Recent UAV/UCAV/UCAS' and Autonomy

[KJ_Lesnick]

I know they can be remotely controlled or operate autonomously...

How autonomous are recent UAV's, UCAV's or UCAS's? Like what are they designed to do autonomously? Can they work as good as a human being in all respects?


KJ Lesnick

 

[Jemiba]

I think, they can follow there flight path from waypoint to waypoint, and I've read that,
as some UAVs (like global Hawk)are intended to operate in civil airspace, too, they can be
fitted with TCAS and react by themselves on collision warnings. So a UAV on a recce mission
wouldn't need to be remotely controlled. On an attack mission, I think, the UAV, once in
the target area would be remotely controlled only, as probably the "rules of engagement"
of every airforce only allow an attack, with "man in the loop". And judging what I've read
about artificial intelligence, that will remain so for quite a time !

 

[Antonio]

Hi KJ,

that's a very exciting question. I have very little information about the subject so any post will be welcome.

From what I see on the media (http://www.airforce-magazine.com/MagazineArchive/Pages/2009/January%202009/0109UAV.aspx) UAV seem to be basicaly remotely controled machines.

Building a machine with the intelectual capability of a human being it is not an easy task. We have very powerful computers but "thinking" is a more complex task than "crushing numbers". In fact, first of all we still haven't a precise idea about how a brain works thus we can't imitate it. Given the complexity of human brain, scientist research is focused on more simple nervous systems like that on bugs (even bacteria show what we name "behaviour"). We can understand bugs and later try to understand humans . I think the Mars Pathfinder works that way. But an autonomous UCAV needs a level of intelligence much higher than that Mars robot explorer. So I think such an aircraft is not going to be developed in the near future. In my opinion we'll have to wait a two or three decades.

 

[KJ_Lesnick]

Pometablava,

Not to mention there is a danger associated with having a completely autonomous system with the intellectual capacity of a human being even if it can be done. If it were to turn on it's master, it would be very dangerous.


KJ Lesnick

 

[Jemiba]

"If it were to turn on it's master, it would be very dangerous."

That maybe a real danger for friendly forces without a correctly working
IFF, but I think, today the possibility of an autonomously operating UAV hitting,
e.g. a kindergarten, because its monkey bars were misinterpreted as missile
launchers by the UAV is regarded as much more dangerous.
Politically only, of course ....

 

[Just call me Ray]

When you think about it, UAVs don't need a lot of processing power to be truly autonomous.

All they need is a good GPS reciever and the ability to follow waypoints/avoid targets. Just take-off, follow the GPS, drop bombs at a designated waypoint, then come back.

 

[Abraham Gubler]

The NASA X-45 trials demonstrated a considerably higher level of autonomy than mentioned here.

http://www.boeing.com/news/releases/2005/q1/nr_050214s.html

 

[sferrin]

Pometablava,

Not to mention there is a danger associated with having a completely autonomous system with the intellectual capacity of a human being even if it can be done. If it were to turn on it's master, it would be very dangerous.


KJ Lesnick

Until it ran out of gas or the batteries died. :

 

[Antonio]

When you think about it, UAVs don't need a lot of processing power to be truly autonomous.

Just take-off, follow the GPS, drop bombs at a designated waypoint, then come back.

I agree with you about defining "autonomous" this way. In fact, the UCAV you're describing is practicable today. But I think it's behaviour is too predictable and thus not survibable out of this simple task of carrying a bomb from point A to B. Besides, as Jemiba noted, it's not a flexible weapon for that mediatic wars of the XXI Century. I mean if an UCAV has to be operated in a highly cooperative environment such a modern combat force it needs to be much more "autonomous".

 

[Antonio]

Thanks Abraham

military aircraft capable of flying over enemy territory, ready to strike and destroy surprise threats immediately

the X-45As autonomously determined which vehicle held the optimum position, weapons and fuel load to properly attack the target. After making that decision, one of the X-45As changed course

Wow, if those robots really can do that...we have a serious achievement. This is what I consider an autonomous machine.


Skynet is approaching

 

[CammNut]

How autonomous are recent UAV's, UCAV's or UCAS's? Like what are they designed to do autonomously? Can they work as good as a human being in all respects?

They do what they are programmed to do by humans. So a human can program them to take off, follow waypoints and land. A human can also program them to react to events by providing them with a matrix of options. ie: detect radar, turn away from radar; or detect radar, attack radar, etc. The X-45As did everything automatically except drop weapons, for which they needed permission from a human in the command loop. But by automatically I mean they followed their programming.

