV that uses AI for tactical decision making appears a straightforward engineering task.9 Indeed, United States Air Force (USAF) plans to repeat the 2020 AI piloted aircraft versus a human piloted aircraft experiment in 2024, but this time not with simulations but with full-scale tactical aircraft.10 An operational, optimised AI-enabled short-range, dogfighting UAV could be smaller, lighter, and lower cost than a crewed aircraft and in the defensive role may not need to be armed to disrupt an incoming adversary air attack.
The UAV might simply be allocated by the command and control system for an adversary aircraft to engage, close, and begin dogfighting. The crewed aircraft would be distracted, and its attack approach disrupted making it vulnerable to other crewed weapon systems. Moreover, if the adversary crewed aircraft maneuvers it will have a higher rate of fuel usage and may need to quickly breakoff so as to be able to return to its more distant home base.
On the other hand, an armed AI-enabled fighter could operate under human in-the-loop or on-the-loop as appropriate. The downside is that arming an aircraft creates engineering design issues, raises connectivity worries, has significant law of armed conflict implications and imposes tactical concerns. For many reasons, it may be preferable to have a UAV that engages and ‘locks on’ to an adversary aircraft and then accompanies it, continuously broadcasting to all the adversary’s track and details.
An AI-enabled aircraft could operate in the combat air patrol (CAP) or ground alert interceptor (GAI) roles.11 For CAP, the UAV would need to be larger to allow a useful endurance on station; although, for a similar sized airframe this is likely to much greater than a crewed aircraft can provide. However, the larger the UAV, the more design and operating complications are introduced.
For GAI, the UAV could be smaller and perhaps more like a missile than an aircraft. For example, USAF’s experimental XQ-58A Valkyrie UAV becomes airborne from a static launcher and lands using a parachute; there are proposals for basing this UAV in relocatable shipping containers.12
If a GAI AI-enabled UAV fighter does not need airfields, defense in depth approaches become easier but crucially new concepts like distributed air defense become possible. Within the IoT sensor field may be dispersed GAI AI-enabled UAV fighters, able to be remotely dispatched by the command and control system on short-range, quick reaction intercept missions, and working in conjunction with crewed aircraft flying CAP. Again, such UAVs do not necessarily need the complexity of being armed to be useful.
Importantly, in such an AI-enabled IADS there would be a separation of tasks between humans and UAVs. The humans would be responsible and accountable for the higher-level cognitive functions such as developing an overall engagement strategy, selecting and prioritizing targets, and approving engagements. The AI would undertake lower-level cognitive functions such as maneuvering the aircraft and dogfight tactics.13
Fool Function AI
The ‘find’ function of AI requires complementing by the ‘fool’ function. An adversary needs considerable information about the target and its defenses to reliably mount successful attacks. AI-enabled ‘fool’ systems could be dispersed across the battlespace both physically and in cyberspace. The intent is to defeat the adversary’s ‘find’ by building up a misleading or at least confused picture. AI-enabled ‘fool’ systems may also be used in conjunction with a sophisticated deception campaign.
In addition, small mobile, edge computing systems widely dispersed could create complicated electronic decoys by transmitting a range of signals of varying fidelity. These systems might be mounted on UAVs for the greatest mobility, although uncrewed ground vehicles using the road network may also be useful for specific functions such as pretending to be mobile SAM systems. The aim is simply to obscure the battlefield for the quite limited time an attack is in progress.
A more costly approach might be UAVs that electronically replicate the defending fighters creating an impression of unexpectedly large numbers of fighters airborne in various CAP stations defending the target area. This may encourage the adversary attackers to retire to avoid seemingly high attrition.
The ‘fool’ function can be further extended and integrated with passive defense measures and operating approaches. An airbase is often established well in advance of hostilities and so can be designed to be resilient under attack. However, modern precision guided weapons have made resilience through hardening less efficacious with dispersion now favored.14 AI could make this approach more practical than it has been for several decades.
