Podcast: Applying Artificial Intelligence To Autonomous Aircraft

The U.S. military services are looking for a range of autonomous aircraft—from expensive high-end aircraft to less costly lower end systems. Aviation Week editors discuss what industry is offering, particularly for mid-range attritable systems, and some of the complications involved with cultivating the new technology. 

Don't miss a single episode. Subscribe to Aviation Week's Check 6 podcast in Apple Podcasts, Google Podcasts, Amazon, Audible and Spotify.

Discover all of Aviation Week Network's podcasts on our Apple Podcasts channel or aviationweek.com/podcasts.

Rush transcript

Jen DiMascio:

Hello, and welcome to the Check 6 podcast. I'm Jen DiMascio, the Executive Editor for Defense and Space, and I'm here with Defense Editor, Steve Trimble, and Pentagon Editor, Brian Everstine. They spent most of the last week at the Air Force Association's annual meeting, and in the week prior, Brian attended the Navy's Tailhook Convention of Naval Aviators.

Jen DiMascio:

The military is trying to figure out how to apply autonomy to aircraft, that is the airframes, the artificial intelligence algorithms, and the enablers of that technology. Steve, I was hoping you could lay out a bit about what the Air Force specifically is looking for, and what you saw at the show.

Steve Trimble:

Sure. So it's an interesting discussion, because the Air Force doesn't really give us a whole lot of detailed information about what it is they really want out of this whole new technology area of AI-piloted, uncrewed aircraft systems. So, we are getting some clues though, based on, really, what industry is showing off right now, and showing us what some of the options are.

Steve Trimble:

And when you think about the technology, as you mentioned, there's three different levels of it. There's the software itself. AI agents and the AI algorithms that will be used as the pilot on these aircraft, making certain decisions, being allowed to have certain degree of autonomy over how they fly and perform the mission. And then there's the platforms themselves. And then beneath that, there's the enablers, like cheap turbo fan engines, or new types of sensors that are also relatively cheap, or in some cases, even more powerful. So, the thing that we learned about probably the most during the show is on the platform side. As you walk around the exhibit hall at AFA last week, you got to see these categories of options that the Air Force could select.

Steve Trimble:

There's the exquisite end. You'd have some kind of flying wing UAS, or perhaps something very advanced that signature matched with the Next Generation Air Dominance fighter, the piloted version of the sixth generation fighter, so that it doesn't give away the position of the sixth generation fighter if they're flying in the same general airspace with each other. So there's that category. We saw the Northrop Grumman SG-101. Previously, Lockheed Martin had shown things like their next generation UAS. They're both sort of flying wing type things. Very generic, really, at this point, which is what you'd expect. But that's really the exquisite level. That's in the hundreds of millions of dollars each. So, like the MQ-25 for the US Navy carrier-based tanker, that's also uncrewed. That's 125 million dollars per aircraft. These would be in that category.

Steve Trimble:

And then at the very bottom end, there's the expendable category. And these are things that you can treat like munitions. Maybe they're recoverable, but you don't really need to recover them because they're priced at the same, an advanced precision guided missile, or a small turbofan powered cruise missile. And they have similar functions of that, but they can also work in a swarming way. So, this is things like Speed Racer. And my colleague, Brian, here wrote about what Skunk Works is planning to do with Project Carrera teaming up the Speed Racer-type decoy, or jammer, or surveillance platform, or could have a whole bunch of different platforms, and teaming that up with an F-35. And that's one thing is that, even though this concept is really being pushed to promote and expand the capability of Next Generation Air Dominance, that doesn't mean they can't work with other aircraft, so we're probably going to see that. That's one example of the options out there on the expendable side.

Steve Trimble:

But the big thing that we saw at the show was a lot more elaboration at the middle layer, what's traditionally called the attritable layer, which would be between this two to 20 million dollar price point, as least as how AFRL has defined it. And in that space, I mean, typically, we've heard things like, especially about Skyborg, the XQ-58 from Kratos. But really, what this show was, was a coming out party for the other alternatives to that. Things which are a bit bigger, and have a different way of operating than Valkyrie was designed to, with a Skyborg AI software agent. So, I'm mixing things up there. But anyway, sorry.

