| +27 31 564 1319 

Connecting Skies Bridging Continents


Royal Air Force Test Pilot School and the Defence Advanced Research Projects Agency have been recognised as finalists for the 2023 Robert J. Collier Trophy, a prestigious award celebrating recent strides propelling the era of machine learning within the aerospace domain.

Collaborating closely, the teams orchestrated groundbreaking tests employing artificial intelligence algorithms aboard the X-62A VISTA aircraft, a pivotal aspect of DARPA’s Air Combat Evolution (ACE) programme.

Secretary of the Air Force, Frank Kendall, expressed, “The concept of autonomous air-to-air combat has tantalised us for decades, yet it has remained elusive until now. In 2023, the X-62A shattered one of the most significant barriers in combat aviation. This marks a pivotal moment, made feasible by the remarkable achievements of the X-62A ACE team.” Secretary Kendall is poised to personally experience AI in a simulated combat setting aboard the X-62A VISTA during an upcoming test flight at Edwards.

Within the span of a mere calendar year, the teams progressed from the initial integration of live AI agents into the X-62A’s systems to conducting the first AI versus human within-visual-range engagements, commonly referred to as dogfights. Over 100,000 lines of flight-critical software were modified across 21 test flights.

Dogfighting, a profoundly intricate scenario, served as a litmus test for the X-62A, successfully demonstrating the safe utilisation of non-deterministic artificial intelligence within the aerospace domain. These AI dogfights pitted the X-62A VISTA against manned F-16 aircraft above the skies of Edwards. Commencing with defensive manoeuvres to ensure initial flight safety, the engagements transitioned to offensive high-aspect nose-to-nose encounters, with the dogfighting aircraft closing in as near as 2,000 feet at speeds of 1,200 miles per hour.

This groundbreaking application of machine-learning-based autonomy in flight-critical systems lays the groundwork for future advancements in aerospace AI, promising enhanced safety and reliability across both commercial and defence sectors.

Colonel James Valpiani, commandant of the Test Pilot School, remarked, “The X-62A serves not only as an exceptional platform for research and advancing testing methodologies, but also for grooming the next generation of test leaders. Industry can draw upon the achievements of the X-62A ACE team as a paradigm shift in ensuring the safety, efficiency, effectiveness, and responsibility of future capabilities.”

Whilst traditional autonomy has been practised for decades, the integration of machine learning has historically been restricted due to elevated risks and the absence of independent control. Safety pilots remain onboard the X-62A with the capacity to disengage the AI independently. However, throughout the dogfights over Edwards, test pilots did not find it necessary to activate the safety switch. “We must have confidence in these algorithms to employ them in real-world scenarios,” remarked Lieutenant Colonel Ryan Hefron, ACE programme manager for DARPA.

While dogfighting served as the primary testing scenario, it did not constitute the ultimate objective. Bill Gray, the school’s chief test pilot, elucidated, “It’s easy to perceive the X-62A ACE programme solely in terms of autonomous dogfighting capabilities, but that overlooks the essence. Dogfighting presented the problem to be solved, enabling us to commence testing autonomous artificial intelligence systems in flight. Every lesson we learn applies to every task conceivable for an autonomous system.”

Share the Post: