The U.S. Air Force Test Pilot School and the Defense Advanced Research Projects Agency (DARPA) have achieved a significant milestone in aerospace machine learning with the X-62A Variable In-Flight Stability Test Aircraft (VISTA). The jet has successfully demonstrated AI’s capability in autonomous air-to-air combat, engaging in combat scenarios against manned F-16 aircraft, performing complex maneuvers at high speeds and proving the feasibility of non-deterministic AI in flight-critical situations.
This work involved over 100,000 lines of software updates and 21 test flights. The safety measures were stringent, with AI flights conducted under the supervision of safety pilots who had the authority to disengage the AI, although this was never needed during the tests.
DARPA’s Air Combat Evolution (ACE) program’s broader objective extends beyond mere dogfighting; it aims to refine the use of autonomous AI systems across various aerospace tasks. The program stands is a collaborative effort involving multiple partners across academia, government, and private industry, including Johns Hopkins University, MIT’s Computer Science and Artificial Intelligence Laboratory, and companies like Lockheed Martin Skunk Works and Shield AI.
Image: DARPA
