Chinese UCAV/UAV/CCA - news & technology

[Deino]

I am not sure if anyone mentioned this before
It's noteworthy that during the parade, Chinese officials used three distinct classifications: 'reconnaissance-strike integrated UAVs,' 'unmanned loyal wingmen,' and 'unmanned air combat aircraft.' Specifically, in order: the first aircraft in the front row (WZ-10) is a 'reconnaissance-strike integrated UAV,' the two slightly smaller ones in the second row are 'unmanned loyal wingmen,' and the two larger ones in the third row are 'unmanned air combat aircraft.' Personally, I believe this may imply that these two larger UAVs possess a certain level of independent air combat capabilities

Thanks for this, but could you please add the full Chinese designation and Chinese text?

 

[donnage99]

It would surely come in handy on topics relating to Chinese or select Russian topics where US and EU based boomers are struggling to comprehend technological advances east of the German border.

Alternatively the rolling eyes emoji would be another excellent choice.

Though western bias is present (as in russian bias or chinese) online or media, it's not always the case. Sometimes doubts come from an understanding of physics.

In fact, here's the cream of the crop type of man from the East (design firm of the J-20) analysis of the Russian T-75 where they cast doubts over certain performance claims by Sukhoi (credit for translation to ZWZ on f-16.net):

 

[Dilandu]

As it stands, yes - the human brain is very important to the fighter and I don't think it'll ever change, but that's a very very broad statement in that field. It's important insofar as figuring out complex and uncommon cases in the broader tactical picture, but it may well become inferior in other, more tactical cases (like flying the perfect trajectory, speed, etc to perform an action). But what is the gap you are speaking of here? The number's / types gap or something else?

Actually, the practical experience of human vs AI pilot tended to indicate that humans are very predictable and not actually capable of grasping broad tactical picture. There were DARPA tests, where during simulated air combat, a team of AI pilots on inferior aircrafts (4th gen fighters, wirh short and medium range missiles) soundly defeated a team of human pilots on superior aircrafts (5th gen fighters, with long-range missiles). The AI won all engagements through better coordination and the ability to plan fast 3D actions in real time. Basically while human pilots were only able to concentrate on immediate goals and threats, and acted more or less individually (and withing predictable tactical schemes), AI planned in advance, so each its action was part of a global tactical scheme.

 

[Dilandu]

Though western bias is present (as in russian bias or chinese) online or media, it's not always the case. Sometimes doubts come from an understanding of physics.

Well, there I wouldn't claim it to be a bias, since West (mainly America) have a rather big lead in stealth technology and its application (they started to dig into stealth since 1950s, and experimented more or less constantly there). So it stands to reason that American experts may notice flaws that aren't obvious for Russian or Chinese ones.

 

[SspewixzzZ]

Thanks for this, but could you please add the full Chinese designation and Chinese text?

The original text of the parade introduction:
人工智能，空天争锋：
接下来通过的是，空中无人作战方队
受阅的新型察打一体无人机(reconnaissance-strike integrated UAVs)、无人僚机(unmanned loyal wingmen)、无人制空作战飞机(unmanned air dominance aircraft)、舰载无人直升机(Carrier-based unmanned helicopter)，
可隐蔽出击、广域覆盖、自主协同（Stealthy operation, extensive coverage, autonomous collaboration），创新未来空战新样式。

Most people seem to be aligning with these designations - this is from Huitong's site (https://chinese-military-aviation.blogspot.com/p/uavucav-iii.html)

Thanks

Edit:I think "unmanned air dominance aircraft" is a better translation.

 

[VTOLicious]

Most people seem to be aligning with these designations - this is from Huitong's site (https://chinese-military-aviation.blogspot.com/p/uavucav-iii.html)

...single WS-10C class engine without a TVC nozzle...

Thx for the summary!
I would assume TVC would be a useful feature for an “unmanned air dominance fighter” without vertical stabilizers.

 

[EmoBirb]

do not misquote me, UCAVs with jet engines are expensive, The Russian build a UCAV with a large jet engine, they can not produce many, regardless is Russian or Chinese or French, the jet engine has the same price regardless is for an UCAV or a manned aircraft, the inlet needed for that jet engine also has the same needs regardless is for an UCAV or a manned aircraft, the speed is the same drag is the same problem for UCAVs or manned aircraft.

The PL-15 or AIM-260 has the same size regardless is fired by a UCAV or a manned aircraft same other types of ammunitions.

Let me get the full quote then

interesting to see these drones, but these are obsolete technology, first they are no VTOL, so easy to attack in their bases, second due to DSI intake or even simpler types of Intakes they are slow, lack of ventral or dorsal fins make complex trailing edge control, no vertical tail also means less control at supersonic flight, no horizontal tail controls affect roll effectiveness, having no canards or LERX means when supersonic flight happens they become nose heavy, change the nozzle into a rectangular cross section also means loss of thrust, keep it with round cross section equals larger IR signature.

They are intended more for propaganda than real combat effectiveness, they might offer a bit of advantages in local small conflicts, but for big wars they are obsolete.