A permanent airbase could have several satellite airfields around it. These airfields can be designed to have a limited life of weeks or months, rather than decades as with the permanent airbase. In time of conflict, aircraft from the permanent airbase can continually move around between it and the short-term airfields. This movement would be closely integrated with the AI-enabled ‘fool’ actions. The intent would be to deceive, perplex, and confuse the adversary so they did not know where to attack and then, after finally having made a decision, attack where no friendly aircraft were. Such a tactic increases the ‘fog of war’, offers some possibilities for manipulating adversary perceptions, and purposefully harms adversary force combat effectiveness.
An adversary has only a limited number of aircraft, stand-off weapons, and ballistic missiles to employ in a counter-air campaign. Attacking airfields where there are no friendly aircraft located exposes crewed aircraft to unnecessary attrition while using stand-off weapons and ballistic missiles simply wastes scarce, and in a short conflict, irreplaceable stockholdings. The combination of ‘fool’ AI and physical dispersion accordingly supports both air defense aims in reducing the effectiveness of adversary air attacks and exposing the adversary to attrition.
A major problem with such aircraft dispersion notions has previously been that operating combat aircraft from several short-term airfields requires significant and costly duplication of logistic support and associated workforce across multiple locations. AI-enabled systems can overcome this issue.
In terms of logistic support, the permanent airbase can have well-established corridors linking its large warehouses and consumable supply storage facilities to the short-term airfields. For the warehousing end of the support and supply corridors, there is considerable existing AI-enabled technology that can be employed.
State-of-the-art warehouses already feature real-time monitoring of inventory; real-time ordering using AI machine learning, the cloud, big data and IoT; order picking by advanced robotics; and stock movement by autonomous vehicles. Some warehouses are now embracing on-demand 3D printing to meet one-time requests for spare parts and save on carrying large part inventories for older equipment. Logistics control towers have been introduced that integrate digital information from numerous sources and use big data analytics to provide a real-time ‘big picture’ of the complete supply chain, including transportation activities.15 The same technologies could be used to control and direct consumable supply storage facilities.
In terms of the corridors along which supplies and support could flow, AI-enabled logistics could use robot trucks employing follow-the-leader autonomy. This capability, also called ‘platooning’, has the lead truck crewed and guiding several uncrewed vehicles following closely behind.16 Devising uncrewed airbase logistics distribution trucks would be a much easier task technically then for land force resupply vehicles. The former would operate principally on pre-surveyed, paved, or graded roads and probably use GPS.
At the short-term airfield end of the logistic corridors, AI-enabled systems could be omnipresent.17 Using AI, machine learning, big data, cloud computing, the IoT, autonomous operations, and robotics, such bases could generate aircraft sorties, faster, and with considerably fewer people than would be needed today. Robot turns of serviceable aircraft, including refueling and weapons loading, could be possible. AI-enabled predictive maintenance would make unscheduled maintenance rare, or at least uncommon. The airfields might appear uninhabited being managed remotely by engineering and logistics personnel at central control centers at the permanent airbases or elsewhere. Such airfields might even generate their own power using renewables and batteries to become semi-self-sufficient.
The equipment needed to make such a short-term airfield functional might be already installed, simply waiting for the conflict for be activated. On the other hand, the airfields could have the necessary networks in place able to quickly incorporate ‘plug and play’ systems and vehicles into the short-term airfield’s own system of systems when these were delivered, possibly in the initial follow-the-leader truck convoys.
AI appears likely to be the modern ‘ghost in the machine’ infusing many, perhaps most, military machines. It will potentially create a different way of waging war-in-the-air as this air defense focused discussion illustrates. Given the decades usually needed to reorient air forces in new directions, there is no time like the present to start the journey.
