Skill level of an automated sentry gun?

[Anthony]

Originally Posted by JackFalcon13

I think the issue with these is that the real difficulty is target recognition. People miss at shooting because their hands aren't steady, or they jerk the trigger, or they didn't dial in the range on their scope properly, or they didn't lead the target enough.

People also miss because they're shooting at shadows. People miss at 10 meters, and there isn't any fancy math required to hit at that range.

[DouglasCole]

I saw a video where an enterprising lad worked up one of these with a full-auto paintball gun . . . and his brother walked into the test zone with predictable results.

[evileeyore]

Originally Posted by The Colonel

As I understand it the current issue with making sentry guns is to do with stopping them from shooting things they are not meant to as opposed to them not being able to hit things - theoretical I know since no-one has actually built one as far as I know...

Depends on what we're classifying as a 'sentry gun'.

I'd certainly put Cemetary Guns and ye olde rigged shotgun at the door into the classification of 'sentry gun'.

[weby]

Originally Posted by Anthony

People also miss because they're shooting at shadows. People miss at 10 meters, and there isn't any fancy math required to hit at that range.

A lot of that missing is due to the stress. Machines will not feel stress(unless we get real full AI).

[Rupert]

Originally Posted by The Colonel

As I understand it the current issue with making sentry guns is to do with stopping them from shooting things they are not meant to as opposed to them not being able to hit things - theoretical I know since no-one has actually built one as far as I know (although I think the Israelites have a remote turret system for AFV that comes close, but still with a human gunner who actually authorizes the fire). The issue cropped up somewhere around the recent "killer robot" panic...

The Phalanx, Goalkeeper, and similar anti-missile point-defence systems used on warships are exactly this. They look for things that match their target criteria (incoming, between X and Y radar cross-section, at between U & V speed), and then they shoot them. Because the targets are small, fast, usually detected only when they're fairly close, and the results of missing are dire, they also use a lot of ammo per target to try to ensure hits.

Note that even though, aside from the speed and short engagement range, these system have a fairly simple job, and use a very high rate of fire cannon, they do not have a 100% success rate. They also aren't perfect at selecting targets (despite the fairly clean environment and simple criteria), leaving the designers and users with the problem of having to choose between 'prone to false positives' (wasting ammo, possibly killing civilians, possibly running the system dry just before a real attack), and 'prone to false negatives' (your ship gets sunk).

Sentry guns intended for use on land against infantry, aliens with acid for blood, and annoying local wildlife have to work in a much more cluttered environment, and will probably be expected to be more economical with ammunition. They are also more likely to be around potential friendly or neutral things that look quite like the things they're supposed to shoot. Even those only employed in 'free fire zones' will be an embarrassment if they shoot up stray goats too often.

Overall, while the targetting solution is fairly trivial on roughly human-sized targets moving at human speeds at normal small arms engagements ranges, the shoot/don't shoot problem is not.

[Anthony]

Originally Posted by weby

A lot of that missing is due to the stress.

That's kind of overestimated. Yes, people take a lot of low probability shots in combat, but it's not stress that makes them do that, it's the fact that waiting around for a high probability shot greatly increases the odds of being shot.

[The Colonel]

Originally Posted by JackFalcon13

Samsung makes an automated turret deployed in the Korean DMZ. The test video I saw of it shooting puts a long burst of full-auto into a heart sized area at a decent range, maybe 50 yards or so. Don't worry - they claim it's smart enough to tell you to surrender first.

Didn't know that the technology had actually been built, let alone deployed, but that level of accuracy doesn't surprise me at all.

Originally Posted by evileeyore

Depends on what we're classifying as a 'sentry gun'.

I'd certainly put Cemetary Guns and ye olde rigged shotgun at the door into the classification of 'sentry gun'.

I'd suggest that, lacking active targeting, they are closer to a form of directional mine.

Originally Posted by Rupert

The Phalanx, Goalkeeper, and similar anti-missile point-defence systems used on warships are exactly this. They look for things that match their target criteria (incoming, between X and Y radar cross-section, at between U & V speed), and then they shoot them. Because the targets are small, fast, usually detected only when they're fairly close, and the results of missing are dire, they also use a lot of ammo per target to try to ensure hits.