You literally called them obsolete, you literally said they're meant for propaganda rather than combat.

Yes jet engines are costly, but they're just one of the many pieces that make an aircraft expensive. One of the most expensive things about a fighter jet is the pilot and the systems that support the pilot and feed said pilot with information. The cost and time to train, feed and house a pilot shouldn't be underestimated. An large unmanned aircraft can cut back on life cycle cost and can create a more compact package because it doesn't have to accommodate one or more people.

You seem to equate "drone" with high volume, low cost, high attrition systems. Like these are DJIs. Instead you should think of these as J-10s but with superior stealth and without a pilot. High end, high cost, high capability drones are going to be their very own niche. These will be the aircraft that one day replace manned fighters. Then you have mid tier UCAVs which are less capable, cheaper and more attritable, think something like the YFQ-42 and YFQ-44. And below that come several levels of low cost, high quantity drones. If you truly believe these large UCAVs are obsolete, then everything that flies are obsolete. Because these drones are closer to J-10s and F-16s with stealth characteristics than they are to a Reaper or DJI quadcopter.

Jet engines and intakes are not what drive the majority of the cost of a capable fighter aircraft (especially when you stick to a single engine design on top of that).

And equating the lack of VTOL capabilities with obsolescence doesn't require a dedicated reply.

 

[EmoBirb]

Though western bias is present (as in russian bias or chinese) online or media, it's not always the case. Sometimes doubts come from an understanding of physics.

In fact, here's the cream of the crop type of man from the East (design firm of the J-20) analysis of the Russian T-75 where they cast doubts over certain performance claims by Sukhoi (credit for translation to ZWZ on f-16.net):

Equating the T-75 to the Su-57 is certainly something else, given the vastly different geometry of the airframes at hand (they really couldn't be more different). But I'd take the opinion of people who have a horse in the race (AVICs J-35) on the export market with a grain of salt anyway.

 

[overscan (PaulMM)]

Equating the T-75 to the Su-57 is certainly something else, given the vastly different geometry of the airframes at hand (they really couldn't be more different). But I'd take the opinion of people who have a horse in the race (AVICs J-35) on the export market with a grain of salt anyway.

Equating the T-75's likely level of stealth to the fighter many of its systems presumably derive from, designed by the same people, and which is likely to be made in the same production processes... yep, no clue why they'd do that.

If China makes same conclusions about Russian tech as US/Europe, then China is lying to sell its potential export fighters. Okie dokie, I see how this works. Only Russia tells the truth, everyone else lies.

We seem to be back at the same argument, in yet another topic.

 

[F-14ATomcat]

Let me get the full quote then


You literally called them obsolete, you literally said they're meant for propaganda rather than combat.

Yes jet engines are costly, but they're just one of the many pieces that make an aircraft expensive. One of the most expensive things about a fighter jet is the pilot and the systems that support the pilot and feed said pilot with information. The cost and time to train, feed and house a pilot shouldn't be underestimated. An large unmanned aircraft can cut back on life cycle cost and can create a more compact package because it doesn't have to accommodate one or more people.

You seem to equate "drone" with high volume, low cost, high attrition systems. Like these are DJIs. Instead you should think of these as J-10s but with superior stealth and without a pilot. High end, high cost, high capability drones are going to be their very own niche. These will be the aircraft that one day replace manned fighters. Then you have mid tier UCAVs which are less capable, cheaper and more attritable, think something like the YFQ-42 and YFQ-44. And below that come several levels of low cost, high quantity drones. If you truly believe these large UCAVs are obsolete, then everything that flies are obsolete. Because these drones are closer to J-10s and F-16s with stealth characteristics than they are to a Reaper or DJI quadcopter.

Jet engines and intakes are not what drive the majority of the cost of a capable fighter aircraft (especially when you stick to a single engine design on top of that).

And equating the lack of VTOL capabilities with obsolescence doesn't require a dedicated reply.

Sorry, but you still do not understand, there are many types of drones, the ones with jet engines have the same aerodynamics that manned aircraft have, same engines, structures and even more complex avionics however Drones with jet engines have the same constraints a manned fighter has.

To give you an example Su-57 is twin engined but its UCAV is single engined, it is only cheaper if it is because it has a single engine.

A drone follows the same physics a manned aircraft has regardless of AI. AI is to control sensors only that, to make automatic some commands.

AI does not make a faster engine per se it is the engine and aerodynamics of the aircraft you can have a very smart drone but the drone can not break physics it is no war thunder or a video game, if the missile out runs it or is more agile or jammed proof, it will down it regardless of AI.

Stealth is another fantasy, if the drone gets close it will be downed, so it also needs to fire as a manned aircraft at BVR, AI is a control system for sensors and actuators, the aircraft still is controlled and constrained by physics .

Drones will be shot down too, the difference is no pilot dies and since they are expendable they are made simpler than manned fighters, less complex and at the end less capable.

