Saab's Gripen Tests AI In Long-Range Air-To-Air Engagements

Saab has begun test flights of its Gripen E fighter loaded with an artificial intelligence (AI) agent that can conduct autonomous beyond-visual-range air-to-air combat. The AI agent, known as Centaur, can be introduced to the Gripen E without any major modifications, thanks to its open-architecture avionics. Under a program funded by the Swedish government, Centaur could find its way into Swedish Air Force Gripen Es as well as other crewed and uncrewed platforms being developed by Saab, which you can read more about here.

The first three test flights of a Centaur-equipped Gripen E were announced today by Saab, which is working on the program together with Germany’s Helsing, responsible for the AI agent. The flights were conducted as part of Saab’s wider Project Beyond, which aims to advance AI capabilities in military aircraft. This project is funded by the Defense Materiel Administration (FMV), Sweden’s defense procurement organization, as part of the country’s future fighter program.

The first of these flights took place on May 28. By the time of the third sortie, on June 3, Centaur was ready to be pitted in beyond-visual-range (BVR) aerial combat against a crewed Gripen D fighter. During a series of dynamic BVR scenarios, targeting data was acquired by the Gripen E’s sensors, and Centaur autonomously executed complex maneuvers on behalf of the test pilot. These culminated in the AI agent providing the pilot with firing cues for the (simulated) air-to-air weapons.

Both the Gripen E and D were supported by an airborne early warning and control (AEW&C) asset providing additional targeting data and command-and-control functions.

Different engagement scenarios explored how Centaur functioned at different starting distances from the target aircraft, as well as various speeds and aspects. Some scenarios also saw command-and-control data disabled to assess the robustness of the AI agent.

“We have been developing human-machine collaboration for a long time,” explained Johan Segertoft, chief of Saab’s Gripen business unit, “allowing the Gripen E to take a lot of decisions on its own. The pilot is still the commander of the aircraft, and will continue to be, but with the assistance of, for example, a BVR agent, you are opening up a lot of free brain power in that commander to do other things.”

Meanwhile, Marcus Wandt, Chief Innovation Officer at Saab and a test pilot himself, said that the test flights so far point to the fact that “it is not a given” that a pilot will be able to win in aerial combat against an AI-supported opponent. “There are still pilots out there that will have a chance, but that will change fast,” he contended.

“This is an important achievement for Saab, demonstrating our qualitative edge in sophisticated technologies by making AI deliver in the air,” said Peter Nilsson, head of Advanced Programs within Saab’s Aeronautics business area. “The swift integration and successful flight testing of Helsing’s AI in a Gripen E exemplifies the accelerated capability gain you can get from our fighter. We are excited to continue developing and refining how this and other AI agents can be used, while once again showing how our fighters will outperform faster than the opponent can evolve.”

In a reflection of the extent of the advantage that an AI solution like Centaur can provide in air combat scenarios, Antoine Bordes, VP of Artificial Intelligence at Helsing, estimated that, in the third test flight, the agent “flew for around 50 years of the equivalent pilot years,” but noted that “it took us only a few hours to do that.” All of this additional ‘flight time’ was generated by Centaur in the virtual world. In the process, Centaur very rapidly built up its experience and its proficiency in terms of decision-making for BVR combat — an arena of combat that Saab describes as “like playing chess in a supersonic with advanced missiles.”

“The fact that we can change and update the [AI] model is one of the key successes” behind Centaur, Bordes added.

As well as the speed with which an AI agent can learn how to make decisions in air-to-air combat, the project has demonstrated how quickly the Gripen E’s software can be updated.

The pre-flight testing work for Centaur started only six months ago, with the AI agent initially trained and tested using Gripen simulation data.

Overall, the Centaur program aims to “explore how trustworthy AI can be used against future threats,” but the relative ease with which it can be inserted into the Gripen E could make a productionized version attractive for the Swedish Air Force or for export operators of the same aircraft (as well as the two-seat Gripen F).

As it stands, the next phase of testing under Project Beyond will see a further series of flights throughout the remainder of the year. Trials after that might involve two AI-enabled Gripens battling two crewed adversary aircraft, although this will be played out in a simulated environment, according to Saab’s Peter Nilsson.

Saab also confirmed that there is scope to expand Centaur into within-visual-range (WVR) aerial combat. BVR is the initial focus, the company says, since this remains the most critical aspect of air-to-air engagements, something that has also been demonstrated in the air war in Ukraine. It also has a particular relevance for the Swedish Air Force, which has long sought to develop innovative technologies and tactics that would allow it, as a much smaller air arm, to deal with a potential large-scale Russian aerial attack. In such a scenario, it is easy to envisage how AI could play a critical role in helping a force of Gripens armed with Meteor BVR air-to-air missiles, for example, prioritize multiple incoming threats and find the best solutions to combat them.

Even if Centaur doesn’t fund its way into frontline Gripens, the lessons from the program will be fed into Sweden’s future fighter program, which aims to select a next-generation air combat platform or platforms by 2031.

In the meantime, there are interesting parallels between the Centaur program and U.S. efforts to explore AI software in fighters. Saab, however, makes a point of the fact that the AI agent can be fully integrated in the Gripen E “without being restricted to solely military test ranges or having to rely on an experimental X-plane to do flight trials with the software.”

In contrast, a similar U.S. Air Force program has made use of an AI-controlled F-16 experimental jet, the X-62A VISTA, which you can read more about here. In 2023, the X-62A was flown in fully autonomous mode during a dogfight against a crewed F-16. For this test flight, a pilot was in the X-62A’s cockpit as a failsafe.

The U.S. Air Force has also been using other platforms to explore advanced autonomy and is converting more F-16s into test jets to support larger-scale collaborative autonomy testing. This is part of a program called Project VENOM (Viper Experimentation and Next-Gen Operations Mode).

Projects like these will feed into the U.S. Air Force’s Collaborative Combat Aircraft (CCA) drone program. This ultimately plans to field hundreds, if not thousands, of lower-cost drones with high degrees of autonomy. Indeed, this autonomy is fundamental to the way the uncrewed aircraft will operate very closely with crewed types, including the F-47 crewed stealth jet being developed under the Next Generation Air Dominance (NGAD) initiative.

The technology being developed under these American efforts will potentially also feed into many other programs across the U.S. military, where there is growing interest in new AI and machine learning-enabled autonomous capabilities in general.

With the recent Centaur trials under Project Beyond, Saab and Helsing have also joined a growing field of industry and defense players who are exploring bringing AI capabilities to military aircraft. While these technologies could assist Gripen E pilots in going into battle in the future, they could have an even more significant impact on Sweden’s next generation of crewed and uncrewed combat aircraft.

Contact the author: thomas@thewarzone.com