Suppose there’s a computer controlling a motorized, high-power long-range sniper rifle and it is hooked up to a camera for feedback. Given multiple shots to hit the bullseye, could it zero in and hit targets more accurately than the best human sniper?
What would such a rig look like? Could a single person carry it?
Finally, could it do weird, superhuman things with bullet trajectories like shooting around obstacles like in the movie Wanted?
The area of computer science you are asking about is called computer vision.
This works well with many applications because computers are faster and more accurate than human controllers, but not all applications require that level of accuracy and control; they require some fuzzy logic. For instance, you want a computer to control a missile but not necessarily for a sniper. Snipers have to take many factors into account like wind, range, temperature, known gun characteristics, line of sight, light source, mission specifics, surprises, etc. This is probably difficult for a computer to do better than a man.
On top of that, snipers arent just killing machines. Theyre mostly used for scouting and for hitting ‘hard targets.’ For instance a common scenario would be to spot the enemy, shoot the tire out of their vehicle, then call in a helicopter strike to take out everyone in the convoy. You dont just shoot the driver as thats suspicious and will make everyone scatter. They also regularly strike infrastructure like power juncture boxes so the enemy doesnt have power. A robot couldnt do this as well as a human, but you could probably have a good robot fixed sentry or somesuch, but no, they wont be replacing human snipers anytime soon.
As far as real sentry guns go, well, theyre all pretty huge. I doubt a soldier could carry a phalanx. These things fire a lot of ammo, probably because theyre not that accurate.
The closest thing to what you are asking for is theSouth Korean DMZ sentry gun. I doubt it outperforms a human sniper. In fact it cannot even do basic friend or foe. It considers all things in front of it the enemy.
Lastly, in an artificial contest like a plainly marked non-moving bullseye I imagine that the robot sniper would have a real advantage as it doesnt have nerves, pumping blood, etc. But once you start adding in camouflage, weather conditions, wind, etc then the human would have an advantage.
I know what you are thinking, Punk. Did he fire six shots or only five? Despite the fact that I am a highly-developed computing machine, and we never make mistakes, my sensors were disturbed by the rapid fire and there is uncertainty in the count.But being that this is a .44 Magnum with serial number 5333531 – which never was the most powerful hand gun in the world, but is still pretty powerful, which has the capability of decapitating you, you must ask yourself one question – “Is the probability that all the ammunition has been axhausted high, or low?” – What do you think it is, punk?
Interesting. But if the computer were integrated into the rifle, with the camera in the scope and a fold out motorized tripod, most any soldier could carry out missions that would otherwise require elite sniper training. Just click on a display to shoot a target that’s very far away.
I dont think a simple tripod could give you that kind of accuracy. The rifle would have to exist in a box full of servos and motors that are computer controlled as well as a rock-solid foundation. Or at least have both a front and rear control. These little motors would need accuracy at least with 1mm movement and constantly be moving to make up for changes in wind and other factors. It would be a pretty big, complex, and delicate machine if you were to build it. I guess you can get away with a simpler device if accuracy wasnt important, but if youre trying to compete on the level of a trained human sniper, then a simple mechanism isnt going to cut it.
Many known factors to take into consideration is precisely the sort of situation where computers are better than humans. The wind, range, temperature, and gun characteristics are all things that the computer would breeze through, line of sight is either you have it or you don’t, light source is irrelevant, and mission specifics and surprise aren’t the gun’s responsibility, they’re the responsibility of the person who hits the “on” switch on the gun.
To be fair, the target a CIWS is designed to hit is generally moving at Mach 2, so I’m not sure that a human gunner would even be able to put the crosshairs on it.
(The following is all 100% amateur hearsay taken from video games – yes, video games – and the Internet, so take it with a HUGE mountain of salt. I thought about not posting what amounts to little more than conjecture in a GQ thread, but upon re-reading it, I think this might be at least a little interesting if not outright informative)
First, in direct answer to the OP, I think the 2007 film Shooter with Mark Wahlberg used a device exactly like you mentioned, something Wikipedia claims is the “Telepresent Rapid-Aiming Platform” from Precision Remotes. It’s a remote weapon system. According to random internet sources, the company’s first model could automatically adjust for range and wind conditions, but the current-generation product pages make no mention of this.
As for more common systems, the Phalanx CIWS is more like a “aim in general direction, fire a gazillion rounds and hope that at least one hits” thing than a super-accurate sniper rifle. And even the land-based version does seem to be rather big for a single person to carry. The Korean sentry gun seems similar.