View: https://www.youtube.com/watch?v=NaY5rF7n6VE


remember if this case is two F-16s one with AI another with a Pilot now what about a manned F-22 versus that AI controlled F-16 at BVR? what matters there is the radar, not AI and the supercruise ability, if the AI controlled uses AIM-7s of short range versus the manned F-22 with AIM-260 who do you think will win, hardware makes software to succeed

 

[Josh_TN]

Actually, the practical experience of human vs AI pilot tended to indicate that humans are very predictable and not actually capable of grasping broad tactical picture. There were DARPA tests, where during simulated air combat, a team of AI pilots on inferior aircrafts (4th gen fighters, wirh short and medium range missiles) soundly defeated a team of human pilots on superior aircrafts (5th gen fighters, with long-range missiles). The AI won all engagements through better coordination and the ability to plan fast 3D actions in real time. Basically while human pilots were only able to concentrate on immediate goals and threats, and acted more or less individually (and withing predictable tactical schemes), AI planned in advance, so each its action was part of a global tactical scheme.

Do have any links to any documents that discuss this? I do not find it hard to believe but I have not heard of testing outside one on one BFM.

 

[EmoBirb]

Equating the T-75's likely level of stealth to the fighter many of its systems presumably derive from, designed by the same people, and which is likely to be made in the same production processes... yep, no clue why they'd do that.

Stealth is predominantly dictated by the geometry of the aircraft. And in that regard the Su-57 and T-75 are just fundamentally different (would be nice if you could move this to the T-75 thread).

And yes, AVIC has all the reasons to downplay an export market rival. Same how Lockheed/USAF has all the reasons to trash talk the J-20 ("muh canards") and J-35 ("cheap knock off copy"). Weapons are a commodity meant to be sold first and foremost.

 

[EmoBirb]

To give you an example Su-57 is twin engined but its UCAV is single engined, it is only cheaper if it is because it has a single engine.

And because it has no IRST, requires less material and structural reinforcement with expensive aerospace grade material, less man hours to produce, to maintain, doesn't need a proper fighter pilot with years of training to operate it and is equipped with less ordnance, doesn't supercruise, has no TVC nozzle...the list goes on.

Enough said.

 

[Dilandu]

Do have any links to any documents that discuss this? I do not find it hard to believe but I have not heard of testing outside one on one BFM.

My memory wasn't as good as I thought, it wasn't DARPA:

 

[F-14ATomcat]

Do have any links to any documents that discuss this? I do not find it hard to believe but I have not heard of testing outside one on one BFM.

tell me how AI increased the radar detection range? that is fantasy because for AI to beat a stealth F-22 manned aircraft has to detect it first, you can not hit what you can not see, no you can not out run what is faster than you, pure fantasy, The F-16 versus F-16 AI against human is logic, because remember the human pilot is constrained by G forces, while the other jet was not, is human physiology what defeated the human but with AI or not you can not shoot down what you can not see, AI can not take decisions if it can not see the target

 

[F-14ATomcat]

And because it has no IRST, requires less material and structural reinforcement with expensive aerospace grade material, less man hours to produce, to maintain, doesn't need a proper fighter pilot with years of training to operate it and is equipped with less ordnance, doesn't supercruise, has no TVC nozzle...the list goes on.

Enough said.

sorry better study physics and about aircraft design, aircraft are not video games

 

[targetpractice]

This is just a crappy drawing, but could be like this:
View attachment 783591

Any real world examples of this set up?

 

[EmoBirb]

sorry better study physics and about aircraft design, aircraft are not video games

I study aerospace engineering, I think I'm good.

 

[Dilandu]

tell me how AI increased the radar detection range? that is fantasy because for AI to beat a stealth F-22 manned aircraft has to detect it first, you can not hit what you can not see, no you can not out run what is faster than you, pure fantasy, The F-16 versus F-16 AI against human is logic, because remember the human pilot is constrained by G forces, while the other jet was not, is human physiology what defeated the human but with AI or not you can not shoot down what you can not see, AI can not take decisions if it can not see the target

Just for example what AI could do, and human pilots couldn't:

ALPHA has the ability to continue to seek positional dominance
even during evasive maneuevers. For example, if evading missiles
from approximately the same direction, two reds will elect to perform
maneuvers in differing directions if possible. Additionally, ALPHA
does not stop tracking defensiveness to all other hostiles if evading a
missile nor does performing an evasive maneuver stop ALPHA from
considering taking shots at all threats. ALPHA also does its best to
prevent the opposing force from acquiring positional dominance
during these times; it will take defensive shots to keep hostile aircraft
from obtaining such an advantage

As I said above; the AI is able to comprehend the whole situation in realtime, and plan even defensive reactions in such way, that its forces would gain advantage from such defensive reactions. It's like a real-time chess play in 3D. The AI main advantage against humans is not the G-force constrains, but the ability to evaluate situation faster and much more comprehensively, and plan accordingly.

 

[Josh_TN]

Just for example what AI could do, and human pilots couldn't:



As I said above; the AI is able to comprehend the whole situation in realtime, and plan even defensive reactions in such way, that its forces would gain advantage from such defensive reactions. It's like a real-time chess play in 3D. The AI main advantage against humans is not the G-force constrains, but the ability to evaluate situation faster and much more comprehensively, and plan accordingly.