Steve Trimble:

Just with the Valkyrie, Kratos was the company that came out with the Valkyrie, but now they've come out with this new family of aircraft called Demigorgon, or Demigorgon, if I've got my Greek pronunciation correct. It's probably in the 10,000-12,000 pound class. Just like another family that General Atomics Aeronautical Systems, they also showed off this thing called Gambit. And these are very similar types of concepts, Demogorgon and Gambit, because they're based on a core hardware platform, which basically defines the wheel base, and the keel structure of the aircraft, power systems, probably the mission computer, and things like that. And then it can bolt on different types of wing configurations, fuselage configurations, engines and inlets, to be optimized, say in theory, for certain missions. The Gambit could be fielded first for adversary air, with one type of wing, one type of airframe, one type of engine, and inlet configuration. But then, if you wanted to do deep surveillance, you could take off this wing, take off this fuselage section, and bolt on more of a flying wing configuration, with a different inlet structure, and everything. So, that's the idea.

Steve Trimble:

It's a very automotive style, automotive inspired way of designing things. Traditionally hasn't been done in the aerospace industry, because aircraft tend to be very sensitive to center of gravity, center of lift, and weight and balance. And so, especially when you move from a straight wing to a swept wing, as they're proposing to do with these families, things get really complicated. They say they think they've been able to figure out how to manage that, so we'll see how that works out. It's a very ambitious thing to do in the aerospace industry.

Steve Trimble:

But we also saw some alternatives to that concept in the attritable class. So MQ-25. Sorry. Not the MQ-25, that's also another Boeing aircraft. But I'm talking about the MQ-28, the Ghost Bat, which is being developed in Australia by Boeing. Similar to the idea of Demogorgon and Gambit, but instead of replacing and modularizing the airframe, fuselage, engine, and inlets, it just does the radome, the nose. If you want it for a radar mission, you put a radar on there. If you want it for a jamming mission, you put a jammer on there. If you want it for a surveillance mission, totally different payload and nose. And so you're just changing that. That means it may not be optimized for all the different types of mission profiles that it would be tasked to do, but it's a lot simpler of a task to just change the radomes. Just like the U2 does, actually, for quite a few missions. So, that's really the idea.

Steve Trimble:

There was also some other concepts out there. Blue Force Technologies brought their concept out for the Fury, which is being built right now for the adversary air role. Which is really interesting, because that's on the software side, where the Air Force really wants to start getting their hands dirty, really, with this AI software. And exposing it to crewed aircraft, basically fighter pilots, in the same airspace, and have them work, in this case, in the adversary air role and beyond visual range type engagements. And then, as the AI agents learn more, and the fighter pilots learn more, and they learn to trust each other about what they can do, maybe that switches over to the blue role, instead of just being a red air asset. So, Fury's got the contract now to do the ADAIR-UX demonstration, but if that goes to the production stage, can expect some pretty intense competition from General Atomics with Demigorgon, and of course, with General Atomics... Did I say...? Well, Kratos with Demigorgon, General Atomics with Gambit.

Steve Trimble:

I think that covers everything. Sorry to ramble on there, but I think that covered the whole thing.

Jen DiMascio:

That's fascinating, Steve. Brian, how did that compare with what you learned out at the Tailhook convention? What is the Navy looking at, and what types of platforms?

Brian Everstine:

Yeah. The discussion's been really interesting on the Navy side, and there's a lot of surprising specifics as they're working through their future plans. Earlier this year, the CNO, Admiral Gilday, put out what he called his navigation plan, looking at what he sees as a future fleet. And within that, he threw out the number of 1,300 combat aircraft for the future carrier air wings. And of this, Navy officials say that they want 60% of that to be unmanned. Leaving about 40% to be a mix of F-35 and the Next Generation FA-XX, which is the manned fighter part of the Navy's NGAD, not to be confused with the Air Force's NGAD sixth generation development.

Brian Everstine:

And for the FA-XX, the Navy says that their manned fighter's in its concept refinement stage. And for this its version of the collaborative combat aircraft, the Navy threw out some specifics for it. They're looking, size-wise, they want it to be about half the size of an FA-18. Obviously, aboard ships, space comes at a premium, so they want to keep these pretty small.

Brian Everstine:

And they're looking at Boeing's MQ-25 as the trailblazer of how they want autonomy and uncrewed aircraft to operate at sea. Navy said that they are planning to use Lockheed's MD5 ground control station which will operate the MQ-25, to operate all uncrewed aircraft in the near future, with the same controller using that system to launch and recover other uncrewed aircraft. And while the Boeing's MQ-25 is, obviously, being pushed for a key mission of refueling and a little bit of ISR, it's broader role in the Navy is the trailblazer for this autonomy development. So far, the Navy said that it's had about 35 test flights with it's test asset, looking at IOC, and then, 2025 on the Roosevelt.

Brian Everstine:

For future CCAs, the Navy is looking at a rapid generation cycle of acquisition, calling back to what the Air Force looked at as a Digital Century Series. Looking at every couple years to iterate on new development. And looking at a lower cost compared to some of the exquisite systems that the Air Force had been looking at, using open mission systems, modular payloads we had touched on. For example, tapping in attaching an infrared search and track on one mission, taking it off, putting on a radar of another. Taking that off, put a jammer, depending on what the mission calls for.

Brian Everstine:

And one interesting discussion that Navy officials had during a panel at Tailhook, is that they're looking at their aircraft to not just be Navy for controlling. For example, aircraft could launch off a carrier, show up to an area and set up a combat air patrol and loiter for a while, and then eventually be handed off to an Air Force operator to handle, have the Air Force control them. Or even a joint military service, have the Aussies take over, depending on the location and the mission.

Brian Everstine:

So, those are some of the interesting specifics, but they're still looking at the requirements. It's going to be a little bit of a process. They're not really having much of a, really, schedule set yet.

Brian Everstine:

But they also already seem to be planning out the entire lifespan of these aircraft. That was an interesting discussion. Following any years of service at sea, they could bring these collaborative combat aircraft back to land, to send out to Fallon for Top Gun, where they could fly alongside Navy pilots in training, to work on the trust issue that pilots will probably have flying alongside an autonomous wingman. And then move over to the red air side for some exquisite training. And then eventually, finish their lives being shot out of the sky as a targeting drone. So, it's interesting that they had the whole lifespan of these things planned out before really their requirements are set. But they're also looking at pretty large numbers, as I'm going back to the 60% discussion.

Jen DiMascio:

Did they give a timeline for when they want to get to 60%?

Brian Everstine:

The Nav plan is in the 2035, 2040 timeframe, so it's still quite a ways away.

Jen DiMascio:

Interesting. So, I wanted to look ahead a couple weeks to October 10th, when the Army has its Association of the US Army Convention, here in Washington DC. What do you foresee that service looking at, in terms of autonomous aircraft?

Steve Trimble:

Well, I could try to pick up on that. So, the Army has talked about two different kinds of what they call air launched effects or ALEs. Very simply, described as small and large. No medium, at this point.

Steve Trimble:

At the large end, it's been the hardest to find out information. We know that there is a classified program that the Strategic Capabilities Office is working on, to support ALE Large. I think I've seen it, because they said it's done by L3Harris. A year ago, I was at one of their experiments called EDGE, at Dugway Proving Ground, and L3Harris was there. And they had a representative show me what they were working on for the Strategic Capabilities Office, and it was called the Red Wolf. Essentially, in that ALE Large class, with a 200 knot speed, and 300 kilometer range. It looks a little bit like a small mini cruise missile, maybe half the size of JASSM. That's something that could be carried by their future attack reconnaissance aircraft, or an Apache.

Steve Trimble:

The idea being that they would also have another class of aircraft, called ALE Small. And either of those families could go forward, be launched by their rotorcraft, and scout for targets, essentially hundreds of kilometers deep into enemy territory. So that they could provide that information back to their... The Navy has the MD5 ground station. The Army's is called Titan. And that would be the clearing house for these targets that would be picked up by them. And these new multi-domain sensing system, high altitude airborne thing, that the Army is trying to develop, as well. And use that. It's in that targeting information to their long range precision fires.

Steve Trimble:

They're going to be fielding all these ground launch missiles in the next couple years, including, at the very high end of the spectrum, it's the long range hypersonic weapon, or Dark Eagle, with a 1,700 mile range. Just below that you've got the mid-range capability, with initially either SM-6, or ground launch Tomahawk cruise missile. But that's going to be supplemented, or replaced, in several years, because the Army's talked about this new version of the Precision strike missile, which would have a combined cycle engine. So, you'd have a rocket booster, and then a ramjet augmenter. Not augmenter, but second stage, really, propulsion system, to extend the range out to about a thousand miles. So you've got that coming down the road.

Steve Trimble:

They've also got several other types of capabilities. GMLRS-ER, and ERCA, the extended range cannon artillery, maybe even the strategic long range canon artillery systems. I mean, they're pushing their striking power hundreds of miles, thousands of miles, deep into enemy territory, where previously they would rely on the US Air Force for that type of battlefield interdiction mission. They're taking it on themselves from the ground. So that's going to be really interesting.

Steve Trimble:

And the air launch effects (ALE) part of that is a key part of it. I don't know exactly what we'll see. Well, we'll be looking to understand more just what they're doing with ALE Small. They've been far more open and transparent about that level of their platform side, and see where they're going with that. Plus hopefully, we'll get some more information about ALE Large, which I think is going to be a pretty interesting program, that is really flying under the radar at this point due to its classification.

Jen DiMascio:

Thanks, Steve. That's a really complete rundown, and makes me look forward to what's coming out of the Army's conference in a couple weeks. Brian, did you have anything to add to that?

Brian Everstine:

I didn't really have anything specifics on the Army to add. But, Steve, you were out at EDGE earlier this year, and had some interesting lessons learned about how the Army's dealing with some bandwidth issues, and how that's shaping operations.

Steve Trimble:

No. I mean, it's a really good point. We're talking about this all like it's happening, and there's a certain schedule by the end of the decade. All these things will be proliferating. But really, what we're finding is they're just starting to learn what the limits are, what the capacity is.

Steve Trimble:

And when I went out to EDGE in 2021, which is this experiment the Army does at Dugway, they launched their ALE Smalls in mass. So, they had 30 of them just launched out. And the idea was they were going to create an ad hoc mesh network, and perform all these different missions amongst themselves. And they would decide, as they're performing this mission as a swarm, which one of them would be doing this part of it, the detection, which part of it would be doing the decoy role. And these would all be changing as the mission was performed. And maybe some of them were getting shot down, or just not working, and they would self-heal, and figure out the network. And they found out they just didn't have the communication bandwidth to make all that happen at the EDGE. They just couldn't send data fast enough to make that work.

Steve Trimble:

This past year, they had changed it. Instead of the Bee Swarm, as they called it, an approach they had done in 2021, they went to the Wolf Pack approach in 2022. In this case, they launched an initial salvo of, it was four to seven aircraft, all the salvos were within four to seven. And that first salvo just went to look for targets. And then the second salvo, once they found the targets, was sent out to be the decoy and light up the radars. And the third salvo of four to seven, however many there were, were launched to hit those targets, to actually be loitering munitions and suicide drones, and to hit those targets, those radiating targets. And then the fourth salvo was launched for battle damage assessment. So, they didn't have to coordinate all those missions among 20 to 30 some odd UAVs that were sent out, they just had to coordinate within their wolf pack, as they said, of four to seven aircraft. And that made it possible to do this.

