AIR 6500 – Closing the gap between artificial intelligence and human performance
Air forces, and defence forces more broadly, must find new ways of responding to an increasingly complex, lethal, and prolific series of threats, in particular tactical threats such as hypersonic weapons travelling at over Mach 5 that deny decision-making time and space.
Time and space are also critical commodities when it comes to investment horizons because – as was identified in the 2020 Force Structure Review – increasingly belligerent actors have now torpedoed the assumption there will be a 10-year warning time to prepare for major conflicts.
Technology will continue to play its part to counter threats and accelerate preparedness, but how does industry combine those commercial investment and technology development priorities with projects associated with national defence systems such as AIR 6500?
Put simply, how do you buy time and space?
Artificial intelligence (AI) is one such technology that can play a defining role in the improved human performance and decision-making necessary to achieve and maintain high levels of readiness, and to manage operational risk. However, there is still much to be done to establish the apparatus and pathways between research and development (R&D) investment, and trusted operational capability.
DOGFIGHTS
The US Defense Advanced Research Projects Agency (DARPA) announced on 12 October 2020 that it, “plans to re-use algorithms employed in the recent AlphaDogfight Trial (ADT) for research on artificial intelligence-enabled intelligence, surveillance, and reconnaissance (ISR); electronic warfare (EW); decision-making; and other purposes”. The USAF’s Advanced Battle Management System (ABMS) project is a likely area for application of the algorithms.
This broadening of the ADT’s scope is important for Australia because system-level applications of AI such as these offer the greatest return on investment. An ongoing focus on tactical application of AI to enhance dogfighting and similar tasks will offer only indirect benefit to projects such as Air 6500.
While the video of experienced F-16 pilot – callsign ‘Banger’ – being consistently outperformed by an AI-enabled simulator in the ADT provided compelling online entertainment, the reality is that the need for dogfighting has been in a steady decline since the end of World War 2.
As a consequence of air-to-air sensors and weapon systems becoming more capable, by the late 1980s air combat manoeuvring (ACM) within visual range was reduced to no more than 15 per cent of a squadron’s annual training syllabus. Fast forward to 2020, and that number is closer to three per cent.
A far higher-order function which demands the highest number of flying training hours involves beyond visual range (BVR) air-to-air targeting. An air combat system optimised for BVR exploits time and space using air power’s speed, technology, reach, and perspective from an advantageous position of height above land or sea.
Rather than being cast as the inevitable culmination of an air-to-air intercept, dogfighting in contemporary air warfare is typically the last-ditch outcome of a breakdown in the air-to-air targeting system. Engaging in a visual dogfight means time and space advantages have been negated. Instead, combat becomes a matter of survival through dynamic energy management, tactical decision-making, and superior performance in a duel defined by physics.
Banger’s contribution to DARPA’s ADT was to meet the desired aim of, “building human trust in AI as a step toward improved human-machine teaming”. He contributes to the establishment of the pathway between new technologies and fielded capability, addressing the future need to mitigate operational risk.
BRIDGING THE GAP
Project AIR 6500 provides a strategic rather than tactical mechanism for buying back time and space by investing in ISR, EW, decision-making, and many of the other purposes described in the DARPA program. But where might AI achieve the greatest return on investment?
AIR 6500 is defined as an integrated air and missile defence (IAMD) project. It fits within a multi-domain program led by Air Force as its capability manager.
A smart IAMD enterprise is designed, built, operated, and sustained at high levels of readiness as a means of providing early warning and managing operational risk. It consistently supports high quality decision-making and understands when an uncontained operational risk becomes a critical operational challenge. And finally, the glue which binds it together is provided by humans, because battles are won and lost in the mind. IAMD is therefore particularly well-suited to the application of AI.
In the insightful article Missile Defence: More than hitting a bullet with a bullet published by the Sir Richard Williams Foundation’s Central Blue blog, SQNLDR Robert Vine describes the doctrinal basis of IAMD, its layers, and its extra-theatre requirements:
• Counter-Proliferation
• Deterrence
• Counter-Force
• Active Defence
• Passive Defence
“Current doctrine treats the five layers of IAMD as separate activities that occur within their own ‘cylinders of excellence’; this will not suffice to create a robust defence against advanced threats,” he wrote. “Such a holistic defence requires Command, Control, Communication, Computers, Intelligence, Surveillance, and Reconnaissance (C4ISR) capabilities to fully combine these layers into a truly integrated defence system.”