If you search the web you will find the DoD's definition of the levels of autonomy. The UCAV was pretty high up, but well short of the autonomy military operators feel they need to reduce the manpower required to operate UAVs. The most autonomous UAV of recent times was to have been Darpa's UCAR, but it was cancelled, although they did some impressive simulations of its ability to fly nap-of-the-earth while operating co-operatively with other manned and unmanned platforms.

 

[Just call me Ray]

I agree with you about defining "autonomous" this way. In fact, the UCAV you're describing is practicable today.

Well...yeah. I was pretty much describing the Global Hawk

But I think it's behaviour is too predictable and thus not survibable out of this simple task of carrying a bomb from point A to B.

Ideally that's where stealth comes in. A commander could assign it waypoints, either pre-flight or during flight, to make it less predictable. It could also terrain-hug, either with the assistance of an onboard radar or (perhaps even preferably) using a GPS command link.

Besides, as Jemiba noted, it's not a flexible weapon for that mediatic wars of the XXI Century. I mean if an UCAV has to be operated in a highly cooperative environment such a modern combat force it needs to be much more "autonomous".

How so? I'm afraid I don't follow. If it needs to loiter, a commanding officer can simply order it to loiter at a waypoint. If it needs to get to a certain location in a hurry, a commanding officer will simply order it to.

Close air support will be a bit tricky, I'll admit. There will have to be some sort of means of visually identifying targets or an ability to illuminate targets by Army grunts or Marines. But this can probably be overcome.

I've outlined what I think should be the target goals of UAV development on here and some other forums before. A small, compact, highly stealthy platform with a very simple flying wing or "2D" planform - perhaps a smaller version of an X-45 or X-47, perhaps even with just a single, low to mid-thrust (but high efficiency) engine like a JD15, and maybe with a warload of no more than 4,000 lbs, with the idea of making cheap "swarming" platforms and reducing the craft's signatures the old-fashioned way of just making it smaller. There would be no air-to-air capability whatsoever for the interest of cost; the idea would be to catch enemy aircraft on the ground or rely on surface-to-air missiles. Of course there would be no reason for the craft to not carry AAMs, but it would have to be with the understanding that the craft is not stressed for high-G flight, so BVR combat would be preferred. It might not even carry an onboard radar, instead relying on a GPS link, and it might be slaved to another radar platform such as AWACS/JSTARS or AEGIS. It might even be programmed to return to base using last known GPS information if that link were to be broken.

The craft would also have a commanding officer who would "pilot" the craft using a stylus on a touch-screen to assign it waypoints and issue fire orders. The commanding officer would also be able to give the craft altitude and speed orders. "Piloting" the craft would probably end up resembling ATC operations more than anything.

The idea would also be to make it as autonomous as possible without a lot of processing power or "AI" behind it. Not only would it be cheaper, but it might also assuage fears of these things going rogue

 

[Jemiba]

As stated before, the answer lies in the definition of "autonomus". A UAV can
take-off on its own fly to a target, accomplish its mission there, fly back and land
while the operator is sleeping in his bed. But if a target of opportunity would appear
beneath its flight path, I think most UAV would complete ignore it. So, I wouldn't
call this real autonomous.
Really autonomous in my view means, for example, flying to a target area, looking what
can be found there, deciding target or non target, chosing an appropriate weapon (e.g.
GBU or hellfire), firing, checking the results, if necessary spend one more weapon and fly
back. That's, what I pilot in, say Afghanistan is doing every day !
Judging, what modern software can do, I think it already could be built into UAVs today.
But, as Ray already mentioned, it's a question of installed processing power, increasing
cost and I would stress again my example with the monkey bars ..

 

[Antonio]

How so? I'm afraid I don't follow

I'm sorry Ray. After reading your last post I realize I'm totally coincident with you. As Jemiba said, the quid lays in the definition of "autonomous" and what degree of autonomy you want for that UCAV.

In my post I was thinking about a 100% degree of autonomy. And, possibly this ability will never be required for a "robot warrior" because there is no such need. Your argument is well balanced because you consider cost which is capital factor.