An additional capability AI likely has or will have is the ability to expend an aircraft to complete a mission or gain advantage. To use the chess analogy again, it might well make a sacrificial play to unbalance an opponent formation, provide mid course guidance updates, or otherwise gain advantage in a situation where as human pilots likely would not. Even if human pilots were so willing to, selection of who and when is likely problematic.

It reminds me of the old 1980s vintage hypothetical that came about after the invention of all aspect IR missiles: the heavier, radar equipped interceptor fires a SARH guided missile at a small cheaper aircraft equipped with all aspect IR missiles and has to maintain a closing geometry. The cheaper fighter closes as well, gets to WVR, and launches a fire and forget missile before the radar guided weapon hits it. The IR missile however does not need the launching aircraft for guidance, and the cheaper aircraft gets a kill on the more expensive one.

This scenario is obviously very artificial and makes a lot of assumptions, one which is that the IR equipped pilot simply runs into an oncoming missile rather than evading. But it demonstrated the potential unequal fight that all aspect missiles might force upon an organization whose doctrine emphasized larger more complex aircraft using SATH missiles.

Software agents that were assigned specific values to themselves and opponent targets might easily come to the conclusion that higher rates of attrition might be forced on an opponent by simply trading aircraft for opposing units. This would be especially true if the AI was controlling far less valuable UAVs and decoys in addition to CCA style aircraft. This could potentially be applied across a wide swath of airspace, trading resources in one area for decisive effects in another.

 

[Reddington777]

Actually, the practical experience of human vs AI pilot tended to indicate that humans are very predictable and not actually capable of grasping broad tactical picture.

The air force as of April 2025 still seems to disagree with this:

"Though AI quickly processes large datasets, it lacks context sensitivity and reasoning. In
HMT, human strengths (e.g. intuition and reasoning) and machines (e.g. fast data
processing) combine to maximize capability. For AI-enabled processes, the exercise of
appropriate human judgment is an important consideration, especially in strategically
high-risk operations."

https://www.doctrine.af.mil/Portals/61/documents/AFDN_25-1/AFDN 25-1 Artificial Intelligence.pdf
Trust but verify. Trust but verify.

There were DARPA tests, where during simulated air combat, a team of AI pilots on inferior aircrafts (4th gen fighters, wirh short and medium range missiles) soundly defeated a team of human pilots on superior aircrafts (5th gen fighters, with long-range missiles).

I read the paper you linked for this and it is absolutely a great read so thank you for that. And yes I think this is a really good simulation of the kinds of uncertainty and randomness faced during air combat. However, I would argue that the results obtained by this paper is what I would expect the general case to be. If the controlling AI can't even perform this, then CCAs are a failed gamble.

A few points of note:

• "The aircraft for both teams are identical in terms of their mechanical performance." then later on "computational cost and complexity of training an AI within AFSIM is a limiting factor"
  • I think you may or may not have mis-remembered or maybe you were quoting some other study. All aircraft involved were the same save for loadout differences.
  • This research group had done different experiments with different mission sets, but keep in mind that all involved aircraft are homogenous in capability and role. There are also limited actors in this. In the real ABMS, it'll be scaled up to include a much wider variety of capabilities and roles. As roles and capabilities increase, permutations between descritized rules also increases, which makes designing them even harder.
• "Each FIS has membership functions that classify the inputs and outputs into linguistic classifications, such as “far away” and “very threatening”, as well as if-then rules for every combination of inputs, such as “If missile launch computer confidence is moderate and mission kill shot accuracy is very high, fire missile”. By breaking up the problem into many sub-decisions, the solution space is significantly reduced. The cost of this approach is a risk of not accommodating for all sources of coupling. Through optimal design this can be minimized if not entirely
  mitigated"
  • As I said before, they turned an infinite solution space into a finite problem space by discretizing the problem space. They also assigned each fuzzy inference system to one task that takes a subset of rules, but that also means that you have to carefully design the rules coupling and discretization heuristics so that important cross effects between each FIS doesn't get missed. Remember that right now the scope is small and descritizing the problem space results in fairly small information loss. In something like ABMS, it's going to be a monumental job to try and design the heuristics that discretizes this space.
  • Even as discretization makes the solution space scalable and more robust, you are still prone to problems like:
• "The fact that the intended opponent is a team of humans isan interesting complexity that is difficult to account for in the typical machine learning setting. The pilots could perform extremely strange, errant, and apparently suboptimal behaviors just to confuse or exploit ALPHA...To account for this and enable post-training testing andadjustments, a simple User Interface was created to allow humans to fly against ALPHA in-house. The human operators have been able to achieve a modest, but reasonable, amount of control over two blue aircraft to compete against ALPHA"
  • These were humans flying against ALPHA initially with the control interface. In a near peer fight, it's going to be AI vs adversarial AI. This isn't insurmountable at all, but it does add further complexity and randomness that you either need suboptimal catch alls for, or huge resources devoted to learning or paring down the problem space.
  • Also note that a UI was built to allow pilots modest control. This doesn't at all mean that this program couldn't test adversarial attempts to confuse and screw with the AI, but it limits the amount of crazy you could throw at it. Again - none of this is to say that this study is invalid or that human pilots flying against AI will necessarily always win, but the key point here is that an AI might excel at even complex tasks, but if you understand the system and its constraints, there may be situations where you can take measures that the AI doesn't understand or know about. And how much the AI can counteract this depends on how good your adversarial training is.
  • Hand in hand with the previous point is that it takes time for an AI to be trained to adapt to new tactics and weapons.
  • As I said before, whether its this study or DARPA's ACE program, we are addressing the 90% here and we try our best to address the remaining 10% too, but as with manufacturing, you always need to account for the fact that you can't account for everything. So at a certain point it becomes a tradeoff for developing a solution to a problem vs accepting partial or total loss of the platform. More often than not, we develop for the 90%, reach for the 10%, achieve 94% and leave the remaining 6% for the human to decide on.
• Geno noted how the first generation of red ALPHA held its own against the blue variant of ALPHA, but the resulting engagements often ended with heavy losses for both sides. Psibernetix and Geno worked together to develop tactics, techniques, and procedures to overcome red ALPHA’s payload and no-AWACS disadvantage, capitalize on blue’s mistakes, and take advantage of numeric platform superiority (when the situation presented itself).
  • This expresses what my point here, which is that there remains many cases when a human is needed to enrich how the AI learns. With new tactics and capabilities introduced, the more disruptive they are, the more you need the human in the loop to provide said context.

