AI in the sky

Air power does not come cheap. The F-35 was supposed to save the US money by replacing a range of different fighters. It will cost the country an estimated $1.7trn through its operational lifetime – making it the most expensive weapon system ever built and an easy target for politicians, stealth design notwithstanding. In March 2021, Washington state representative Adam Smith, the top Democrat on the House Armed Services Committee, compared the programme to “throwing money down [a] rathole”, which wasn’t entirely fair. The F-35 may not be as agile as some of its predecessors, but none of those aircraft can match its ability to penetrate the modern air defence systems used by Russia and China.

Admittedly, that would be more useful if F-35s were not so prone to getting stuck in hangars waiting for repairs or parts. On average, only 61% of the US fleet was available at any one time in 2021, well below the military’s desired 80% mission capable rate. “Is there a way to not keep spending that much money for such a low capability?” Smith asked.

Made to be expendable

Well, yes. For all the doom mongering about the fact that, close up and one-on-one, the 50-year-old F-16 can outmanoeuvre the F-35, the latter aircraft neither has to get close nor fight alone. Just days before Smith took aim at the “rathole”, the Royal Australian Air Force tested its first Boeing Loyal Wingman, later rebranded as the MQ-28 Ghost Bat – a relatively cheap unmanned combat aerial vehicle (UCAV) designed to work with and support its F-35s. It’s a match made in heaven – or at least optimised for contested skies. As Lockheed Martin puts it, pilots of fourth-generation fighters like its F-16 are sensor operators consumed with the job of maintaining situational awareness; F-35 pilots, equipped with a spectacular – and heavy – array of fused sensors and avionics systems, are battlespace commanders. Their first troops are now taking off to join them.

“I think the technologies are there now where we can talk about a formation, if you will, of a manned aircraft controlling multiple unmanned aircraft,” said US Air Force Secretary Frank Kendall in January 2022. Networked with sufficiently advanced fighters and bombers, these drones can act as forward sensors, swarming decoys or weapons platforms while protecting pilots and their high-value planes. To Mark Gunzinger and Lukas Autenried at The Mitchell Institute for Aerospace Studies, as is made clear in their paper “Understanding the Promise of Skyborg and Low-Cost Attritable Unmanned Aerial Vehicles”, they are “a new class of force multipliers that could help the air force balance its requirements and modestly grow its force capacity”. Or, as Kendall – with the F-35 front of mind – put it, “If we only do very expensive aircraft for the air force, we’re not going to be able to afford an air force anywhere near the size that we either need or have today. These have got to be cheaper platforms.”

That is exactly what they are. The USAF calls them attritable/reusable (A/R) UCAVs, reflecting the fact that they are cheap enough to risk losing (an XQ-58 Valkyrie could cost as little as $2–3m, versus $78m for the cheapest F-35A) and that they can complete their missions without destroying themselves – unlike, for example, cruise missiles or loitering munitions drones. “That opens up a whole range of tactics that we currently would not contemplate because we’d be sacrificing manned aircraft,” Kendall added.

A/R UCAVs are also designed to be modular, meaning they can swap out weapons and sensors to perform different offensive, defensive, support and surveillance roles as required. As then RAAF head of air force capability Air Vice-Marshal Cath Roberts said after two Australian MQ-28 Ghost Bat drones flew in formation in November 2021, “This opens up significant capability agility for the air force, particularly with features such as the reconfigurable nose. We’re heavily engaged in the payload development and the element of surprise that it gives us in the battle space. You never really know what’s in the nose.”

That’s a big change from the development model that brought about multirole platforms like the F-35, or any of the mission-specific aircraft that preceded it. Efficient as it might be to rationalise supply chains around fewer types of standard aircraft, Gunzinger and Autenried argue that increasingly homogeneous allied air forces are easier for adversaries to predict and counter. With the US’s combat air force now smaller than it ever has been, that contributes to a reduced ability “to pose multiple simultaneous operational challenges to an enemy”. In that context, “A/R UAVs could help create a future force that is more heterogeneous, less predictable and more capable of distributed operations.”