A living beign has an inbuilt set of instructions (a program) we call instinct (a sort of ROM in computer parlance) while the brain is a processor that can rectify instinct instruction if it doesn't lead to the optimal solution. A computer that could mimic the brain should be very expensive (at least in the near future).

regards,

Antonio

 

[KJ_Lesnick]

Abraham Gubler,

The NASA X-45 trials demonstrated a considerably higher level of autonomy than mentioned here.

http://www.boeing.com/news/releases/2005/q1/nr_050214s.html

military aircraft capable of flying over enemy territory, ready to strike and destroy surprise threats immediately

the X-45As autonomously determined which vehicle held the optimum position, weapons and fuel load to properly attack the target. After making that decision, one of the X-45As changed course

Highly impressive, and totally autonomous if necessary.

BTW: How exactly did it determine fuel-load?

There is something worrisome as I pointed out that if it malfunctioned or for one reason or another decided to attack it's master. Contrary to SFerrin's statement, there are nuclear-powered UCAV's under development which would have plenty of endurance.


sferrin,

Until it ran out of gas or the batteries died. :

Well, they have been proposing nuclear powered UAV's...


KJ Lesnick

 

[Antonio]

there are nuclear-powered UCAV's under development

which program, could you give more details?

 

[Jemiba]

"And, possibly this ability will never be required for a "robot warrior" because
there is no such need"

The real problem with total autonomy is, that you cannot court martial a UAV ... :

 

[Just call me Ray]

As stated before, the answer lies in the definition of "autonomus". A UAV can
take-off on its own fly to a target, accomplish its mission there, fly back and land
while the operator is sleeping in his bed. But if a target of opportunity would appear
beneath its flight path, I think most UAV would complete ignore it. So, I wouldn't
call this real autonomous.

That also brings up another question - is full autonomy even wanted or needed? Do we want the plane itself designating targets or do we want it slaved to an Air Force Captain's touchpad who will ultimately be the one identifying targets and issuing fire orders?

 

[CammNut]

Here's the scale of autonomy levels that was promulgated during the UCAR program to "explain" why it would be a leap ahead of the UCAV...

 

[Just call me Ray]

I'm sorry Ray.

Heh, no need for apologies

 

[Antonio]

 

[Antonio]

Drone compilation

http://www.defensetech.org/archives/cat_drones.html

 

[KJ_Lesnick]

Jemiba,

The real problem with total autonomy is, that you cannot court martial a UAV ... :

Good point

 

[Just call me Ray]

Drone compilation

http://www.defensetech.org/archives/cat_drones.html

Hmmm, that's pretty interesting.


You know what? Just a shot in the dark...is there any way I can get a job on a drone development project?

 

[Jemiba]

"any way I can get a job on a drone development project?"

Don't think so, but in former times, there were people, walking along the beach an collecting
everything worthwhile, debris, etc. Such a job seems to become fashionable again :

" ...although it could be recovered via a beach, "

If this system is used near a beach, I wouldn't think it's a friendly one! So, who should recover
the UAV ? North Korean troops? The Hisbollah or revolutionary guards ?? I hope, that wouldn't
be fitted with booby traps ! ;D

Nevertheless an interesting system, although those drones seem to be small and probably not
having much autonomy.

 

[avatar]

Hey machines are alive too, from their perspective we might be non-sentient objects for all you know.

But seriously fully autonomous machines may turn out to be a major headache , because even not very autonomous machines behave pretty peculiarly at times.

needless to say full autonomy will also create huge political problems during employment. a mistake by one of these would receive far more press attention than desirable.


of course machines becoming self aware from "our" ( human hopefully) point of view is a non sequitur. they are already self aware from their point of view.

 

[Just call me Ray]

"any way I can get a job on a drone development project?"

Don't think so, but in former times, there were people, walking along the beach an collecting
everything worthwhile, debris, etc. Such a job seems to become fashionable again :

Huh, well, that's encouraging :

Either way, maybe I'll just submit a resume to Boeing, Lockheed, NG etc. I don't have a lot of engineering experience, but at least I have some. Maybe those 4 years at aerospace school will amount to something after all.

 

[KJ_Lesnick]

Hey machines are alive too, from their perspective we might be non-sentient objects for all you know.

What causes you to believe they are "alive" and self-aware?

I thought if you were self-aware, had conscious control over your actions, and could communicate your desires and intent you would be classified as sentient?

But seriously fully autonomous machines may turn out to be a major headache , because even not very autonomous machines behave pretty peculiarly at times.

