AFRL to Test Skyborg Autonomy Algorithms This Summer


A Skyborg conceptual design for a low cost attritable Unmanned Combat Aerial Vehicle. Air Force Research Laboratory artwork.

The Air Force’s unmanned Skyborg project will get its first flight test this summer, putting autonomous flight controls to work on a “small, but representative, high-speed surrogate aircraft.” The ultimate goal is to create a combat-ready, autonomous aircraft comparable to a fighter jet by the end of 2023.

Skyborg is envisioned as a robotic wingman for other pilots, using artificial intelligence to fly and control the aircraft and managing some combat mission tasks itself. After initial test flights, Skyborg will scale up to larger test platforms that are closer to the final airframe, according to Air Force spokeswoman Capt. Cara Bousie.

Bousie declined to specify when this summer’s testing will occur at Edwards AFB, Calif., but said it is part of a two-year experimentation campaign designed to see how advanced autonomy performs in a controlled setting. The tests get at a key question at the heart of AI and autonomy research: How can the military ensure the smart systems it develops are “thinking” correctly

Upcoming flights will focus on how Air Force Research Laboratory-built software can test that an autonomous system is proposing reasonable next steps. The flights will also look at how to correct the system with a better solution if it misses its mark. Just as a human pilot tells their plane what to do in combat, the algorithms will suggest how the Skyborg aircraft can proceed.

“The autonomy algorithms can be viewed separately from the vehicle system, similar to how a pilot is viewed separately from the aircraft,” Bousie said.

Installing more advanced autonomy is the experimentation campaign’s next step to declaring Skyborg ready for operations. Its development can also spill over into other research areas, like a special AFRL AI team known as ACT3 that is working on air-to-air combat algorithms for existing fighter jets.

“The ACT3 effort aims to develop an artificial-intelligence capability to execute air-to-air tactics,” Bousie said. “In this way, ACT3 and Skyborg are complementary, but separate efforts. The Skyborg project plans to leverage the work from ACT3 and other applicable technologies” that have been judged useful to the military and are mature enough to be operational in the next few years.

ACT3 is still deciding what hardware and software it needs to launch full-scale development for its fighter project. The Air Force is working on other “wingman” initiatives like Kratos’ XQ-58A Valkyrie.

“When I was in [the Office of the Secretary of Defense], I thought there was a powerful role to give a high-end fighter a wingman that you could take risk with, that you could attrit, that would not necessarily have to return and land,” Air Force Acquisition Executive Will Roper told lawmakers in May.

The question of how exactly the Air Force will use airframes that can pair with other jets, crunch data themselves, potentially select targets, and more—and what role humans will play in that world—is still under consideration. Even if the AI is programmed to “think” within certain guidelines, there is more work to do for humans to trust these sidekicks. Skyborg is helping drive those discussions, Bousie said.

One main effort is bringing together experts from across the defense community to ensure these development initiatives are following current policy guidelines, while also considering what language might need to change to fully take advantage of AI and technology that helps machines learn from their experiences, she said.

A 2012 Defense Department directive dictates that warfighters must be able to “exercise appropriate levels of human judgment” over autonomous and semi-autonomous weapon systems, though there are still many unknowns about what the military may deem “appropriate” and in what contexts.

“Incorporating collaborative, teamed, autonomous aircraft into the Air Force will require a myriad of policy and regulation changes, similar to the challenges private industry is attempting to overcome with autonomous vehicles,” Bousie said. “Some of these challenges include airspace integration, supervision, collaborative communication, cybersecurity, resiliency, mission planning, accountability, and many others. While Skyborg is not attempting to solve all of these challenges, the project provides a catalyst for the necessary changes to occur.”