The Air Force Test School of USA and the Advanced Research Projects Agency Defense(DARPA from Defense Advanced Research Projects Agency) claim to have achieved a major breakthrough in machine learning by demonstrating that the software artificial intelligence can fly a modified F-16 fighter jet in a dogfight against humans.
They refer to the application of machine learning by the USAF and DARPA to an X-62A VISTA.
"We've imagined the potential for autonomous air-to-air combat for decades, but reality has remained a distant dream until now," said Air Force Secretary Frank Kendall. “In 2023, the X-62A has overcome one of the most significant obstacles in aviation. This is a moment of transformation, made possible by breakthrough innovation.”
DARPA has been testing an artificial intelligence agent software to handle airplane simulations for several years. Her schedule Air Combat Evolution (ACE) dates from 2020, when the AlphaDogfight tests pitted human pilots against each other in a flight simulator against an opponent AI.
The AI software won that competition, but it had one advantage – it was allowed to fly at speeds that would be too stressful on a real aircraft and create g-forces that would injure a human.
"The machine learning approach relies on analyzing historical data to make informed decisions about both present and future situations, often discovering insights that are subtle to humans or difficult to express through rules," said Daniella Rus, Director of MIT CSAIL. in a DARPA video you can see below.
"Machine learning is extremely powerful in environments and situations where conditions fluctuate dynamically making it difficult to establish clear and robust rules."
Dogfights between military aircraft are largely a dynamic scenario. But machine learning has a drawback that needs to be overcome. It must be sufficiently explainable and verifiable for military personnel to trust it and for aviation authorities to be able to certify systems that implement such code.
The X-62A is essentially an F-16 that has been integrated with a flight simulator so that machine learning applications can operate the plane.
Watch the video