The AI won all engagements through better coordination and the ability to plan fast 3D actions in real time.

Yes indeed. Humans cannot compare in reaction time and the number of variables we can adequately consider at any given moment.

Basically while human pilots were only able to concentrate on immediate goals and threats, and acted more or less individually (and withing predictable tactical schemes), AI planned in advance, so each its action was part of a global tactical scheme.

You'd conclude that if you strip away the context in which this observation was made - which was in a bounded and limited simulation in the context of this report. It's a long stretch to say that humans acted individually and didn't do long term planning without actually understanding in greater detail how the control of the human controlled blue team was set up. Note that blue was described as taking overly agressive shots and maneuvers while ALPHA controlled red was conservative and precise about its timetables. It's not terribly hard to understand why blue was aggressive given the loadout advantage.

• The test never flipped the loadout and situational advantages. Although I'm sure ALPHA may well win that too, there's no telling whether it makes similar decisions to human pilots i.e high aggression and shooting first
• ALPHA was trained by EVE on the baseline controller AFRL uses, which means it's trained specifically to fight with and against the predictable and common doctrine rules. To what degree the interface allows one to stray from this doctrine and to what degree ALPHA can counter these isn't clear.

Having said that, yes - humans usually don't make make optimal decisions, but it's not necessarily because AI plans ahead in ways that humans can't or can't be assisted to achieve, it's the decision loop and considered parameters that outpaces humans.

What this doesn't account for or doesn't account for completely are the rest of the 10%. The paper does talk extensively about using real world harvested data that accounts for some level of system associated failures, but that's only the tip of the iceberg here. In practice, you have immediate jamming, deceptive EW, lossy links, and delays and connection time delays. You have elements that may or may not be currently connected to the system. You have manned elements already performing non-optimally compared to what AI would do. You also have to remember that large models with a ton of parameters do begin to do poorly unless you can structure that data using well designed algorithms. You also can't fully trust AI to train itself to discover these algorithms either because in most instances you must be able to interpret why the AI decided to do what it does to verify that it isn't misleading itself. So ultimately, at best, AI is at the stage where at the very minimum a human is needed to trust and verify critical decisions. It's why there's a whole sub-field of ML dedicated to interpretable machine learning.

Another thing of note here is that ALPHA is a central node controller that omnipotently understands the state of each commanded aircraft. Realistically, you aren't going to have that in the real world. In the real world ABMS or the BMS system in controlling fighters is going to be taking on the high level controls of the flight while the state of each part of the system requires transmission in order to be communicated. The federated nature of this now introduces another factor, which is that your central BMS that acts so well coordinated in the simulation now relies on those transmission links in order to make the critical decisions - links that can be jammed, degraded, fooled, lossy, high latency or lost entirely due to various hardware / software problems.

As I said above; the AI is able to comprehend the whole situation in real time, and plan even defensive reactions in such way, that its forces would gain advantage from such defensive reactions. It's like a real-time chess play in 3D. The AI main advantage against humans is not the G-force constrains, but the ability to evaluate situation faster and much more comprehensively, and plan accordingly.

Again - that "whole situation" is a situation reduced in scope and simplified greatly by having homogenous participants, limited numbers and an unclear control mechanisms given to the human in controller. It's feedback loop is faster and it can account for more variables. It can probably be built to ameliorate many of the difficulties I've listed already, but when it encounters things that forces it into catch alls, those catch alls may or may not be optimal anymore.

To conclude, while this is all very promising, when you actually account for implementation, limitations and encountered operational difficulties, the reality is far away from the simplistic and over generalized conclusions being drawn.

 

[Dilandu]

The air force as of April 2025 still seems to disagree with this:

The Air Force fought tooth & nail against anything that might challenge the dominating culture of fighter jocks elite. Of course they would be downplaying the advantages of AI, too. For compairson - imagine reaction of late Medieval knights, when their king started to inquire "maybe we should reduce the heavy cavalry and hire more pikemans and arquebusier instead?"

 

[Reddington777]

The Air Force fought tooth & nail against anything that might challenge the dominating culture of fighter jocks elite. Of course they would be downplaying the advantages of AI, too.

And... where would the evidence for this be again?

For compairson - imagine reaction of late Medieval knights, when their king started to inquire "maybe we should reduce the heavy cavalry and hire more pikemans and arquebusier instead?"

Pikemen and arquebusiers didn't spell the end of the cavalryman either did it? Knights -> men at arms -> totenkopf cuirassiers -> lancers -> french heavy cuirassiers & Dragoons & Reiters. The Cavalryman never truly went away until machines guns and tanks became a thing - literally. Lancers were still a thing at the start of WWI and even now, the cavalry lives on vicariously through various MBTs.

Here's what the USAF had to say about introducing new technologies for air superiority 2030:

"The speed of capability development and fielding will be critical to retain the U.S. advan-
tage in the air. As the pace of technological advancements continue to increase the Air Force must
leverage experimentation and prototyping to more rapidly infuse advanced technologies into the
force. Additionally, the Air Force must reject thinking focused on “next generation” platforms.
Such focus often creates a desire to push technology limits within the confines of a formal pro-
gram. Such efforts should be accomplished within the S&T portfolio and proven through effec-
tive prototyping, harvesting when mature to a sufficient level for transition. Pushing those limits
in a formal program increases risk to unacceptable levels, resulting in cost growth and schedule
slips. This put such programs at risk of cancellation due to their nearly inevitable underperfor-
mance, and results in delivery of capabilities “late to need” by years or even decades."

So how is this impractical again? Being prudent and waiting for technological maturation in existing systems isn't at all limited to the air force. It's the product of politics and overall political environment. If anything, the fighter jock elite as of late has found greater prominence posting shitty takes on youtube or selling their services to China by training their NavAir nowadays than they are actually pulling strings and guiding the overall direction of development.

 

[Dilandu]

Pikemen and arquebusiers didn't spell the end of the cavalryman either did it? Nor did cannons either did it? Knights -> men at arms -> totenkopf cuirassiers -> lancers -> french heavy cuirassiers & Dragoons & Reiters. The Cavalryman never truly went away until machines guns and tanks became a thing - literally. And even then, the cavalry lives on vicariously through various MBTs.

My point was more about asking the peoples who are personally interested in maintaining status quo about their interest to change it.

 

[Reddington777]

My point was more about asking the peoples who are personally interested in maintaining status quo about their interest to change it.

Well sure, but I was referring to the application upon AI replacing pilots. Keeping the human in the loop and trusting but verifying has nothing to do with maintaining the status quo. The efficacy of AI and the corresponding reduction in role for humans hasn't necessarily made the human pilot obsolete.

There may well be people interested in maintaining the status quo, but being scared of adopting AI is just not something anyone is blocking. If anything, from joint command structures all the way down to the infantryman, AI is being implemented and integrated everywhere they can be. The status quo of manned fighters was already fading during the UCLASS competitions 10 something years ago let alone now. I don't think being reluctant to relinquish complete control to AI agents whose applications and caveats are still being investigated counts as obstructionism or maintaining the status quo.

 

[F-14ATomcat]

Just for example what AI could do, and human pilots couldn't:



As I said above; the AI is able to comprehend the whole situation in realtime, and plan even defensive reactions in such way, that its forces would gain advantage from such defensive reactions. It's like a real-time chess play in 3D. The AI main advantage against humans is not the G-force constrains, but the ability to evaluate situation faster and much more comprehensively, and plan accordingly.

Pure blah blah blah, a pilot suffers G LOC and blurred vision do you get that? a Pilot at 9Gs will have his blood going to places that can render him unconscious.

A human is helped by computers got it? that can process the same information a Drone has but a pilot suffers G LOC, what unpais a human pilot is excessive G LOC

The fighter also is dependant upon aircraft performance envelop and sensor detecting capability.

If you think any of these drones are superior by AI you are totally wrong, the most important factor is the detectability of the target and the radar set the fighter has.



This drone for example has a fixed air intake which limits pressure recovery and reduces uninstalled thrust at speeds beyond Mach 1.1.

Try to compete with F-22 in supercruise is a joke, the F-22 will out run it, if you out run a rival the missile fired will very likely miss, if your missile is shorter ranged sorry UCAV can not shoot down the rival flying at super cruise speed.