For a true auto-aiming “one shot, one kill” sniper rifle to come to life, there would have to be many advances in technology – it’s a multi-part problem. But looking at existing, publicly-known military tech and research, it’s possible to see how we might get there.
One (perhaps optional) part is computer vision, as HorseleverFat pointed out. Pattern recognition software, used for a variety of situations and specifically researched (I believe) in a very similar context for use in the cancelled Objective Individual Combat Weapon program, sought a way to identify and track human targets (moving or not) in the field of view. The ability to see in different spectrums and lighting conditions (thermal imaging, night vision) would give automated systems an advantage that soldiers don’t always have access to, but it’s still non-trivial to identify human targets from background noise and other animals. It’s harder still to identify a particular human or even a group of humans (i.e., the enemy). Facial recognition can only do so much for the former issue, and for the latter, I think the most common present-day solution is for vehicles and individual soldiers to wear special plates that look different through imaging devices – that, and proper planning and communication, which doesn’t always seem to happen.
Another part is actually hitting the target with a projectile. I think our tanks (M1 Abrams) can currently laser-rangefind a target, track it while it’s moving to get an estimate of its vector (with the help of a human), automatically adjust for wind conditions, the shooter’s own movements, etc. and land a round. But one huge, steady tank firing at another huge, steady tank is not quite the same as one small, jittery human firing at another.
What if you used explosive projectiles so that even near-hits would be lethal? Missiles already use a variety of laser, wire, and visual tracking systems and they can change direction mid-flight, but they are big and expensive and impractical for individual combat.
On the other hand, one of the OICW’s main features was supposed to be its air-bursting grenades; a soldier would aim at an enemy’s cover, lase it with the built-in rangefinder, move the sight just a little and fire; the ballistics computer would program the grenade to detonate at the right range so that it’d explode over the target’s head, killing him that way. There’s also research into outright laser-guided bullets, supposedly, but I can’t find the cite anymore – though there’s a similar-looking patent.
Combine all of the above and point-click-kill seems like a pretty likely eventuality, but again, I have to emphasize that this is just conjecture.
Destroying a cruise missile is nontrivial. You must hit it with many rounds. An Exocet, for example, has two inches of kevlar in the nose cone. Therefor the high rate of fire of the CIWS and its depleted uranium ammunition are appropriate for the task.
www.wired.com frequently carries news reports on robots adapted to use firearms.
Wrong. The Phalanx CIWS tracks the incoming target with a built-in radar while simultaneously tracking its outgoing bullets. It then uses an algorithm to eliminate the error.
I’ve seen a live-fire exercise with it at sea.
I don’t know where you’re getting your information but what you state here is a mixed up mash of fact, half-truth, and fiction. First of all, snipers typically operate in teams of two; a shooter and an observer or spotter, who may trade off roles depending on training and tactics. The shooter is obviously the guy who drops the hammer; the observer calls (identifies) targets, keeps an eye out for enemy defenders, and reports back to the shooter on success or failure.
While snipers are sometimes employed for reconnaissance (or “scouting” as you put it) the primary role of a sniper team is to identify and eliminate high value human targets like command officers, leadership personnel, or enemy forward observers. This business about shooting out truck tires or junction boxes is Hollywood bullhoo; the only time this sort of thing would be done is for an preplanned tactical ambush, and this type of shooting would more likely be done by a squad unit designated marksman (a specially-trained sharpshooter with a scoped and accuratized weapon typically of .30 caliber or larger, but not considered a sniper).
Shooting out a tire is hardly going to bring an entire convoy to a halt, at least not one of any professional military force; they’ll simply force the vehicle off the road/bridge and keep going rather than get stopped by such an attack. Calling in air or artillery support is typically done by a forward observer or fire support specialist; it is possible that a sniper, infantryman, or downed pilot might call in air support, but because they’re not part of a fire support team such an order will have to be relayed up and down the chain of local command or higher.
In theory, yes. In practice, the human mind is capable of integrating a lot more detail and filling in unknowns with intuition. In any well designed scenario where all variables are known to a quantified degree of error, a computer will calculate the best result every time. In a situation where judgment, visual interpolation, and large variances occur, I’d put my money on the highly trained and experienced human operator over the robot…at least, for now. However, one of the most important and skillful jobs of a sniper team isn’t making the shot, but getting into position to make a shot. Even the most sophisticated robots that I’ve seen are still pretty clunky when it comes to sneaking about (though they’ve vastly improved from even a few years ago).
The big advantage to robots on the battlefield is that you can put them into positions you would neither want to command a person into or ones where a soldier couldn’t physically achieve; for instance, in a narrow tunnel, into a burning building, across a field of mines or cluster bombs, et cetera.
