AI drone for air superiority?

[jyl]

I have a question about the potential use of drones and AI for air-to-air combat.

Suppose an unmanned aerial platform were developed, that could carry a long range air-to-air missile (something like a modernized version of the AIM-54 Phoenix), that could loiter in an area for a long time (several hours), and that had sufficient AI to operate autonomously if communications with ground controllers fail or are jammed.

Could these be used to control airspace, instead of manned fighter aircraft?. Would it be more effective and less expensive?

I'm thinking our air force could use several of these to patrol an area 24/7. They'd be in contact with ground and airborne radar and operated by pilots on the ground, and would also carry their own radar, IFF, countermeasures and AI. If ground control was disrupted, the AI drone would revert to pre-set instructions, such as "shoot down anything flying in this area that is not verified as friendly".

Is this feasible? Should we have such a program?

I have never quite understood why we don't currently have a long range air-to-air missile. AFAIK, our current air superiority tools all involve getting a manned aircraft relatively close (1 to 20 km) to the enemy aircraft. Manned aircraft are expensive, can't stay airborne very long, and risk pilot loss. 30 years ago, a single F14 Tomcat flying over San Francisco could shoot down any six aircraft flying within a 200 mile radius, from as far as Sacramento to almost as far as Santa Barbara (in theory; I don't believe many AIM-54s were actually fired in anger).

Granted a human pilot can make decisions and visually identify aircraft. But I'll bet an AI can make decisions faster and identify aircraft better. And a human can't visually identify anything that isn't close by, while an AI drone can visually identify at longer range (via optics) and even far beyond visual range (via camera in the missile).

[stomachmonkey]

No such thing as software that is 100% bug / error free.

Even if one got there there is still no such thing as a system that can't be compromised.

Far too much stuff in the air at any given time to risk it.

https://www.theguardian.com/world/ng-interactive/2014/aviation-100-years

[HardDrive]

Its a great idea until a 747 full of passengers who's transponder has a fault gets shot down.

I work on a massive software suite. Even with a deep testing regime, you can still get subtle errors that require an exact set of circumstances that no one could have possibly thought up.

I don't know about flying AI weapons, but in the nearer term, perimeter defense is certainly possible.

[Wetwork]

The Navy is doing the same projects with subs. There's already a bunch of prototypes. The plan is they turn them loose and they find foreign subs and just follow them. Supposedly.. Spooky

Unmanned Sub-Hunter To Begin Test Program

[jyl]

Well, I'm assuming you'd use a weapon like this when you are in a full on war. At that point, there's not going to be any passenger airliners flying around. And if they are, they're going to get shot down by land based air defense or manned aircraft anyway, since presumably United Airlines flight 312 isn't going to be equipped with military IFF.

Theoretically, with enough AI, the missile could visually identify a civilian Airbus and abort on its own.

Yes, the AI could make mistakes. But humans make mistakes all the time.

[svandamme]

The Phoenix was meant to intercept hordes of Badgers, Cruise missiles and what not , coming at the carrier.

It was not designed to take out 1 single target at 200 miles out..
Either way it really did not have such range they only did120 miles/190km
ROE typically requires visual confirmation or at least being fired upon by the target.


The AIM 54 was heavy, could not be trapped on the carrier if the Tomcat had to many of them coming back. And it was expensive to maintain.

So situations where you can actually launch a missile at 200 miles out, so exceptionally rare that they don't warrant the extreme cost involved with such designs and if such range is needed it's probably going to be easier and more cost effective to build a boosted rocket that can fire from land or ship..
Running costs= Next to Nothing.


As for used in anger, the Iranians did make good use of them.. The Iraqi's feared the Tomcat to the point that they would turn and run as soon as they noticed the distinct AWG9 radar.
In US use a few were fired in anger but they missed or misfired.. By then they weren't new anymore

Amraam can get to 85% of the distance that Aim54 could, 160km
And it get's there only a few seconds slower

190km @ mach 5 = 111 secs
160km @ mach 5 = 117 secs

For all intent and purpose at those ranges and speeds what really matters is detection and response time/capabilities by the target.. Amraam can fly inertial guiding and then switch to active homing resulting in very short detectiontimes or switch to home on jamming if it detects it. So that should be interesting.