Increased capability

At the same time, loyal wingman technology plays directly into the advantages offered by manned aircraft like F-35s. Gunzinger and Autenried highlight that fifth-generation fighters are already force multipliers “in the sense that they can share their operational picture of the battlespace with other air, ground and sea weapon systems and perform as battle managers in contested areas”. As loyal wingmen do not have to stay within shouting distance, a single battle manager can use them to extend that operational picture over a much greater range.

For example, A/R UCAVs equipped with various sensors can be spread out across a wide area while transmitting data to F-35s with the tools on board to fuse their readings. From there, pilots can command their robotic wingmen to track or target specific threats, either using the data to engage enemies from a safe distance or passing it on to other shooters that can also stay out of range without impacting accuracy. Gunzinger and Autenried are clear about how important this could be for the US military, which has long relied on its ability to dominate the skies. “Extending the sensor and weapons reach of counterair aircraft in this manner would help ensure the USAF maintains its first look, first shot, first kill air superiority advantage.”

In fact, introducing A/R UCAVs can help address some of the vulnerabilities of linear kill chains, which can be broken if only one link – an AWACS aircraft, for instance – is lost. By contrast, fifth-generation fighter and A/R UCAV-supported ‘kill meshes’– consisting of potentially hundreds of sensors, shooters, and command and control nodes – could continue to function even if multiple elements are jammed or shot down, increasing the resilience of offensive operations. As such, relatively few manned fifth-generation penetrators should be able to quickly find and engage high-value targets while minimising their exposure to enemy fire. Even if they’re not providing sensor data for the kill mesh themselves, A/R UCAVs could disrupt air defence systems with electronic weapons, confuse the enemy’s operational picture by acting as decoys, or enable US commanders to conduct simultaneous operations that overwhelm the enemy’s ability to react.

In that last example, A/R UCAVs could be used as off-board stations carrying extra air-to-air weapons for manned aircraft. Autenried and Gunzinger point out that that would address “a capability gap the air force has long sought to fill”, particularly as fifth-generation aircraft can only carry four to six missiles while maintaining their low radar profile. Accompanied by loyal wingmen with air-to-air weapons, manned aircraft could carry more air-to-ground munitions, enabling them to deliver more strikes per sortie without increasing their vulnerability to enemy interceptors. As Gunzinger and Autenried note, that “can have a critical impact on the time needed to achieve a theatre commander’s objectives”.

Trust the system

The potential of loyal wingman programmes is undeniable, but, to realise it, pilots and AIs need to work out how to fly together. Typically, they are presented as competitors. DARPA’s Air Combat Evolution (ACE) initiative made headlines when it showed that an AI system can beat a USAF F-16 pilot in simulated dogfights – by 5–0, in fact. That does not mean the end of humans in the cockpit, though. Dogfighting in clear, virtual skies is a relatively easy task for an AI, but it’s a long path from there to full battlefield autonomy. As the above examples illustrate, loyal wingman drones are not designed to match human pilots’ ability to make decisions under pressure, but to perform specific tasks under their direction. Testing each other’s flying skills is just how pilots like to get acquainted.

“Regardless of whether the human or machine wins the final dogfight, the AlphaDogfight Trials is all about increasing trust in AI,” Colonel Dan Javorsek, ACE programme manager, told journalists before the 2020 event. “If the champion AI earns the respect of an F-16 pilot, we’ll have come one step closer to achieving effective human-machine teaming in air combat, which is the goal of the ACE programme.”

That’s still the vision today. “They operate as a team […] and they’re under the control of the manned aircraft,” said USAF secretary Kendall in January 2022. “The commander for that formation calls the play; then, with a reliance on a high degree of autonomy, the other members of the formation operate in a way which is optimised against the threat that they see.”

We are on the cusp of seeing that in action. The US Air Force Research Lab’s (AFRL) Skyborg programme has made important strides towards developing a platform-agnostic Autonomy Core System (ACS) that enables a variety of drones to fly in coordination with and support crewed aircraft. In October 2021, two AFRL MQ-20 Avengers flew in coordination using the ACS. That puts the US slightly ahead of Australia in the race to operationalise loyal wingmen, but by no means are they the only countries to realise the potential of A/R UCAVs. India, China and Russia are also testing their own drones and teaming systems. A money-saving measure is becoming an arms race.