Chinese UAV/UCAV development

I think you misunderstood what I meant Bltizo.

I do not think the Dark Sword is meant to be a BVR drone. And I think BVR drones will not likely be all that useful against top of the range manned fighters, especially not 5th gen ones.

With the proliferation of stealth technology, I can see the importance of BVR diminishing, especially against stealth-optimised small drones that do not emit active radar until they are in WVR range.

As I mentioned before, I think the technology for a remote controlled WVR drone is already within reach, and if serious work began today, you may have an operational UCAV before the end of the decade.

Such UCAVs will reap the maximum benefits of UAV if they are designed for WVR, since that means they can be made cheaper as they will not need a top of the range AESA radar; they will have no pilot, so will not be limited to 9Gs like manned planes, and that advantage will be most telling in WVR; WVR is also the most dangerous part of air combat, so you would expect higher losses in return for more kills. UCAVs are far more expendable than manned assets, so who better to send into higher-risk and heavy casualty missions?

As for where such UCAVs will fit in the PLAAF's force structure, well I would think that they would be their primary anti-5th gen weapon if such a system becomes operational, and will go in after the PLAAF BVR missiles, but before the manned assets.

In a way, they will do the job J7s would have done if war broke out across the straits a few years ago - they will move in as the first wave of fighters, under cover of BVR missiles from friendly high-end fighters, and engage the enemy fighters in WVR to eliminate, or at the very least, soften them up to allow your own high-end fighters to mop up with minimal losses.

^ Ah apologies, I somehow didn't take in the "limited to no BVR capablities" part.

Reading over again, I suppose my main question is whether a ucav can have the situational awareness of a manned fighter, even if it was remote controlled and had say an EODAS like system... or would it be easier to preprogramme a set of "manouevers" into a drone when it comes into WVR combat?

There is no question that a UCAV can have far superior situational awareness compared to a pilot as it can have a continuos 360 degree view of the entire world around it, while a pilot's situational awareness is limited by his field of vision.

The F35 already offers avionics to try and improve the pilot's situational awareness by having the computer project an image of a target onto the pilot's HMD even if the enemy is outside of the pilot's field of view, for example, if the enemy was below the plane, the HMD would display the image of the enemy fighter as if the pilot could see through the floor of his cockpit.

There are still limits to the system imposed by the pilot's physical limitations, so if the enemy was behind and below the plane, even though the same avionics system can track it, it would not be able to present the information to the pilot in a way he would understand other than as an arrow to where the enemy is. But to take full stock of the situation and devise tactics to use against the enemy, the pilot would first need to turn his plane around to where the enemy is within his field of view. That is an example of a limitation a fully automatous UCAV will not have to face.

Personally, I think that there are just too many possible variables and permutations to write pre-programmed responses. Computers are not very good at making any sort of adjustments or modifications on the fly, so even the slightest deviation for the programmed response might be enough to throw them off.

Computers are able to beat humans at chess because there are so many rules and only so many possible moves you can make, and a computer and use brute force to number crunch all possible choices of itself and the human response to each to find the winning solution. But take a game with fewer rules and more possible variations, like the Chinese game of Go, and you can quickly see how even the best computers will loose to even decent human players, never mind the world's best.

Go is still only a 2D game, so if a computer cannot beat a human in 2D, what are the odds of it doing so in 3D? So I really struggle to see how current computer and software will have much of a chance against human pilots.

What more, computers cannot adapt without someone re-writing their software, whereas human pilots can. So even if you do design a few amazing tricks into your UCAVs, I can see human pilots learning that and its counters very quickly in combat, whereas the UCAV will not be able to adapt anywhere near as fast to new tactics and moves human pilots might develop.

With a remote controlled UCAV, the only issue will be one of bandwidth. If you can get the data rate fast enough, it will be pretty much as if the pilot was inside the cockpit himself. And I see that as a far easier challenge to overcome than to be able to programme computers to think, adapt and learn, as you are into true AI territory if you can make computers do that.