This aircraft regardless uses AI has aerodynamic constraigns




which Drone will fly better the one with swept angle or the one straight wing? of course it depends in the speed.

https://eaglepubs.erau.edu/introductiontoaerospaceflightvehicles/chapter/supersonic-flight-vehicles/When transitioning from transonic to supersonic speeds, the aircraft's center of pressure (CP) moves backward due to the formation and movement of shockwaves on the wings. As the airflow over the wings exceeds the speed of sound, shockwaves develop, creating areas of high pressure that shift the overall lift distribution rearward.

An aircraft with canards like Eurofighter has for a reason a canard because this reduce the center of pressure movement at supersonic speeds, the drone without a canard will become less agile as it flies faster regardless it has AI or the aircraft is manned by a pilot,





AI will not be able to do a thing about that beyond what a computer in a manned aircraft can do


For a reason the F-14 had vanes that deployed at high speeds
Man on the Spot

“UAVs have a great future, increasing popularity and usefulness, but no machine can replicate the human gift of discovery and situational awareness,” says Air Marshal Vinod Patney, former Vice Chief of Air Staff, Indian Air Force (IAF). Today’s technology is insufficient to allow unmanned aircraft to make independent, complex judgements in an unknown environment and so must be linked to human judgement, albeit on the ground. The data links can be denied or deceived. Threat of cyber attack today is more potent than Al-Qaeda. Delays in commanded guidance could be crucial if not lethal. All UAV data is still analysed and interpreted by a human.

As it stands, till date, unmanned aircraft have flown only in uncontested airspace. As the focus of the US military shifts to the Pacific, the Air Force could be confronted by enemy fighters in unfriendly skies, a mission UAVs aren’t yet designed to handle. Of the 25 most expensive programmes of the US Department of Defense (DoD), none is unmanned. The same is true for all major air forces of the world. Even with a disruptive technology like UAVs, one doesn’t let go tried and tested systems, analysts feel.

The first recorded air-to-air engagement was in December 2002, between an Iraqi MiG-25 and an American Predator UCAV armed with Stinger missile. Both fired a missile at each other. The UAV was shot down and the same could be seen through the video downlink. A fighter aircraft has the positioning and speed advantages and better chance of an air-to-air kill. The United States Air Force UAV accident rate at 0.9 per 10,000 hours is much worse than 0.2 per 10,000 hours for manned aircraft. UAVs also require significant manning for their operations. At approximately $75 million ( Rs. 450 crore), the Global Hawk, a Class III HALE UAV, is costlier than many a modern fighter. Collateral damage from UAV attacks continues to be high at 32 per cent. The USA nearly lost a friend in Pakistan when a large number of civilians were killed in one such attack.
https://www.sps-aviation.com/story/?id=1278

 

[Josh_TN]

@F-14ATomcat :

I think the discussion between @Dilandu, @Reddington777, and myself is outside of the PLAAF vs USAF CCA debate. We discussing the practical applications and limits of AI more broadly. You seem to think we advocating for a particular side and I do not think that is the case.

 

[donnage99]

Equating the T-75 to the Su-57 is certainly something else, given the vastly different geometry of the airframes at hand (they really couldn't be more different). But I'd take the opinion of people who have a horse in the race (AVICs J-35) on the export market with a grain of salt anyway.

Huh? Where did the chinese equate T-75 to Su-57? Also, ad hominem fallacy. Yes there are several erroneous points but not every single one. The point to illustrate here was that the advertisement by Sukhoi, especially its range against the scrutiny of math, doesn't check out. And that was done by the Chinese, not some Western boomers. If your math shows something different, please share

Also, ignore F-14ATomcat. We all did after it was clear his original comment was, in fact, not sarcasm.

 

[F-14ATomcat]

@F-14ATomcat :

I think the discussion between @Dilandu, @Reddington777, and myself is outside of the PLAAF vs USAF CCA debate. We discussing the practical applications and limits of AI more broadly. You seem to think we advocating for a particular side and I do not think that is the case.

The first recorded air-to-air engagement was in December 2002, between an Iraqi MiG-25 and an American Predator UCAV armed with Stinger missile. Both fired a missile at each other. The UAV was shot down and the same could be seen through the video downlink. A fighter aircraft has the positioning and speed advantages and better chance of an air-to-air kill. The United States Air Force UAV accident rate at 0.9 per 10,000 hours is much worse than 0.2 per 10,000 hours for manned aircraft. UAVs also require significant manning for their operations. At approximately $75 million ( Rs. 450 crore), the Global Hawk, a Class III HALE UAV, is costlier than many a modern fighter. Collateral damage from UAV attacks continues to be high at 32 per cent. The USA nearly lost a friend in Pakistan when a large number of civilians were killed in one such attack.
https://www.sps-aviation.com/story/?id=1278

My whole point was a superior airframe engine combination will beat an inferior airframe engine combination regardless of AI.

And I said obsolete because, because for big war, SLBM, ICBMs and missile interceptors are far much more important specially high energy weapons that can down missiles.

Cool any way I leave it here enjoy the topic, I do not want to take it off topic

 

[Josh_TN]

That is simply historically inaccurate: there are numerous instances of technically inferior aircraft being utilized with sufficiently more skill by specific organizations or formations such that the superior aircraft failed routinely.

