Artificial intelligence controls unmanned vessels at sea up to 300 km away

An artificial intelligence-controlled unmanned aerial vehicle has reached a new stage in autonomous air combat testing. The X-62A VISTA aircraft performed 27 intercepts against T-38 Talons during eight flights conducted at Edwards Air Force Base in California.

During these tests, an artificial intelligence system was used to analyze real sensor data and execute maneuvers without relying on pre-modeled information. The tests were part of the HAVE HEAT series and utilized tracking data received in real time from the Legion Pod IRST infrared system.

Thanks to this, the experimental aircraft was able to transition from a simple automated navigation mode to dynamic reaction to an aircraft considered a test threat. This progress was made possible by the 'Supermassive' artificial intelligence system developed by Lockheed Martin's Skunk Works division.

The technology was integrated into the X-62A in just three months, allowing the aircraft to process infrared information and convert these readings into maneuvering decisions during missions.

Testing an aircraft capable of operating independently in a combat situation requires more than just demonstrating the ability to follow a set trajectory. The system must simultaneously handle flight dynamics, assigned tasks, and a large volume of information coming from sensors installed on the aircraft or external platforms.

This scenario becomes particularly complex because real-world data can contain inaccuracies, ambiguities, and interference. Therefore, previous experiments with autonomous aerial vehicles often relied on digital environments where radar signals, ground station data, obstacles, and even the presence of enemy aircraft were artificially reproduced.

The HAVE HEAT series took the X-62A beyond this controlled environment. During the flights, the computer received information actually captured by the Legion Pod IRST system, which uses optical sensors to detect and track targets via their heat signature. Thus, the aircraft had to interpret this imperfect data and act upon it.

This experience also allowed for the abandonment of active radar during intercepts. Since the T-38 was tracked passively, using the heat emitted by the target, the X-62A could perform this task while maintaining electronic silence.

Despite the level of autonomy achieved, a pilot instructor remained onboard during the tests. Human presence remains necessary for conducting such tests in accordance with United States Air Force protocols and for the possibility of regaining manual control if needed.

For Lockheed Martin, the main result of the campaign is the link between perception and action. Ron Fellen, Vice President and General Manager of Skunk Works, stated that the tests demonstrated the system's ability to process classified infrared information and execute maneuvers considered critical for combat in real time.

The development of 'Supermassive' also draws attention due to the integration timeline. According to Lockheed Martin, ground preparation and full software integration into the X-62A were completed in only three months.

This result represents a significant shift in the type of tasks facing artificial intelligence within the experimental program. Instead of working only with pre-prepared information, the system had to convert real-world sensor data into flight responses during an intercept.