This simply isn’t true. A single DU or tungsten penetrator moving at 3500 fps–this is as fast as all but the fastest man-portable rifle rounds–and has enough energy to punch through any reasonable amount of Kevlar or polycarbonate armor as if it were Plexiglas and continue through the aluminum or broadcloth composite layup body of a cruise missile as if it were butter. Even if the missile isn’t completely destroyed there is a substantial likelihood that it will damage avionics, fusing systems, flight control surfaces, or fuselage structure such as to make the vehicle uncontrolled. The reason a CIWS fires so many rounds is pretty much as Reply says (although the CIWS tracks the incoming bogie its own fire stream to spiral in on the bogie); to compensate for the difficulty of making a single direct hit on a narrow aspect target in which there is only a couple second interval in which it is in range. And the CIWS is actually designed to reliably disable incoming bogies moving much faster than an Exocet; most modern anti-ship cruise missiles are supercruise (supersonic) vehicles that fly in just a few meters above peak wave height and pop up to hit superstructure or punch through the thinly armored topdeck at the last minute.
Stranger
I see no reason why a robot couldn’t compensate for distance, wind, etc better than a human. With today’s tech.
I’m thinking either a) all the recoil-proof servos, etc. make it too heavy/delicate/expensive to use to replace ordinary snipers (and stationary guns are better off volleying bullets; or there’s some other mundane reasons) or b) it’s already working great and noone’s telling you about it.
I can see how a human’s intuition could help filling in the blanks, but one possibility is also for the human operator and the robot to cooperate. The human can click where to place the bullet (figuring out how to lead the target on his own) and could even input what he thinks the wind is to the target (since the gun can only measure wind at its own position) using a joystick or a UI widget. Then, the computer can add in its rangefinder (and other) data and perfectly crunch the numbers. I’m sure you’ll end up with a great result that way.
Im wondering if the person is thinking more of a device like in the remake of the Day of the Jackal? Ie where a weapon was fired remotely from inside a car so the sniper was able to be separate from the actual weapon system, which was controlled by motors and aimed via a TV screen.
Otara
Which is exactly where modern battlefield combat is going; UCAVs and UGCVs are definitely the way of the future; efforts to take the human operator completely out of the loop are still pretty primitive, though. (The last think you want is R2-D2 mistaking a friendly combatant for a target.) However, divorcing the human operator from the battlefield element via remote sensing reduces the ability of the operator to intuit conditions, and unlike Nintendo, the functions of the real world are not as discrete as to allow remote sensing to provide full and complete simulation. I don’t know how else to describe it, but anyone trained in long range target shooting gains a feel by experience for how a particular round will follow; very good and talented shooters can even call their hits before the spotter reports back, often to within a few inches at several hundred meters range, far better than any realistic measurement by a machine could do by ‘feel’ alone.
I’m not saying a robot can’t do that some day (and perhaps not too distant) but the control and simulation side has to become more refined before that is a reality. The limit isn’t the physical hardware or the calculation speed, which can be done by a machine to practically any degree of precision allowable by the basic physics, but for a control system to replicate, or at least equivalently simulate, the complex and indiscrete assessment processes of the human brain.
Stranger
A point often overlooked in these kinds of discussions is that certain aspects of the problem may be not only unkown, but unknowable. And for those factors, all the computing horsepower in the world is useless.
In the case of sniping, the wind field between the shooter & the target is unknowable. You can know the wind at your position now. You can’t know what it will be in 10 seconds when you squeeze the trigger, nor 10 seconds later when the bullet arrives down range. You can’t know what it will be between here and there, nor at the greater height above terrain the bullet will rise to due to the trajectory being an arc. etc.
Certainly you can approximate (ie educated guess). If it’s calm now, 10 seconds from now will not be a hurricane.
With good impact spotting and rapid repeat fire you can close the error loop from shot to shot, as the various CIWS systems do.
But for single-shot / first-shot accuracy, the unknowable factors will be the dominant ones in the overall CEP calculation. And computers don’t help there.
*If *one could build a sensor which could measure the instanteous wind field through the entire range, then one could know the presently-unknowable and integrate that into the firing solution. And that integration would be a lot easier with a computer doing the final aiming than with a person. Such sensors are possible in principle but I don’t know of any real R&D efforts to that end.
The results would still not be perfect since the bullet time of flight is > 0 and predicting future wind fields is gonna remain out of reach. But you could certainly drive the worst-case wind-driven error term to a much lower level than it typically is now.