The integrated defence system espoused by SQNLDR Vine is a multi-agency effort, with its C4ISR mission system elements delivered substantially by AIR 6500.
AI can be applied to each of these elements to enable human-machine teaming and, where appropriate, trusted autonomous capabilities. Other examples include tools to identify and track critical components of a hypersonic weapon supply chain, improved preparedness enabled by training systems, asset management systems, and autonomous logistics support systems using intelligent virtual reality and data analytics to enhance performance.
At its core, IAMD is substantially an integrated counter-air system with the addition of the specialist functions which address the increasingly complex requirements of missile defence. Existing counter-air systems already employ AI but these are largely constrained to individual tactical platforms. The challenge now is to scale up to the force level, and leverage AI to deliver enterprise-level outcomes which see greater synchronisation with resources in other domains and information-related capabilities.
In this regard, the most important applications of AI for AIR 6500 are in the management of data to enhance early warning, situational awareness (SA), and decision-making.
SCALING UP
The first human factors engineer and former Chief Scientist of the USAF Mica Endsley defined situational awareness as a tactical phenomenon which is, “the perception of the elements in the environment within a volume of time and space, the comprehension of their meaning, and the projection of their status in the near future.” Converting SA to its higher-order function – situational understanding – involves the integration of intelligence and a broader assessment of the battlespace.
AI can perform data analytics tasks faster and with greater accuracy than a human and identify risks across each IAMD layer to counter an adversary’s attempt to deny, degrade, deceive, disrupt, or destroy the enterprise. It can be used to generate high definition visualisations providing commanders with common operating pictures and ground truth.
Like SA, situational understanding exists in the mind, but it is the product of the integration of the C4ISR capabilities highlighted by SQNLDR Vine. These capabilities can be employed across strategic distances, imposing additional costs on adversaries, and making it much more difficult to encroach on decision-making time and space. At some point the cost becomes disproportionate to the desired outcome, requiring the adversary to think again.
PATHWAYS
For a large country like Australia with a relatively small but well-educated population, AI can make a difference, but it will need a clear pathway from R&D to be fielded operational capability. The first steps have already been taken.
Minister for Defence Industry Melissa Price announced a new initiative on 23 November 2020 to encourage Australian industry to develop new AI applications across Defence’s Information and Cyber, Maritime, Air, Space, and Land domains. The two priority areas highlighted were Intelligence Mission Data (IMD) to improve situational awareness in the battlefield, and intelligent virtual reality to enhance simulation, modelling, and training. Minister Price describes AI as, “a critical enabling technology which can deliver a decisive Defence advantage”.
To maintain this advantage and buy back the time and space denied by increasingly belligerent actors, investment in force level C4ISR capabilities, enterprise resource management tools, and decision support systems should complement the acquisition of ISR and firepower. Moreover, as demonstrated by DARPA, the Commonwealth needs to establish a robust, integrated commercial and technology framework for IAMD projects with the dual purpose of developing real capability while building trust and relationships along the way.
Commercialising AI technology within projects such as AIR 6500 will be a challenge as Australia does not have organisations like DARPA or rapid capability acquisition program offices like the US. The Rapid Capabilities Office (RCO), for example, is used for the acquisition of highly specialist and sensitive capabilities such as the B-21 bomber and, as recently announced by USAF Chief Architect Dr Will Roper, the Advanced Battle Management System (ABMS).
DARPA and the RCO between them have significant expertise and experience, as well as the clearances necessary to gain access to the most sophisticated technology available across the defence enterprise. And they have the ways and means to develop it and buy it.
The Australian leadership role will likely become a partnership between Defence and industry. Prime contractors, small and medium enterprises, academia, Defence Science Technology Group, Capability Acquisition & Sustainment Group (CASG), and organisations such as the Defence Artificial Intelligence Centre, and Trusted Autonomous Systems Defence Cooperative Research Centre will need to form an ecosystem to ensure the delivery of both operationally relevant and commercially viable capability.
The people in this partnership and their mission focus are as important as the technology itself. Improving situational understanding through investment in new sensing and decision-making apparatus is one way of creating time and space based upon an increasingly sophisticated relationship between human and machine. It needs more than just doing things faster.
The alternative is to get stuck in an energy-sapping strategic dogfight with the initiative seized by an adversary. It is time to get real about Artificial Intelligence – just ask Banger!