[RKDinOKC]

Drones are currently used for situational awareness. The information they collect provides data to the command and piloted aircraft. With the capabilities of modern aircraft I would assume piloted aircraft launch their munitions when hostile aircraft that have already been classified as hostile by the battle area command are no more than a blip on their own radar.

There is a description in the book "Snow Crash" where defense systems are integrated with dog intelligence, i.e. guard dogs. Each device can communicate with the others to not only help determine what is or is not a threat, but warns of an intruder approaching their own protected area.

[svandamme]

Originally Posted by RKDinOKC

Drones are currently used for situational awareness. The information they collect provides data to the command and piloted aircraft. With the capabilities of modern aircraft I would assume piloted aircraft launch their munitions when hostile aircraft that have already been classified as hostile by the battle area command are no more than a blip on their own radar.

There is a description in the book "Snow Crash" where defense systems are integrated with dog intelligence, i.e. guard dogs. Each device can communicate with the others to not only help determine what is or is not a threat, but warns of an intruder approaching their own protected area.

Why bother with piloted planes if you already ID'ed the target with drone intel?
Much cheaper to either launch shipboard Long Ranger SAMs or Shoot AA missiles from drones.

But the fact of the matter remains, that for the most part , for more complicated situations, you still need the Mark 1 Eyeball up there, along with the Mark 1 human brain.. Nothing can adapt to changing parameters and make judgement calls like a human being.

And there is such a thing as relying to much on missile tech or even drone tech.
Vietnam taught you guys that lesson.

[svandamme]

Additonally, if drones can do the intel, and drones or Sam's can launch from a distance.
Why even bother with aircraft carriers or manned missile cruisers.

Could just install remote controlled Missile pack units in Shipping containers, and keep a round the clock 24/7/356 rotation on commercial container ships.. captain doesn't even have to know what he has on board. Just find some method to make sure you have enough missile packs circulating the right area at any given time.

And put some AI on that mofo

[flipper35]

Well the AIM-120C has a range of 50-70km and the AIM120D should run about 50% farther. One of the new European missiles is being tested with a ram jet for extended range.

They have already performed some tests with the F35 radar guiding ship launched SAMs and there is also the use of F22/F35s in forward areas feeding info to missile trucks back in the safe zone. A B52 or B1R could carry a large number of missiles to supplement the F22 stores.

All that is in a perfect world and against non-maneuvering targets it works great. So if we have a bunch of Bears/Badgers/Blackjacks on the way we could engage them that far out. Against smaller Fulcrum/Flanker type aircraft that are maneuvering you generally end up where the missile "thinks" the target should be but wouldn't necessarily be there. The AIM-120 does have mid course updates if someone would car to keep the target lit up the entire engagement.

We could use AI drones as missile trucks as well.

The AIM-54 was not used with any luck in the 1991 gulf war. I believe there were ZERO shoot downs with the few that were fired. There were a few AIM-7 hits but most were AIM-9 missiles. There were a couple wins where the enemy flew into terrain.

One F-15E used a paveway to shoot down an Iraqi helicopter.

[Seahawk]

I have spent the last 16 years of my life working DoD and commercial UAS.

My company has teamed with a large defense contractor to build and fly a UAS that was my idea: A UAS that can be taken to the fight on the wing of a jet and launched.

We fly in our first big exercise in August, called Flex 17, Netted Sensors. We are building the air vehicle, the LDC is doing the sensor integration. We will talk directly to an F-18G. We are not an air superiority vehicle, rather a node gathering information.

Netted sensors: That is the future capability all services (the Air Force and Marines are involved) are headed.

UAS, ground and sub surface UV's can do everything well if they know or are taught why.

[Captain Ahab Jr]

I have a friend who is working on AI to control a flock/school of drones so they are totally autonomous.

Creating artificial neuron networks so the drones learn from each other rather than directed by some one in a control center.

Scary but fascinating stuff