USAF/US NAVY 6G Fighter Programs - F/A-XX, F-X, NGAD, PCA, ASFS news

[NMaude]

Principal Deputy Undersecretary of the Air Force for Acquisition Darleen Druyun

She ended up having a "Nice" government paid "Vacation" at "Club Fed".

 

[Explorer121]

Not necessarily. If you're referring to the twin seat J-20B, so far we don't know if it will enter service (the more recent consensus is it's a series of tech demos that assists in the development of the new J-20A variant).
Even if J-20B does enter service, we don't know how its CCA/UCAV control is in an absolute sense, only that it would likely be superior to a single seater.

For example (just throwing a number out there), a generic single seater of the near future may be able to control six CCAs while carrying out its regular missions, while a generic twin seater of the near future may be able to control twenty CCAs while carrying out its regular missions, and having both single seater and twin seaters in service could prove useful (regular CCA control versus augmented uber CCA control).

4-6 CCAs might give you air dominance over the modern battlefield, but when competitors begin fielding CCAs in a decade or two, you'll want to add more airframes and increase specialization to maintain the advantage. Eventually at some arbitrary size, managing a swarm of combat drones will be a full-time job unless you want to pass all mission authority to the AI.

 

[Ozair]

4-6 CCAs might give you air dominance over the modern battlefield, but when competitors begin fielding CCAs in a decade or two, you'll want to add more airframes and increase specialization to maintain the advantage. Eventually at some arbitrary size, managing a swarm of combat drones will be a full-time job unless you want to pass all mission authority to the AI.

I'd be very surprised if AI advances didn't allow CCA parent drones managing child drones on behalf of the manned platform. The manned platform still manages overall strategy and intent with the parent CCAs having devolved authority for specific actions. The slowest part of this whole process is the manned portion trusting the AI to operate autonomously. I expect a conflict or two will either remove those barriers or enforce them and given the pace of the sector I'd say removing barriers seems more likely.

 

[B2Blain]

I'd be very surprised if AI advances didn't allow CCA parent drones managing child drones on behalf of the manned platform. The manned platform still manages overall strategy and intent with the parent CCAs having devolved authority for specific actions. The slowest part of this whole process is the manned portion trusting the AI to operate autonomously. I expect a conflict or two will either remove those barriers or enforce them and given the pace of the sector I'd say removing barriers seems more likely.

Whatever the case, the USAF needs to manage risk of relying on unmanned combat aircraft for a new role and function. It didn't do this when the US didn't develop the F-4 with a canon, believing that a simpler autonomous weapon than CCAs, the air to air missile, had made guns as the primary air to air weapon on fighters obsolete.

AF leadership seems to be engaging in group think. Everyone has been sold the idea of affordable mass, but like other forms of technology the reality may be different than the promise.

 

[Ozair]

Whatever the case, the USAF needs to manage risk of relying on unmanned combat aircraft for a new role and function. It didn't do this when the US didn't develop the F-4 with a canon, believing that a simpler autonomous weapon than CCAs, the air to air missile, had made guns as the primary air to air weapon on fighters obsolete.

For sure the technology is new but if they don't push it now then they will likely either obtain parity or fall behind. The US leads AI research and more than any other country in the world has the ICT startup culture to develop and push the boundaries of what is possible with AI. Fortunately for the US Military there are multiple startups such as Andruil who are committed to the defence cause.

As an aside I was always curious if manned NGAD, which seems less and less likely now, would also be equipped with a cannon. Noting F-35B/C lost theirs would that be present on NGAD, F/A-XX, FCAS, GCAP etc.

AF leadership seems to be engaging in group think. Everyone has been sold the idea of affordable mass, but like other forms of technology the reality may be different than the promise.

Hence why phase one and phase two etc, incremental steps forward. If they build it right then the underlying platform hardware acquired, so airframe and engine and cooling and generic interfaces can remain relatively static, but the systems can be upgraded with new software and sensors.

 

[Scott Kenny]

As an aside I was always curious if manned NGAD, which seems less and less likely now, would also be equipped with a cannon. Noting F-35B/C lost theirs would that be present on NGAD, F/A-XX, FCAS, GCAP etc.

Once more nations get most of their Air Force equipped with stealthy aircraft, then yes the gun is still going to be needed.

In Vietnam, the problem assumed in the Phantom was that most combat would be BVR. Which would have been accurate in the event of a US v USSR fight. But when you have USN, USAF, USA, RAAF, ROKAF, ROKA, RVNAF, ARVN etc all flying overhead and you don't have reliable IFF systems, you need to get positive visual ID to avoid fratricide. And now you're close enough to need guns.

When only one side in a conflict has LO aircraft, they control the range at which the fights happen. And it'll be overwhelmingly BVR. Once both sides in the fight have LO aircraft, we're looking at WVR (or pretty close to that using IRSTs). So you're going to need guns again.

 

[Ozair]

Once more nations get most of their Air Force equipped with stealthy aircraft, then yes the gun is still going to be needed.

I'm not sure the presence of stealth aircraft enforce use of a gun. Reasoning expanded below.

.In Vietnam, the problem assumed in the Phantom was that most combat would be BVR. Which would have been accurate in the event of a US v USSR fight. But when you have USN, USAF, USA, RAAF, ROKAF, ROKA, RVNAF, ARVN etc all flying overhead and you don't have reliable IFF systems, you need to get positive visual ID to avoid fratricide. And now you're close enough to need guns.

The ROE constrained the US forces in Vietnam, some of that was perhaps limited IFF but I'm not sure it can all be attributed to that. It could well be argued the very poor performance of the AIM-4 was as much an issue as the ROE constraints,

Putting some number to it, US gun kills were approx 20% of A2A kills over Vietnam and a majority of those were F-105s with no AAM on the aircraft so no other option. Hence I'm not sure the presence of a gun would have significantly changed the kill ratio. Having a better performing AAM over the AIM-4 likely would have made a difference, at least in the early years.

When only one side in a conflict has LO aircraft, they control the range at which the fights happen. And it'll be overwhelmingly BVR. Once both sides in the fight have LO aircraft, we're looking at WVR (or pretty close to that using IRSTs). So you're going to need guns again.

If you are close enough to see the other aircraft then I'm not sure a gun is required. HOBS missiles are good enough now that as long as the target is detectable then a missile seems far more effective than a gun which, with obviously the absence of a turret, is limited to nose pointed engagements. Maybe min fuzing distances for the missiles might be a factor but I doubt it. The presence of a gun would IMO be more a factor in A2G tasking.

To conceive a scenario, a manned NGAD is shot down behind the FLOT and while supporting a CSAR the NGAD wingman needs an appropriate sized weapon to keep adversary infantry away from the downed aircrew's position. A perfect use of a gun. Of course that is a contrived scenario and I'm not anti gun, just not sure the USAF will go there.

Taking it a step further then, will we see a gun equipped CCA? Seems ideal for interdiction tasks against soft targets as well a CAS and potentially then also A2A against other drones. They are probably out there but I haven't seen any concepts that have a gun as the primary or even secondary weapon on a CCA.

 

[TomcatViP]

But can a CCA be really relevant in a dogfight? I mean, you need a lot of engine power and CCA while, being built for range and persistence, doesn´t look like performers there.
So why someone would want them to have a gun for DCA?
Similarly, what kind of profile attack would a CCA use while strafing? Dive and pull-up at 5G? I am not sure they can do that efficiently and have the power to not linger at low alt within ground defense reach.

 

[Ozair]

But can a CCA be really relevant in a dogfight? I mean, you need a lot of engine power and CCA while, being built for range and persistence, doesn´t look like performers there.
So why someone would want them to have a gun for DCA?
Similarly, what kind of profile attack would a CCA use while strafing? Dive and pull-up at 5G? I am not sure they can do that efficiently and have the power to not linger at low alt within ground defense reach.

All valid points. Phase One doesn't appear to have a gun. The GA aircraft doesn't appear overly manoeuvrable, the Andruil Fury may be a bit better but the control surfaces don't appear large either.

I expect that in CAS scenarios then perhaps the attritable portion of the CCA comes in to play. You likely wouldn't use the aircraft for gun runs until it had exhausted its other munitions. Alternatively maybe we will see CCA's custom designed for CAS with gun armaments that trade the internal weapons for higher thrust, perhaps armour in certain locations and better manoeuvrability?

I'm again not saying this will happen, just a thought exercise given we tend to focus on missiles and smart bombs as the primary weapons.

 

[DWG]

As an aside I was always curious if manned NGAD, which seems less and less likely now, would also be equipped with a cannon. Noting F-35B/C lost theirs would that be present on NGAD, F/A-XX, FCAS, GCAP etc.

Given the air-combat AI being tested by USAF reputedly developed a liking for head-on gun shots during the merge, I wouldn't rule out a gun from any future combat platform, manned or unmanned.

 

[siegecrossbow]

Once more nations get most of their Air Force equipped with stealthy aircraft, then yes the gun is still going to be needed.

In Vietnam, the problem assumed in the Phantom was that most combat would be BVR. Which would have been accurate in the event of a US v USSR fight. But when you have USN, USAF, USA, RAAF, ROKAF, ROKA, RVNAF, ARVN etc all flying overhead and you don't have reliable IFF systems, you need to get positive visual ID to avoid fratricide. And now you're close enough to need guns.

When only one side in a conflict has LO aircraft, they control the range at which the fights happen. And it'll be overwhelmingly BVR. Once both sides in the fight have LO aircraft, we're looking at WVR (or pretty close to that using IRSTs). So you're going to need guns again.

Combat between LO or even VLO aircraft take at longer ranges than most people assume. While stealth fighters may not be able to detect and engage each other using onboard radar, that’s not really the case if they can get info from AWACs.

 

[TomcatViP]

Given the air-combat AI being tested by USAF reputedly developed a liking for head-on gun shots during the merge, I wouldn't rule out a gun from any future combat platform, manned or unmanned.

That within the constraint of demo testing where adversary actions are summarized and filtered through assumptions. A do X hence me do Y.
Here the drone probably interpret their low frontal aspect and the linear flight path from the target as an advantage when a pilot might exploit that to decipher the merge picture and ruin the day of the robot (but do they have days? )

 

[Shoes31]

I'd be very surprised if AI advances didn't allow CCA parent drones managing child drones on behalf of the manned platform. The manned platform still manages overall strategy and intent with the parent CCAs having devolved authority for specific actions. The slowest part of this whole process is the manned portion trusting the AI to operate autonomously. I expect a conflict or two will either remove those barriers or enforce them and given the pace of the sector I'd say removing barriers seems more likely.

I'd say the slowest part of this whole process is the development of the AI command logic - not the manned side. Its seriously hard to make good AI, something that seems to get missed whenever we talk about CCA's.

If all the autonomy you want out of a CCA is a simple fly here while avoiding the A2A bubble, bomb this target, fly back thats easy enough to do but requires a greater workload for the manned element. If you want less workload on the manned fighter, you need software that all you need to do is tell it to suppress enemy defenses here, fly out to this area, and then select best targets of opportunity for attack. It has to be able to defend itself with no intervention. For the air to air side, the software has to be able to designate enemies, figure out which CCA will attack which target, best plan of intercept etc etc and suggest all that to the pilot or be able to take action by itself. That is hard to do, and I don't see that software being ready in the next few years at all. Not with any sort of reliability that I'd want to base my NGAD design on the assumption that CCA's will do a large share of the job.

 

[DWG]

That within the constraint of demo testing where adversary actions are summarized and filtered through assumptions. A do X hence me do Y.
Here the drone probably interpret their low frontal aspect and the linear flight path from the target as an advantage when a pilot might exploit that to decipher the merge picture and ruin the day of the robot (but do they have days? )

Can't remember where I saw it, maybe a thread here, but ISTR the AI was a lot more willing to commit to head-on, and more accurate when doing so.

I suspect it's a good way to FOD your own engines even if it works, but a CCA is probably a good exchange for an NGAD. (The whole air-combat kamikaze option opens up bigtime if there's not actually anyone aboard and the target is worth it...).

 

[TSARb]

Guys, what are the goals of a sixth-generation fighter? What should it be capable of? I mean, 5th gen is already great, and wingmans (wingmen?) expand capabilities even further.

 

[dark sidius]

Guys, what are the goals of a sixth-generation fighter? What should it be capable of? I mean, 5th gen is already great, and wingmans (wingmen?) expand capabilities even further.

Be in front of the enemy this is the reason you can't let China having 6th gen capacity before USA. F-22 start to be a 15 years plane with avionic the same age and F-35 is not a air dominance fighter.

 

[Josh_TN]

I'd be very surprised if AI advances didn't allow CCA parent drones managing child drones on behalf of the manned platform. The manned platform still manages overall strategy and intent with the parent CCAs having devolved authority for specific actions. The slowest part of this whole process is the manned portion trusting the AI to operate autonomously. I expect a conflict or two will either remove those barriers or enforce them and given the pace of the sector I'd say removing barriers seems more likely.

I think the U.S. CCAs rather rapidly will evolve to adopt specific behaviors as specified by the pilot: for instance, perhaps a pilot sets one drone to defensive wingman and two others to offensive sweep. The wingman only concerns itself with acting as a radio relay to minimize the manned aircraft’s exposure and synchronizes its flying to put itself in between the fighter and any incoming threat as a stand in jammer, CM dispenser, and if necessary sacrificial decoy. The sweepers move ahead and act as forward sensor platforms with an optional engagement capability if the pilot approves. All of the aircraft reorient themselves with the principle’s movement so the pilot does not have to give new instructions, rather like carrier escorts responding to course changes of CV to launch and recover aircraft and such. This would minimize the heads down time.

Long term I think the controller is ground based and the orders dispensed via satellite. The U.S. is going to have a global low latency system of multiple tactical radio formats inside several years, including laser cross links and downlinks as well basic Link16 and a host of UHF formats. Dedicated high altitude communications gateway UAVs could potentially directly receive optical communications as well as UHF and L bands. I think this combined with higher levels of automation rapidly makes the control of drones via aircrew obsolete, or at least more of a fall back option for redundancy.

 

[Jimmo952]

Once more nations get most of their Air Force equipped with stealthy aircraft, then yes the gun is still going to be needed.

In Vietnam, the problem assumed in the Phantom was that most combat would be BVR. Which would have been accurate in the event of a US v USSR fight. But when you have USN, USAF, USA, RAAF, ROKAF, ROKA, RVNAF, ARVN etc all flying overhead and you don't have reliable IFF systems, you need to get positive visual ID to avoid fratricide. And now you're close enough to need guns.

When only one side in a conflict has LO aircraft, they control the range at which the fights happen. And it'll be overwhelmingly BVR. Once both sides in the fight have LO aircraft, we're looking at WVR (or pretty close to that using IRSTs). So you're going to need guns again.

In Vietnam, another problem was that the air-to-air missiles available at the time were just not the good. There were serious performance and reliability issues.

The promises of the manufactures exceeded the capabilities delivered.

 

[TomcatViP]

Can't remember where I saw it, maybe a thread here, but ISTR the AI was a lot more willing to commit to head-on, and more accurate when doing so.

I suspect it's a good way to FOD your own engines even if it works, but a CCA is probably a good exchange for an NGAD. (The whole air-combat kamikaze option opens up bigtime if there's not actually anyone aboard and the target is worth it...).

Your memory is correct. I did also read the same.
But as was emphasized at the time by the DARPA, the dogfight envelope chosen for the test scenario of ACE was more for practicality and rapid assessment.
If they had gone BVR, CCA demonstrators would have been way more complex, costly and range allocation more scarce, delaying the assessment and development of the capacity.
Yet, no particular airframe have been revealed with the accent put on WVR*. Their dedicated payload being even AMRAAM missiles.

*Although that might be where GA-SI and Anduril slightly differs

 

[Josh_TN]

Once more nations get most of their Air Force equipped with stealthy aircraft, then yes the gun is still going to be needed.

In Vietnam, the problem assumed in the Phantom was that most combat would be BVR. Which would have been accurate in the event of a US v USSR fight. But when you have USN, USAF, USA, RAAF, ROKAF, ROKA, RVNAF, ARVN etc all flying overhead and you don't have reliable IFF systems, you need to get positive visual ID to avoid fratricide. And now you're close enough to need guns.

When only one side in a conflict has LO aircraft, they control the range at which the fights happen. And it'll be overwhelmingly BVR. Once both sides in the fight have LO aircraft, we're looking at WVR (or pretty close to that using IRSTs). So you're going to need guns again.

Gunfire is still going to be hopeless hard to achieve versus firing a 60g HOBs IIR guided AAM at even a couple miles/kilometers range. It a world that will be filled with wrap around IRST and lock on after launch missiles, the gun is dead weight.

 

[Josh_TN]

More general CCA statement: if you think the problem is software or processing, then that seems like an easy backfit to any open architecture platform. The main benefit I see to Incr1 is that it introduces aircraft as soon as possible so these concepts are tested en mass with real hardware. IMO, the US has a significant lead in AI right now and one of the ways it can maintain that lead is to get out of virtual environments and surrogate testing and put some planes in the air and see how they behave. The U.S. also has a uniquely large number of relatively off the shelf LO vehicles in the 10,000 lb MTOW range, and is greatly helped in this weight class by its off the shelf civilian turbofan market. There are already four different models of UCAV type aircraft in this weight class using at least 2-3 different commercial engines.

It may well be that Incr1 as a combat aircraft is largely a flop, but I suspect the experience can still catapult the USAF into an undisputed UCAV lead in terms of both human and machine learning.

 

[Josh_TN]

And to add to that, I would argue that the unique American position of having to take the fight all the way to the Pacific with a peer competitor having dozens of readily available air bases and an equivalent if not superior fighter aircraft production infrastructure demands a paradigm change. If AI driven UCAVs are unworkable, well then so is the current and future manned air fleet anyway. There is not another way the U.S. can achieve air dominance in the Pacific without cheaper AI driven aircraft; full stop. It’s all or nothing.

 

[Shoes31]

Gunfire is still going to be hopeless hard to achieve versus firing a 60g HOBs IIR guided AAM at even a couple miles/kilometers range. It a world that will be filled with wrap around IRST and lock on after launch missiles, the gun is dead weight.

I wonder how much guided gunfire rounds would change that. Would be lower heat signature when fired if that's important.

Even if guns are not of much use when up against fighters, they may be very useful for drone defense. Might be worth it to keep the internal gun just for that mission set if required.

 

[Josh_TN]

I wonder how much guided gunfire rounds would change that. Would be lower heat signature when fired if that's important.

Even if guns are not of much use when up against fighters, they may be very useful for drone defense. Might be worth it to keep the internal gun just for that mission set if required.

For an A2A UCAV it’s useless. Look at the weight of a gun installation, even the much lighter Russian ones, and now factor that into the MTOWs we are seeing for Incr1.

 

[Ozair]

I'd say the slowest part of this whole process is the development of the AI command logic - not the manned side. Its seriously hard to make good AI, something that seems to get missed whenever we talk about CCA's.

If all the autonomy you want out of a CCA is a simple fly here while avoiding the A2A bubble, bomb this target, fly back thats easy enough to do but requires a greater workload for the manned element. If you want less workload on the manned fighter, you need software that all you need to do is tell it to suppress enemy defenses here, fly out to this area, and then select best targets of opportunity for attack. It has to be able to defend itself with no intervention. For the air to air side, the software has to be able to designate enemies, figure out which CCA will attack which target, best plan of intercept etc etc and suggest all that to the pilot or be able to take action by itself. That is hard to do, and I don't see that software being ready in the next few years at all. Not with any sort of reliability that I'd want to base my NGAD design on the assumption that CCA's will do a large share of the job.

So if you believe GA and Andruil then they are confident the AI portion is present today and the hard part is convincing the USAF to allow it to operate to its full extent. This corresponds to other comments made by the USAF about the laws of war and having a human in the loop.

This Mitchell Institute article, which may have been linked here already shares the following,

The team also engaged in considerable discussion on if, and when,future counterair CCA should be allowed to go autonomous, a decision they likened to “launching anAMRAAM with a booster” toward a beyond visual range target. In this mode, CCA could continue tooperate in an engagement area until they found a target or were attritted.

...

The HVAA defense team also agreed that CCA command and control, autonomy, and rules of engagement(ROE) were interdependent factors. The team debated linkages between autonomy and rules of engagement(ROE) for using CCA as lead forces without crewed aircraft. Requirements to constantly maintain humansin decision loops could constrain the warfighting potential of CCA in highly contested environmentswhere an enemy is capable of degrading or temporarily denying beyond line-of-sight communications.With software to share situational awareness within CCA force packages and appropriate ROE parameters,the CCA could autonomously perform an expanded set of mission actions without continuously availablelong-distance links to E-7s or other command and control centers.

https://mitchellaerospacepower.org/wp-content/uploads/2024/02/The-Need-For-CCAs-for-Disruptive-Air-Warfare-FULL-FINAL.pdf

 

[Ozair]

I think the U.S. CCAs rather rapidly will evolve to adopt specific behaviors as specified by the pilot: for instance, perhaps a pilot sets one drone to defensive wingman and two others to offensive sweep. The wingman only concerns itself with acting as a radio relay to minimize the manned aircraft’s exposure and synchronizes its flying to put itself in between the fighter and any incoming threat as a stand in jammer, CM dispenser, and if necessary sacrificial decoy. The sweepers move ahead and act as forward sensor platforms with an optional engagement capability if the pilot approves. All of the aircraft reorient themselves with the principle’s movement so the pilot does not have to give new instructions, rather like carrier escorts responding to course changes of CV to launch and recover aircraft and such. This would minimize the heads down time.

I agree and expect that will be standard fare to CCAs. Pilot interaction will be intent based and less specific action based. "Go to kill Box X and patrol" or "protect this asset" or "sweep mode" etc.

Long term I think the controller is ground based and the orders dispensed via satellite. The U.S. is going to have a global low latency system of multiple tactical radio formats inside several years, including laser cross links and downlinks as well basic Link16 and a host of UHF formats. Dedicated high altitude communications gateway UAVs could potentially directly receive optical communications as well as UHF and L bands. I think this combined with higher levels of automation rapidly makes the control of drones via aircrew obsolete, or at least more of a fall back option for redundancy.

I'm still not convinced that CCAs will require so much human control. Even with highly capable satellite constellations there are situations where communications will be degraded or denied from potentially both aircrew and ground based controllers. In that context then the autonomy must be trusted.

 

[Josh_TN]

I'm still not convinced that CCAs will require so much human control. Even with highly capable satellite constellations there are situations where communications will be degraded or denied from potentially both aircrew and ground based controllers. In that context then the autonomy must be trusted.

I think if you are willing to just send UAVs into a no return pitbull mode, pure automation and local directional datalinks are enough. But if you want to have and flexibility strategy and reassigning assets to new missions mid sortie, some kind of over arching organization from outside is preferable.

Though here’s a monkey wrench to thrown that thought process : what if the ground based controller also was not human?

 

[Ozair]

More general CCA statement: if you think the problem is software or processing, then that seems like an easy backfit to any open architecture platform. The main benefit I see to Incr1 is that it introduces aircraft as soon as possible so these concepts are tested en mass with real hardware. IMO, the US has a significant lead in AI right now and one of the ways it can maintain that lead is to get out of virtual environments and surrogate testing and put some planes in the air and see how they behave. The U.S. also has a uniquely large number of relatively off the shelf LO vehicles in the 10,000 lb MTOW range, and is greatly helped in this weight class by its off the shelf civilian turbofan market. There are already four different models of UCAV type aircraft in this weight class using at least 2-3 different commercial engines.

It may well be that Incr1 as a combat aircraft is largely a flop, but I suspect the experience can still catapult the USAF into an undisputed UCAV lead in terms of both human and machine learning.

Agree 100%. Phase one systems are just the start and we may not see more than 100 of them before phase two comes along. The key will be manufacturing infrastructure to support rapid build and then rapid. As I said earlier the external hardware doesn't need to change much, CCAs will be using the same munitions ten years from now as they are today but the software will have evolved exponentially. When most of these will rarely fly then the ability to upgrade existing CCAs to new Blk standards will be relatively simple.

 

[Ozair]

I think if you are willing to just send UAVs into a no return pitbull mode, pure automation and local directional datalinks are enough. But if you want to have and flexibility strategy and reassigning assets to new missions mid sortie, some kind of over arching organization from outside is preferable.

At which point you get into swarm tactics and local autonomy. You could easily program in tactical, operational and strategic goals with associated priorities but agree, overarching direction will always be better.

I wonder as well if much of this will be somewhat irrelevant until CCAs are present in larger numbers and with deeper weapons payloads. Two AIM-120 or four SDB won't go very far before the CCA has to RTB to reload.

Though here’s a monkey wrench to thrown that thought process : what if the ground based controller also was not human?

Agree and if so why host that ground based controller then. Operate a parent platform so the long range comms become a non factor and then you have the laser or MADL style datalinks

 

[Josh_TN]

I also want to speak to the ability to upgrade the control agents in a separate post. I am going to refer to an episode of Battlestar Galactica (the new one, not the one with Face from A-Team). Bear with me.

There is an episode where an ace pilot raider with a mark on its outer surface is grinding down viper pilots. The pilots name it "Scar" for the mark on the fuselage. One of the android cylons reveals that the raiders get downloaded whenever they die, so if they lose a dogfight, they gain all of that experience for their next fight. It is surmised that Scar has "died" several times already.

That is not the extent of the true horror of an AI agent pilot. The full on effect is that all AI platforms will be recording as much as possible and sending it home as much as possible, bandwidth allowing, constantly. How opponent platforms and missile perform compared to estimates, wartime reserve jamming modes, new observed tactics, effective evasion maneuvers for specific enemy weapons, etc.

The real horror is not that any one AI agent keeps coming back; its that the entire system and network keeps getting smarter hour by hour across every platform as more data becomes available, and that the test virtual environment back home that the AI agents get trained on becomes dramatically more realistic with continued combat with the enemy. It would not be one ace pilot coming back; it would be every pilot being continually upgraded to that level. This is not science fiction any more; we are practically there now and certainly will be by end of decade.

The US can either be on the forefront of this kind of technology or be defeated by it.

 

[NMaude]

I also want to speak to the ability to upgrade the control agents in a separate post. I am going to refer to an episode of Battlestar Galactica (the new one, not the one with Face from A-Team). Bear with me.

Snip.

The US can either be on the forefront of this kind of technology or be defeated by it.

That's a very interesting observation there and it has significant implications.

One quibble:

One of the android cylons reveals that the raiders get downloaded whenever they die

Not an android, a Replicant*, they are CINOs (Cylon In Name Only).

*The Replicants weren't robots (Despite the accretions of the Ones) they were genetically-engineered, mass-cloned Human replicants.

 

[Training_Dummy]

humans are generally rate limiting elements of OODA loops so taking the manned component out of whatever timescale CCAs operate on in terms of optimizing the lethality of a dispersed group of sensor and shooter CCAs at different vectors, altitudes and EM profiles (the Boyd kind) as it prosecutes a highly dynamic engagement against a future peer opponent seems very logical.

There’s a broader idea that I keep writing draft posts about that never get completed, but it’s along the lines of aerospace superiority in the future is lot less Top Gun/Maverick and a lot more Enders Game, where a human or small group of humans meshes with multiple AI combat nets controlling assets across a broad AO to achieve very specific temporally and spatially permissive conditions that allow a strategic action to be achieved that would otherwise be impossible.

 

[Ozair]

There’s a broader idea that I keep writing draft posts about that never get completed, but it’s along the lines of aerospace superiority in the future is lot less Top Gun/Maverick and a lot more Enders Game, where a human or small group of humans meshes with multiple AI combat nets controlling assets across a broad AO to achieve very specific temporally and spatially permissive conditions that allow a strategic action to be achieved that would otherwise be impossible.

That is the definition of air superiority, to control the airspace for a specific period of time and allow you to conduct operations without prohibitive interference from air and missile threats. Clearly the same principle applies to AI controlled CCAs being used as the means to gain air superiority and then sustain that for the required time and additionally being happy to relinquish that superiority when it isn't needed.

 

[Training_Dummy]

That is the definition of air superiority, to control the airspace for a specific period of time and allow you to conduct operations without prohibitive interference from air and missile threats. Clearly the same principle applies to AI controlled CCAs being used as the means to gain air superiority and then sustain that for the required time and additionally being happy to relinquish that superiority when it isn't needed.

I agree and perhaps optimal solutions may not be obvious at all to a human under the stress of a full on high threat environment. The problem is bounding behavior so that Kobayashi Maru situations can’t happen.

 

[quellish]

The real horror is not that any one AI agent keeps coming back;

I disagree.

The true horror is that before every command the human sends to the CCA AI he will be presented with a cookie consent form, privacy policy, and end user license agreement.

 

[Josh_TN]

I disagree.

The true horror is that before every command the human sends to the CCA AI he will be presented with a cookie consent form, privacy policy, and end user license agreement.

How to keep the CCAs from being hacked? Captcha!