Steve Trimble:

And they're going to be learning more and more about this. I think we're really still at the crawl stage of this whole thing. They're really trying to learn how to use autonomy, how to use artificial intelligence, how smart it really is, how intelligent it really is, how that intelligence interacts with how humans think about things. Because the artificial intelligence agents, they think in terms of probabilities, and calculate billions of these probability equations every second, and come up with different strategies than a human would use. And if they're being teamed up with humans, the human has to manage that, and understand that they're going to be thinking differently about how to approach something than a human would, based on our own of biological impulses, and all that. We don't have a lot of interaction yet to know what that relationship is going to look like. And over the next several years, we're going to learn all that.

Steve Trimble:

And as they learn those things, that's going to inform what kind of platforms they want. Whatever you hear about what they want to do right now is probably going to change significantly over the next several years, as they understand what these things actually are, how they operate, and how they operate best in this collaborative team.

Jen DiMascio:

That made me think of something that you mentioned, Brian, with the Navy and planning for a lifespan of its assets. But one of those would be building trust with humans, and how you trained for that.

Brian Everstine:

And in that vein, you have to build trust. But also, I had an interesting interview with the head of Air Force Futures, General Hinote, a bit ago, and he said you also need to build a healthy level of distrust. So, you need to understand that when these systems are operating well, and if maybe something's wrong, that the pilot can really understand and recognize that something is wrong. Because, he said that in a future conflict, the Air Force plans to attack the enemy systems any way they can, so they need to be prepared for the same thing to come back to us.

Brian Everstine:

And also, I thought it was pretty interesting, we talked about understanding the level of autonomy now. And the Air Force, specifically, there's been a lot of focus on the Skyborg program, and how that is about to set some of the building blocks for this. The Skyborg is very rudimentary. It's very basic. But that doesn't seem to phase the Air Force. General Hinote says they don't really need these things to be able to dog fight right now. We'll get there eventually, but we need to just start to bring in autonomy, and learn its lessons right away. To get these programs started sooner rather than later.

Steve Trimble:

And just on that AI side, the AI agent, probably timed it, to just say it, that Skyborg is not the only game in town anymore. There are AI agents proliferating all across industry. Every company seems to have their own AI agent that they want to stick on a platform. It's going to be really interesting, because I think we're going to see different AI agents that will go into the same aircraft. Right? I mean, one day it could be L3Harris' AI agent, and next day Shield Ais, and next day Raytheon's. And they'll have slightly different behaviors, because they'll be programmed slightly differently. So, they'll be just like humans, where they'll be slightly different, of different personalities, in a way. I mean, that's probably pushing way too far. They're not people. But it's just going to be a little bit different, and so we're going to have to see how we manage that.

Steve Trimble:

The Navy's been very quiet about what they've been doing. So, the Air Force has been very loud about what they're doing with Skyborg. It's sort of like, "Hey, look at us. We've got Skyborg. This is what we're doing." Navy's been doing that. I'm sure of it. They've just been a lot quieter about it. And you can look at a Raytheon press release from, I think it was September, October, 2021, where they talked about using BQM-34s which were painted up with US Navy target drone colors, because they were US Navy assets, being used for autonomy demonstrations. And AI agent pilots that were funded by the Strategic Capabilities Office, which is an OSD organization, that appears to be working with the Navy, in this case. So, I think the Navy has its own version of Skyborg, but neither that or Skyborg itself, is probably the thing that will go into service as the AI agent. It's just what they're practicing with right now, just to get some idea what this is, as industry is working on its own approach to the same thing.

Jen DiMascio:

Well, unfortunately, that's all we have time for today, but I think this is a discussion that we're going to keep revisiting over the next many years because the technology is evolving so rapidly.

Jen DiMascio:

Join us again next week for another episode of Check 6, which you can hear by subscribing to the podcast at Apple Podcasts, or wherever you listen to them. And one last request. If you're listening to us on Apple Podcasts and want to support us, please leave us a very high star rating, or review. Bye for now.