The US Air Force’s experimental combat drone, the XQ-58A Valkyrie, has successfully completed a three-hour flight at the Eglin training ground in Florida, under the control of artificial intelligence. This test flight is part of the Skyborg Vanguard program, which aims to develop AI and machine learning capabilities to safely control the drone and perform tactically relevant tasks. The AI was trained using a simulator, with algorithms developed by the Afrl’s Autonomous Air Combat Operations (Aaco) and spending millions of hours on high-precision training and flights on the X-62 Vista.
The XQ-58A Valkyrie is a joint project between Kratos Unmanned Aerial Systems (Kuas) and the US Air Force Laboratory (AFRL), with the main goal of developing autonomous combat drones that can operate independently, in cooperation with other drones, and in conjunction with human operators. Colonel Tucker Hamilton, AI testing and operations head in DAF, stated that the mission confirmed the effectiveness of the multi-level security system on the drone controlled by AI, and demonstrated the ability to solve tactically relevant problems in the air. This flight officially confirms the possibility of developing AI agents/ML capable of performing modern air tasks.
The project’s key idea is to provide the US Air Force with a robotic alternative that can accompany and cooperate with manned fighter aircraft, as well as operate in particularly dangerous areas, thus reducing costs. The Valkyrie drone has a wingspan of 6.7 meters, a maximum speed of 1050 km/h, a ceiling of 13,715 meters, a flight range of up to 3941 km, and can carry eight types of weapons onboard. Brigadier General Scott Kane, commander of Afrl, emphasized the importance of AI in future military operations and the need for collaboration between the government, academic institutions, and industry partners.