This article is published in Aviation Week & Space Technology and is free to read until May 24, 2024. If you want to read more articles from this publication, please click the link to subscribe.
The U.S. Air Force will start modifying F-16s to serve as autonomy testbeds; collaborative combat aircraft are set to enter service by the end of fiscal 2028.
A dedicated fleet of flying testbeds may soon inject the next level of realism into combat training for the artificial intelligence-generated pilots that the U.S. Air Force expects will be ready for operational service by the end of fiscal 2028.
Artificial intelligence (AI) pilot training for the service started with the DARPA Air Combat Evolution program’s 2019-20 AlphaDogfight trials, in which eight research teams developed algorithms to compete one-on-one in Lockheed Martin F-16 flight simulators. AI pilot training continued through flight tests under the Air Force Research Laboratory’s Skyborg program, in which a government-owned algorithm learned to fly in formation, navigate and respond to pop-up situations on real autonomous aircraft such as the Kratos XQ-58 Valkyrie.
- Modifications begin on first VENOM Lockheed Martin F-16
- DARPA is expected to name AIR participants soon
Now the Air Force plans to take the next step with the Viper Experimentation and Next-Gen Operations Model (VENOM). AI-generated pilots will fly specially modified F-16Cs on training flights for defensive and offensive counter-air missions of increasing complexity, factoring in beyond-visual-range engagements, sensors and electronic warfare capabilities. A complementary set of algorithms will also be trained to run in the cockpits of fighters flown by humans to ease their transition from low-level tacticians to high-level mission commanders.
The first of several flying testbeds—a 32-year-old Lockheed Martin F-16C reassigned from the 16th Weapons Sqdn.—landed on April 1 at Eglin AFB, Florida. By adapting F-16Cs into testbeds, the Air Force expects to accelerate the pace of testing compared with that of uncrewed aircraft systems, such as the XQ-58, which face more restrictions.
A human-piloted F-16C can take off like any other aircraft in normal airspace and fly to a dedicated military training range. There, the human pilot would transfer control of the fly-by-wire fighter to an AI-generated pilot while the human pilot serves as a safety observer.
The F-16Cs also carry an onboard suite of sensors that the AI-generated pilot can control, enabling a degree of realism unavailable to the perfect situational awareness granted to the simulated AI agents during the Alpha-Dogfight trials. The flight of a VENOM aircraft would not be a onetime demonstration; rather, the fleet would establish the capacity to test and train algorithms for autonomy and manage collaborative combat aircraft (CCA) for decades to come.
“One of the things VENOM does is it gives us multiple aircraft,” Air Force Test Center Commander Maj. Gen. Evan Dertien said at the Air & Space Forces Association Warfare Symposium in February. “Two, we can tie it in with the sensors. But more importantly, it helps set up the foundation to make sure we have a test range ready to develop CCA and autonomy.”
The VENOM testbed program is launching at a key moment: The Air Force plans to select at least two companies in the coming weeks to compete for the CCA Increment 1 development and production contract. Anduril, Boeing, General Atomics Aeronautical Systems Inc., Lockheed Martin and Northrop Grumman are the finalists, each with a unit cost target of roughly one-quarter to one-third of a $92 million Lockheed F-35A. The winning aircraft are scheduled to be operational within 3.5 years.
Those developers will benefit from the second generation of AI-generated pilots that DARPA has in development. The agency is expected to announce several participants in its AI Reinforcements (AIR) program, a follow-on to the AlphaDogfight trials.
“For unpiloted platforms, AIR will enable vehicles to perform missions with minimal human oversight,” the agency’s latest budget documents say. “The outcome of this program will be an AI air combat capability that works in dynamic, operationally representative environments.”
One of the first moves of the industry players has been revealed. Alpha-Dogfight participant EpiSci has partnered with Northrop Grumman to provide AI-generated pilot algorithms for the defense prime’s advanced autonomous aircraft. EpiSci Chief Technology Officer Dan Javorsek also indicated the company’s involvement in the AIR program on LinkedIn, saying that its software would fly on the newly delivered VENOM F-16s.
—With Brian Everstine in Aurora, Colorado
Related Content
