A glimpse into a possible AI-enabled defence future
15 Oct 24
Technical and Strategic Services OBUElectronic Warfare
BY Andy Rogerson
Andy Rogerson, our EW & Electromagnetic Spectrum Operations Training & Consultancy Expert muses.
Let's set the scene...
“The S-700 Plamya (‘Flame’) system, codenamed by NATO as GARGANTUA, had searched its designated region of airspace. The system was optionally crewed and had been remotely operated to its operational location. It was then placed into fully autonomous mode where it relayed its ‘picture’ into the wider air defence system.
It autonomously selected a pseudo-random search pattern that optimised its search efficiency whilst simultaneously presenting as small an electromagnetic presence as possible to any adversary electromagnetic surveillance.
It detected the presence of two potential targets and autonomously switched waveform to try and provide identification via Non-Cooperative Target Recognition.
The ‘unknowns’ were F-45s. Electromagnetically savvy and controlling their emissions. Very few – an almost undetectable few – returned in the direction of the S-700.
It was irrelevant. The S-700 had communicated autonomously with other sensors in its network, across air, land and space. Those sensors had received those stray F-45 transmissions, direction finding and relaying sufficient data for the S-700 to compute 3D coordinates.
Recognising the targets as two Western F-45’s, the GARGANTUA identified a risk of counter-detection from the F-45 Electromagnetic Support Measures and switched mode again to its track pattern. It used ‘negative’ tracking to avoid placing signal on target and pseudo-random noise within its pulses as a modulation. It had the processing and memory to integrate and identify its own ‘noise’ from background.
The S-700 launched its missiles passively to an inertial guided point en-route to the target. The F-45’s could have no knowledge that the missiles were inbound…”
Tom Clancy it isn’t, but it might be a (slightly fanciful) story from today. Modern systems have the ability to datalink, communicate and present a threat with minimal warning time to the target. Modern systems can present such threats as to require concurrent jamming or kinetic denial of multiple parts of the threat system simultaneously.
How could this change in the future with the development of 6th generational platforms? What defines 6th gen?
Let’s moot a possible future and continue the story…
Conveying a futuristic 7th or 8th generational aircraft. Created using DALL-E
“The wingman F-45’s Electromagnetic Support Measures equipment had detected the S-700 as soon as it switched mode. It did not rely on library matching to recognise the S-700 – to try would be folly for the threat had an almost inexhaustive range of parameters to select its parametrics from. Instead, the F-45’s processing of complex reinforcement learning, semi-supervised systems and behaviour-relating, machine-learning systems had deduced the probable threat from the exhibited S-700 behaviours and from virtual models that resided in its memory.
The F-45 autonomously communicated the presence of the S-700 across its intra-formation datalink, to its semi-autonomous AI wingmen and wider to multi-domain intelligence, surveillance and reconnaissance assets.
The F-45’s high-gain aperture was used to transmit a probing jamming pulse at the S-700. Suspecting counter-detection, the AI within the S-700 switched modes.
For the F-45, instead of becoming a complex problem, this action only served to highlight a potential weakness in the S-700 that would be exploited. The F-45’s Electromagnetic Counter Measures was not simply responding to incoming waveforms accordingly to a library-like look-up response. Instead, it contained a reinforcement learning system able to respond to the change in its environment. The machine system was ‘rewarded’ every time it produced an optimised state. In this case, protecting the F-45 and defeating the S-700. Like traditional engagements before it, it was still a contest between two players. Now they were two AI systems, and the S-700 would lose.
No threat system has developed the ability to physically teleport location, and the S-700 was no exception. That meant all emissions regardless of mode or parametric originated from the same location. The F-45 now electromagnetically tracked the S-700 emission and produced near isochronous counter transmissions to keep the S-700’s AI in a sequence of consecutive mode changes that the S-700 could not escape from, rendering it almost useless.
The lead F-45 AI had commanded its AI drone semi-autonomous wingmen to transmit similar adaptive deception jamming into the seekers of the two missiles. The wingman’s reinforcement AI had calculated the optimal location and parameters and was continually calculating apparent electromagnetic attack performance. Ultimately, if it calculated a risk that the jamming was not effective, the autonomous wingmen would act as sacrificial barriers to preserve mission success of the crewed assets.
Simultaneously, all the transmissions from the S-700 were being tracked by the F-45 but also by other remote ISR sensors. This location and parametric information was already being analysed by AI engines and passed to other units by way of updated mission data. Communications intelligence between the S-700 and associated units was already being analysed by Natural Language Models and used to support operational decision making.”
The notion of artificial intelligence, shaped like a human. Created using DALL-E
Is this a realistic look at things to come? Perhaps.
What I think is realistic is that data dominates. Reports indicate 90% of the world’s data was made in the last two years. In a defence context, the period of 2006-2011 saw a 1500% increase in the volume of US STRATCOM data collected. To continue to analyse these volumes of data would require personnel resource increases in the millions.
‘Stealth’ has had its genesis, and whilst it will remain a vital element, it is unlikely to define future generations of air power. Data and information have dominated 5th gen. Perhaps 6th generational air power will be defined by the successful implementation of AI systems across sensing, sense-making, correlating, classifying, recognising and communicating across both AI-supported crewed and AI-autonomous uncrewed platforms, comprehending that data flood to achieve mission goals and working in harmony.
AI is not tomorrow’s technology. It is today’s technology and its influence in every aspect of our lives will only grow. We will see AI derive potential courses of action that will appear illogical to us, yet we must develop methods of assurance and learn to trust in our new toolsets.
The alternative is to drown in data.
Latest News
We're celebrating International Volunteers Day
05.12.24
In honour of International Volunteers Day, we spoke with a few generous members of team Inzpire to find out more about their volunteering efforts, and why they find it so rewarding.
Need Help?
Our team are here to help. If you have any questions, no matter how big or small, please get in touch.
+44 (0)1522 688231
.(JavaScript must be enabled to view this email address)