The alternative is to literally train a computer the same way one would a regular pilot. I have read of drones tested with learning computers, that had to go through all the steps, from learning how to take off, to basic maneuvers. The advantages are obvious. Once taught, you can back-up their programs so you retain their experience should they die, and also really copy and paste experienced pilots as many times as needed.

kyanges said:

The alternative is to literally train a computer the same way one would a regular pilot. I have read of drones tested with learning computers, that had to go through all the steps, from learning how to take off, to basic maneuvers. The advantages are obvious. Once taught, you can back-up their programs so you retain their experience should they die, and also really copy and paste experienced pilots as many times as needed.

Click to expand...

Computers are nowhere near as adaptable as humans (or even most animals) at present. You will have to literally tell the computer what to do in every conceivable scenario with every conceivable variable account for. Obviously you cannot literally program the thing to respond to every possible variation, as there are pretty much infinite numbers of possibilities. This means that the best you can do is teach it how to react to likely moves and counter-moves and so on.

Just that is a massive task, and if we leave aside the obvious problem that ultimately, you are training these UCAVs to best counter your own pilots, and even if you managed to come up with a good enough program for the UCAV to be competitive against a human pilot, every single one of your UCAVs will do exactly the same thing when faced with the same situation. A good human pilot will be able to learn this, and damn quickly and it will only be a matter of time before someone stumbles upon a course of action the programmers haven't thought of, or came up with a winning combo of his/her own based on how he knows a UCAV will react to a certain maneuver.

Once that happens, all enemy combat pilots will learn this new strategy and adopt it almost instantly as a single briefing or memo would be enough to get the word out. As soon as that happens, all of your UCAVs would become easy meat and will pretty much need to be grounded till new programming can be done to address the new enemy tactics. This could easily take weeks or months or even years. And even if you do adapt, it is possible that the enemy would find another maneuver to beat your UCAVs and you will have to do the whole thing again.

Until computers are able to organically adapt to a real world situation and independently formulate their own solutions to a new situation, all your automatous UCAVs will be clones of each other, and they will not be able to adapt to new tactics and so there is a big risk that they will be rendered useless by an enemy every time they engage in combat.

That kind of adaptive learning what exactly what I was referring to having read about. No one programmed the drone on how to take off, they just let it work up to it in baby steps. I'll have to find the material on it again and reread it to make sure I'm not recalling it incorrectly. It has been a few years.

I'm thinking about the idea of pairing up the likes of J-20 with Anjian.

Maybe we could have a two seater J-20 version where the pilot controls the J-20 and the co-pilot can control Anjian alongside it. Both will be optimised for LO.

Whereas J-20 can be utilized for BVR combat, as enemy approaches closer range the anjian can start to play a more important role with its bag of close range missiles or even completely take over the action while the J-20 can be on standby or even allowed time to retreat.

The single seater J-20 can still be used for deep strike missions.

Also what's the potential of one or even multiple anjians being controlled from and acts as body guards for AWACs. They can even protect other important assets such as tankers or dispatched to help out local hot spots with high speed dashes. On landing their control will be transferred to ground units.

Landing can certainly be automated. The Soviet space shuttle Buran landed autonomously a quarter of a century ago. That's a very long time in computer and sensor development.

about UCAC/UAV :

it would be pointless to teach them to do better than what human do best (e.g decision making), instead they should be tasked with what human can't do well (like loitering for hours or sorting and generating data)

it would also be pointless to design expensive UCAV to replace manned fighters, instead they both should work in complement, with that said i hope Anjian is a cheap+light weight+agile+stealthy missile lobber (no more than 2-4 missiles) companion for manned fighters, with built in EW/ECM perhaps, say around 15-20 mil a pop

and no, J-20 will not do that UAV controlling thing, it can/will cooperate or work with UAV/UCAV in the future but direct control is too much work for pilot, AWACS/ground control suits the job better

Or perhaps in a two-seater J-20, the radar/munitions operator can take over control of the UCAV in the event of transmissions from ground control and AWACs being jammed.