 

[Scott Kenny]

The ROE constrained the US forces in Vietnam, some of that was perhaps limited IFF but I'm not sure it can all be attributed to that. It could well be argued the very poor performance of the AIM-4 was as much an issue as the ROE constraints,

Putting some number to it, US gun kills were approx 20% of A2A kills over Vietnam and a majority of those were F-105s with no AAM on the aircraft so no other option. Hence I'm not sure the presence of a gun would have significantly changed the kill ratio. Having a better performing AAM over the AIM-4 likely would have made a difference, at least in the early years.

If you are close enough to see the other aircraft then I'm not sure a gun is required. HOBS missiles are good enough now that as long as the target is detectable then a missile seems far more effective than a gun which, with obviously the absence of a turret, is limited to nose pointed engagements. Maybe min fuzing distances for the missiles might be a factor but I doubt it. The presence of a gun would IMO be more a factor in A2G tasking.

To conceive a scenario, a manned NGAD is shot down behind the FLOT and while supporting a CSAR the NGAD wingman needs an appropriate sized weapon to keep adversary infantry away from the downed aircrew's position. A perfect use of a gun. Of course that is a contrived scenario and I'm not anti gun, just not sure the USAF will go there.

All that anyone has to say about not including a gun is "F-4 in Vietnam" and the decision will be reversed.

When you're talking about a 100klb aircraft, including ~3000lbs for gun and ammunition is trivial (if you even need that much for ~500rds of 25mm and the gun).

Taking it a step further then, will we see a gun equipped CCA? Seems ideal for interdiction tasks against soft targets as well a CAS and potentially then also A2A against other drones. They are probably out there but I haven't seen any concepts that have a gun as the primary or even secondary weapon on a CCA.

Maybe. We haven't seen any concepts for CAS CCAs yet.

My mental image was the CAS CCAs not having a gun, leaving their controller with one. So if there's a unit that needs Danger Close gunfire to keep the enemy off of them, there isn't just some drones overhead but also a pilot. We can debate whether that's a USAF pilot or an Army pilot later.

Note that there have been a lot of ideas for drones armed for A2G kicked around, up to and including the LockMart Cormorant.

But can a CCA be really relevant in a dogfight? I mean, you need a lot of engine power and CCA while, being built for range and persistence, doesn´t look like performers there.
So why someone would want them to have a gun for DCA?
Similarly, what kind of profile attack would a CCA use while strafing? Dive and pull-up at 5G? I am not sure they can do that efficiently and have the power to not linger at low alt within ground defense reach.

Not the current increments, no.

If we end up making a CCA that is CAS/ground attack optimized and has a gun, then needing to make that dive and 5-gee pull will be part of the requirements.

I'd say the slowest part of this whole process is the development of the AI command logic - not the manned side. Its seriously hard to make good AI, something that seems to get missed whenever we talk about CCA's.

If all the autonomy you want out of a CCA is a simple fly here while avoiding the A2A bubble, bomb this target, fly back thats easy enough to do but requires a greater workload for the manned element. If you want less workload on the manned fighter, you need software that all you need to do is tell it to suppress enemy defenses here, fly out to this area, and then select best targets of opportunity for attack. It has to be able to defend itself with no intervention. For the air to air side, the software has to be able to designate enemies, figure out which CCA will attack which target, best plan of intercept etc etc and suggest all that to the pilot or be able to take action by itself. That is hard to do, and I don't see that software being ready in the next few years at all. Not with any sort of reliability that I'd want to base my NGAD design on the assumption that CCA's will do a large share of the job.

That was demonstrated back in about 2005, dude. At least for ground attack. A flight of 4 drones were all told "attack this location, there's SAM/AAA that needs to be suppressed as well." The drones chose which bird was going for the main target and which were going to be decoys and which were on SEAD, based on which drone was where. Then they attacked.

 

[4decaa]

Speaking of CAS and unmanned platforms… do u even Mojave bro?

 

[Scott Kenny]

View attachment 742008
Speaking of CAS and unmanned platforms… do u even Mojave bro?

I fully expect the US Army to drop their MQ-1C Gray Eagles for Mojaves here pretty soon. More capacity, uses the same ground stations, only a little retraining on the engine side. (I'm sure there's stuff in the US Army inventory using PT6s)

 

[CiTrus90]

I disagree.

The true horror is that before every command the human sends to the CCA AI he will be presented with a cookie consent form, privacy policy, and end user license agreement.