Dr. Peter Layton
Dr. Peter Layton is a Visiting Fellow at the Griffith Asia Institute, Griffith University, and an Associate Fellow at the Royal United Services Institute. He has extensive aviation and defense experience and, for his work at the Pentagon on force structure matters, was awarded the US Secretary of Defense’s Exceptional Public Service Medal. He has a doctorate from the University of New South Wales on grand strategy and has taught on the topic at the Eisenhower School. His research interests include grand strategy, national security policies particularly relating to middle powers, defense force structure concepts and the impacts of emerging technology. The author of ‘Grand Strategy’, his posts, articles and papers may be read at: https://peterlayton.academia.edu/research.
1 Peter Layton, “Fighting Artificial Intelligence Battles Operational Concepts for Future AI-Enabled Wars,” Joint Studies Paper, No. 4, 2021, https://www.defence.gov.au/.
2 Peter Layton, “Algorithmic Warfare: Applying Artificial Intelligence to Warfighting,” Air Power Development Centre, 2018, https://airpower.airforce.gov.au/.
3 Steve Ranger, “What Is the IoT? Everything You Need to Know about the Internet of Things Right Now,” ZDNet, 3 February 2020, https://www.zdnet.com/.
4 Maj Peter W. Mattes, USAF, “What is a Modern Integrated Air Defense System,” Air Force Magazine, 1 October 2019, https://www.airforcemag.com/.
5 Duncan Stewart et al., “Bringing AI to the Device: Edge AI Chips Come into Their Own,” Deloitte, 9 December 2019, https://www2.deloitte.com/.
6 Michael Spencer, “Pseudosatellites: Disrupting Air Power Impermanence,” Air Power Development Centre, 2019, https://airpower.airforce.gov.au/.
7 Sarah Lewis, “OODA Loop,” TechTarget, June 2019, https://searchcio.techtarget.com/.
8 Chris Westwood, “5th Generation Air Battle Management,” Air Power Development Centre, 2020, https://airpower.airforce.gov.au/.
9 Joseph Trevithick, “Navy Establishes First Squadron to Operate Its Carrier-Based MQ-25 Stingray Tanker Drones,” The Drive, 1 October 2020, https://www.thedrive.com/; and Kyle Mizokami, “Russia’s ‘Hunter’ is Unlike Anything in America’s Arsenal,” Popular Mechanics, 10 August 2020, https://www.popularmechanics.com/.
10 Patrick Tucker, “An AI Just Beat a Human F-16 Pilot in a Dogfight — Again,” Defense One, 20 August 2020, https://www.defenseone.com/; and Secretary of Defense Dr. Mark T. Esper, “Secretary of Defense Remarks for DoD Artificial Intelligence Symposium and Exposition,” US Department of Defense, 9 September 2020, https://www.defense.gov/.
11 “Combat Air Patrol,” Wikipedia, https://en.wikipedia.org/; and Lt Col Ernani B. Jordao, “An Investigation of the Combat Air Patrol Stationing in an Integrated Air Defense Scenario,” (BS Thesis, Brazilian Air Force Academy, 1971), https://apps.dtic.mil/.
12 Joseph Trevithick, “This Containerized Launcher for the XQ-58A Valkyrie Combat Drone Could Be a Game Changer,” The Drive, 16 October 2019, https://www.thedrive.com/.
13 Col Daniel Javorsek, USAF, “Air Combat Evolution (ACE),” DARPA, https://www.darpa.mil/.
14 Miranda Priebe et al., “Distributed Operations in a Contested Environment: Implications for USAF Force Presentation,” RAND Corporation, 2019, https://www.rand.org/.
15 Stefan Schrauf and Philipp Berttram, “Industry 4.0: How Digitization Makes the Supply Chain More Efficient, Agile, and Customer-Focused,” Strategy& and PWC, 7 September 2016, https://www.strategyand.pwc.com/.
16 “Oshkosh Defense Delivers Autonomous Vehicles,” Nation Shield, Military and Strategic Journal, 2 February 2020, http://nationshield.ae/.
17 Peter Layton, “Surfing the Digital Wave: Engineers, Logisticians and the Future Automated Airbase,” Air Power Development Centre, 2020, https://airpower.airforce.gov.au/.
Wild Blue Yonder Home