Agreed

of course machines becoming self aware from "our" ( human hopefully) point of view is a non sequitur. they are already self aware from their point of view.

How exactly would they qualify as self-aware?


INRM

 

[donnage99]

All this talks made me remember the incident in which a predator was supposed to bait an Iraqi mig-25 and than run, but it fired the mig-25 on its own anyway with a Stinger and lost.

I remembered reading that the x-47b would have the ability to discriminate whether what it see as friends or foes, flying in group and deciding and work out a strategy once approached by targets such as who will strike, who will provide intelligence, etc.

The UCAR will take it to another level by further enhancing "reasoning skills" and "team work." The human pilot will no longer control the air vehicles, but rather participate in a cooperation with them as a team and oversee their missions from a manned helicopter. The team of UCARs would be able to select among themselves a "leader" for the job, decide whether to avoid or engage threats to itself, or to sacrifice itself to protect the manned aircraft, or protecting friendly targets. As the scenarios enfolding itself, they will have the ability to switch the "leader role" among themselves and replan the missions, depend on the rules of engagement. Relay the informations it gathered back to the manned aircraft, suggest an appropriate strategy, waiting on the pilot to make the final decision. The pilot has x amount of time to approve, or it will be automatically approved. The pilot will talk to the unmanned vehicles, and they talk back, and go as far as to negociate with the pilot so that top priorities will precede lower ones. Using 5 sensors, they will be able to determine concealed and disguised targets, combatants from non-combatants, etc.

 

[Jemiba]

"But seriously fully autonomous machines may turn out to be a major headache"

As all these talks about autonomy basically deal with software, this story may not be too OT:
In a documentation about the making of "Lord of the Rings", it was told, that for the fully animated
large battles scenes a software was used, whith separate software modules for every single figure
on the screen. And of course, those figures, gathered in two groups, would have had to run to the
other group to start the fight. But surprisingly, there always were some figures running in the wrong
direction, not entering the fight at all ... just a bug, or self-awareness and cleverness ? ;D

It's the funny thing with modern software, that even after comprehensive testing, you can never be
absolutely sure about it, as you'll never be able to test all combinations of inputs !

Anybody remembering the Y2K panic ?

 

[Just call me Ray]

As all these talks about autonomy basically deal with software, this story may not be too OT:
In a documentation about the making of "Lord of the Rings", it was told, that for the fully animated
large battles scenes a software was used, whith separate software modules for every single figure
on the screen. And of course, those figures, gathered in two groups, would have had to run to the
other group to start the fight. But surprisingly, there always were some figures running in the wrong
direction, not entering the fight at all ... just a bug, or self-awareness and cleverness ? ;D

In all fairness that's a common bug in game software too. From what I understand it just ends up being a conflict or misinterpretation of instructions for that particular character or object.

If there's anybody really into programming, they might be able to talk more about it.


Also:

The UCAR will take it to another level by further enhancing "reasoning skills" and "team work." The human pilot will no longer control the air vehicles, but rather participate in a cooperation with them as a team and oversee their missions from a manned helicopter. The team of UCARs would be able to select among themselves a "leader" for the job, decide whether to avoid or engage threats to itself, or to sacrifice itself to protect the manned aircraft, or protecting friendly targets. As the scenarios enfolding itself, they will have the ability to switch the "leader role" among themselves and replan the missions, depend on the rules of engagement. Relay the informations it gathered back to the manned aircraft, suggest an appropriate strategy, waiting on the pilot to make the final decision. The pilot has x amount of time to approve, or it will be automatically approved. The pilot will talk to the unmanned vehicles, and they talk back, and go as far as to negociate with the pilot so that top priorities will precede lower ones. Using 5 sensors, they will be able to determine concealed and disguised targets, combatants from non-combatants, etc.

A helicopter? A fixed-wing aircraft that could keep up makes more sense.

 

[KJ_Lesnick]

This could be a major problem if they give these UCAV's too much autonomy.

Technically what they are currently thinking of is already giving the UCAR the ability to make up it's own mind without an instruction (if one isn't given fast enough). What if it somehow due a to a glitch or some other reason decides to ignore all inputs and make it's own decisions completely?

 

[Jemiba]

"The UCAR will take it to another level ..."
As UCAR is mentioned in the future tense, am I wrong,or should it
be "UCAR would have taken it to ..." ?