Note that even though, aside from the speed and short engagement range, these system have a fairly simple job, and use a very high rate of fire cannon, they do not have a 100% success rate. They also aren't perfect at selecting targets (despite the fairly clean environment and simple criteria), leaving the designers and users with the problem of having to choose between 'prone to false positives' (wasting ammo, possibly killing civilians, possibly running the system dry just before a real attack), and 'prone to false negatives' (your ship gets sunk).

Sentry guns intended for use on land against infantry, aliens with acid for blood, and annoying local wildlife have to work in a much more cluttered environment, and will probably be expected to be more economical with ammunition. They are also more likely to be around potential friendly or neutral things that look quite like the things they're supposed to shoot. Even those only employed in 'free fire zones' will be an embarrassment if they shoot up stray goats too often.

Overall, while the targetting solution is fairly trivial on roughly human-sized targets moving at human speeds at normal small arms engagements ranges, the shoot/don't shoot problem is not.

Hadn't considered point defence guns in the category, but I suppose they fit - as for the shoot/don't shoot problem, I recall a rumour from Gulf War 1 that someone tried to use one to shoot down an Iraqi shore launched anti-ship missile and instead watched it hose down the battleship they were escorting...

[JackFalcon13]

Originally Posted by The Colonel

Didn't know that the technology had actually been built, let alone deployed, but that level of accuracy doesn't surprise me at all.

Kind of scary, considering it was made over a decade ago. Plus, they made my phone. There's some good videos of it available, though, like anything fun, it's actual abilities are classified. Pretty sure Samsung will be the first cyberpunk megacorp.

There's a more recent video I saw, but can't find now because it looks like Youtube took it down, of a CROWS system with a .50 cal attached that automatically tracks to a downed opponent and, either autonomously or because the operator had the trigger held down, just engaged him with a few seconds of fire. It's not pretty, even in IR. The scary part is it clearly identified both the non-moving body and a wounded soldier crawling away with targeting brackets. Even if there's a human operator, he's probably conditioned to shoot at the boxes.

[JackFalcon13]

Originally Posted by Rupert

The Phalanx, Goalkeeper, and similar anti-missile point-defence systems used on warships are exactly this. They look for things that match their target criteria (incoming, between X and Y radar cross-section, at between U & V speed), and then they shoot them. Because the targets are small, fast, usually detected only when they're fairly close, and the results of missing are dire, they also use a lot of ammo per target to try to ensure hits.

My understanding is that it's not that they don't hit, just that they aren't fast enough to hit the target within a range where the ship won't be damaged anyway by the blast. That's why Phalanx is being phased out for longer ranged solutions. I think Goalkeeper, using a 30mm based around the A-10 gun rather than a 20mm based around the Vulcan, has a better track record. In testing, obviously, I don't think either are battleproven. The actual intercept isn't the problem, it's doing it fast enough. And let's be honest, a bad guy running at you is going to give you a lot more time than a missile at near hypersonic speed.

I bet if we put a quad Bofors on a ship with the guidance of a Phalanx and a fast enough tracking turret, it would do well. Problem is, it would weigh way more than the equivalent in missiles. The RIM-116 launcher weighs half what the Goalkeeper system does and gives you 21 missiles with a near certain hit at far longer ranges than a gun does.

[Rupert]

Originally Posted by JackFalcon13

My understanding is that it's not that they don't hit, just that they aren't fast enough to hit the target within a range where the ship won't be damaged anyway by the blast. That's why Phalanx is being phased out for longer ranged solutions. I think Goalkeeper, using a 30mm based around the A-10 gun rather than a 20mm based around the Vulcan, has a better track record. In testing, obviously, I don't think either are battleproven. The actual intercept isn't the problem, it's doing it fast enough. And let's be honest, a bad guy running at you is going to give you a lot more time than a missile at near hypersonic speed.

I bet if we put a quad Bofors on a ship with the guidance of a Phalanx and a fast enough tracking turret, it would do well. Problem is, it would weigh way more than the equivalent in missiles. The RIM-116 launcher weighs half what the Goalkeeper system does and gives you 21 missiles with a near certain hit at far longer ranges than a gun does.

No, the high rate of fire is needed to allow for inaccuracies in the firing solution and the translation of that to bullets in the air. The radar's resolution isn't perfect, there's some slop and error in the turret's training mechanisms, and the stabilisation isn't perfect. It all adds up, and even a 'big' Russian missile isn't a huge target (and the big Russian missiles are very fast, so you can't afford to miss the first time).