The University of Iowa and Lockheed Martin have come together to demonstrate the use of AI in air-to-air intercept scenarios. The AI controlled the jet, controlling its heading, speed, and altitude when in command. The AI flew a full-scale Aero L-29 Delfín jet.
The Aero L-29 Delfín (meaning “Dolphin”) is a trainer jet developed and manufactured by Aero Vodochody, a Czechoslovakian aviation manufacturer. It is the country’s first locally designed and constructed jet aircraft.
Eight test cases were performed for each flight to test the AI pilot in different scenarios, such as standard head-to-head fights, off-aspect encounters, missile support, and missile defeat situations.
The Lockheed Martin team was impressed by the overall tests and results. They observed a “smooth transition of learned behaviors” from simulation to the real world and noticed that the AI agent acted “purposefully and decisively.”
“This was the first live exercise of the new flight interface; seeing the separate components successfully integrate on the L-29 to demonstrate new capabilities is thrilling. The complete system performed even better in live flight than simulation,” said Dr. Tom “Mach” Schnell, OPL professor at Iowa Technology Institute.
“Live flight tests are a crucial aspect of advancing our expertise in AI and autonomy. These flights are powerful demonstrations of our ability to quickly and affordably develop and test operationally relevant AI capabilities,” said Matthew “Gabe” Beard, Lockheed Martin Skunk Works autonomy/AI and machine learning engineering manager.
These tests are part of a more extensive program to develop and evaluate AI-driven autonomy for air-to-air missions quickly. In the same context, additional flights are scheduled for later this year.
This development follows the news that the USAF successfully pitted an AI-piloted F16 against a real human being. The X-62 did have a human “failsafe” pilot in the cockpit but was otherwise entirely AI-controlled. The AI pilot did win the duel, but the catch was that the X-62 broke the rules.
Lockheed Martin explains that future tests will build on these accomplishments by incorporating more aircraft into offensive counter-air and battle management scenarios, increasing complexity.
Lockheed Martin also intends to provide machine learning solutions that augment human performance by processing, combining, and analyzing large volumes of data to provide our customers with actionable intelligence and a strategic edge.