AFAIK, UCAR is short for "US Cancelled Another Rotorcraft"

;D

 

[avatar]

The best that the "science is God " types can do is program what they think is rationality into a machine they build. Example- "there are ten routes out there , my computation reveals that x is the shortest and safest i.e I will take route x."


. nobody even wants to program humanity into a machine. Nor will A machine become "self aware " by what it "learns" on account of some clever algorithm . if it "seems" self aware then it must have crossed a rubicon (on its own and not due to any programming) that separates what is termed as a "certain point of view". Given that Humans are increasingly unable to do that between themselves,the chances of some machine doing that is slimmer.

 

[donnage99]

The UCAR will take it to another level by further enhancing "reasoning skills" and "team work." The human pilot will no longer control the air vehicles, but rather participate in a cooperation with them as a team and oversee their missions from a manned helicopter. The team of UCARs would be able to select among themselves a "leader" for the job, decide whether to avoid or engage threats to itself, or to sacrifice itself to protect the manned aircraft, or protecting friendly targets. As the scenarios enfolding itself, they will have the ability to switch the "leader role" among themselves and replan the missions, depend on the rules of engagement. Relay the informations it gathered back to the manned aircraft, suggest an appropriate strategy, waiting on the pilot to make the final decision. The pilot has x amount of time to approve, or it will be automatically approved. The pilot will talk to the unmanned vehicles, and they talk back, and go as far as to negociate with the pilot so that top priorities will precede lower ones. Using 5 sensors, they will be able to determine concealed and disguised targets, combatants from non-combatants, etc.

A helicopter? A fixed-wing aircraft that could keep up makes more sense.

What do you mean?

 

[Just call me Ray]

Oh, I take it from Jemiba's explanation that the UCAR is a rotorcraft, in which a helicopter would make sense as a drone controller. I was under the impression UCAR was a fixed-wing program.


That said, I do see a very long-term, lingering need for a high-performance manned tactical fighter aircraft, but before much longer it will really end up being a one or two seat drone control platform with some really impressive (primarily) defensive capability, with maybe an offensive capability thrown in as a "what the heck" measure.

I don't remember if I said this in the F/A-X thread or on What-If Modelers, but in the next 50-75 years we could see about 100 "combat" pilots in the entirety of the USAF. If even that.

 

[KJ_Lesnick]

I didn't know a UCAR was a rotorcraft.

 

[donnage99]

yep, cancelled though. Here's some images. These are northrop and Lockheed designs (boeing and raytheon got kicked out early in the game).

Lockheed proposal

Lockheed Martin, working with Bell, is proposing a compound helicopter with propulsive anti-torque system - a highbypass propulsion system in the tailcone that eliminates the tailrotor while providing forward thrust, giving a dash speed exceeding 180kt (335km/h). Lockheed Martin's UCAR has a gross weight of 2,500kg (5,5001b), payload of more than 317kg and an endurance of over 9.5h.

Northrop proposal

Northrop Grumman, working with Sikorsky and Kaman, is proposing an intermeshing-rotor design that eliminates the tailrotor and delivers all the power to the rotors for lift, says Zwernemann. Derived from Kaman's K-Max external-lift
helicopter, the vehicle has a new rotor aerofoil, blade servo-flap and forward-tilted mast to increase thrust and reduce drag, pushing speed above 160kt. Northrop Grumman's has a gross weight of 2,890kg carrying four Hellfire missiles,
payload of 1,135kg internally or 1,450kg externally, and 10.6h endurance
with auxiliary fuel

 

[yasotay]

The technology on the UCAR program was really advanced for its day. Both teams had some very impressive technology going in the labs and had made great strides in developing compadible technology to work with manned rotorcraft. Both teams had active duty aviators working with them to make the man-machine interface work in combat.

Alas when it came time for the Army to provide matching funds those funds were not available.

 

[donnage99]

If the UCAR was sufficiently funded, it would be the most ambitious program we have ever pursued in the past decade, I think. Artificial intelligence seen in sci-fi conspiracy thriller, along with stealth, advanced rotorcraft technology, etc. Northrop commented that even if the program wouldn't achieve its vision, it still advanced the Army greatly. When news that the Army wanted to shift funding from the program to other stuff, DARPA did seek support from the civilian crowd of the government (since most UAV programs were funded by the DoD instead of the branches of the military) or the Navy, as they shown some interest of the UCAR for their littoral combat ships (they instead chose the fire scout).