 

[Dilandu]

Pure blah blah blah, a pilot suffers G LOC and blurred vision do you get that? a Pilot at 9Gs will have his blood going to places that can render him unconscious.

I'm kinda puzzled, why for Holy Kiten sake you suddenly got the strange notion that I do not realize human G limitations. My point was, that AI pilot advangates went far beyond merely being unlimited by G-forces.

 

[SspewixzzZ]

Try to compete with F-22 in supercruise is a joke, the F-22 will out run it, if you out run a rival the missile fired will very likely miss, if your missile is shorter ranged sorry UCAV can not shoot down the rival flying at super cruise speed.

I'm really not quite sure how you reached this conclusion... I think there's no absolute causal relationship between a missile's ability to lock onto a target and whether the launch platform is faster than that target.

 

[TwoTribes]

 

[Tomboy]

Someone did some measurement for the smaller aircraft. As a reference F-35A is 15.7m long and could carry 4 AMRAAM internally.

 

[Reddington777]

I'm really not quite sure how you reached this conclusion... I think there's no absolute causal relationship between a missile's ability to lock onto a target and whether the launch platform is faster than that target.

Hes not talking about being able to lock on.
Though in modrn radars, theres still a small penalty for locking receding targets anyway.

Hes talking about the missiles kinematics. If launched at a cold target, your missile has longer to travel because its trying to catch up to the target. As long as the target is not inside the no escape zone, the target can maneuver and force the missile to maneuver with it and the missile eventually loses enough speed and cant catch up. Even a mach 5 missile isnt going to catch a lot of cold target its fired at thanks to this.

However, if you are in the no escape zone, that means kinematically, you have no way of forcing the missile to lose enough energy not to hit you. Then you'd pray that the missile either wont acquire you, or acquires the wrong targets (counter measures of various kinds). With dual pulse motors and air breathing missiles though, the NEZ is much more expanded compared to earlier missiles.

 

[Josh_TN]

+1

I think it comes down to detection/engagement range - if you are inside the NEZ when you’re fired upon, the extra speed is not going to be decisive. And it is worth noting even if you are outside the zone, you are still required to make a lot of hard maneuvers and acceleration to get out. The article on ALPHA noted that it would fire missiles, presumably single shots, just to force a firing opponent into disadvantageous geometry and kinematic profile. One on one, that probably does not do much, but in a group setting, it could be quite successful, especially for the larger group.

So initial detection and counter detection range probably play a big role in whether you can apply energy to escape a shot, and the more energy the AAMs have, the more likely the playing field gets leveled down to mostly countermeasures and numbers rather than individual aircraft performance. Note that it likely does not even matter who detects who first - the AAM launch is likely to be such a visible event that counter detection is extremely likely almost immediately, and software agents seem a lot more inclined to shoot first before evading. So the initial detection range of whoever sees the other first probably sets the engagement range in most encounters, although the lower performance aircraft are incentivized to hold fire and let the range close.

 

[Jemiba]

Interesting discussions, interesting opinions, though the tone somehow has deteriorated …
But all of this got miles away from the original theme ! So, please, back to topic, which is „Chinese drones & CCA thread“, just as a reminder !
(From now, OT posts will be deleted here, without any mention or comment)

 

[lancer21]

Some additional higher res pics of the UADFs.

 

[Josh_TN]

Interesting discussions, interesting opinions, though the tone somehow has deteriorated …
But all of this got miles away from the original theme ! So, please, back to topic, which is „Chinese drones & CCA thread“, just as a reminder !
(From now, OT posts will be deleted here, without any mention or comment)

It might be worth organizing some of the posts above into a CCA/AI tactics and doctrine thread; there was some interesting stuff I had not heard of before.

 

[SspewixzzZ]

Hes not talking about being able to lock on.
Though in modrn radars, theres still a small penalty for locking receding targets anyway.

Hes talking about the missiles kinematics. If launched at a cold target, your missile has longer to travel because its trying to catch up to the target. As long as the target is not inside the no escape zone, the target can maneuver and force the missile to maneuver with it and the missile eventually loses enough speed and cant catch up. Even a mach 5 missile isnt going to catch a lot of cold target its fired at thanks to this.

However, if you are in the no escape zone, that means kinematically, you have no way of forcing the missile to lose enough energy not to hit you. Then you'd pray that the missile either wont acquire you, or acquires the wrong targets (counter measures of various kinds). With dual pulse motors and air breathing missiles though, the NEZ is much more expanded compared to earlier missiles.

Thank you for the explanation.
However, I think there are too many factors affecting the effectiveness of CCA launched AAM — many of which remain entirely unknown, such as their radars, target acquisition algorithms,AAM itself and others — that discussing them could turn into a lengthy discussion.

But I do have significant doubts about the cruise capabilities of these CCA. In particular, I'm skeptical whether relying on a single WS-10 engine can truly allow them to keep up with the J-20, which is renowned for its high-speed performance, let alone J-36 which seemingly even faster.