Sneaky Sneaky: The Stealth Thread

Alright, stealth in most games is pretty simple. There are only so many variables when you are dealing with pre-defined locations and/or opponents.

We are not dealing with pre-defined locations or opponents.

And in fact, stealth, and more broadly, deception, is one of the most central and one of the most complex parts of life. Virtually every animal in existance uses deception to an extent, and not all of it comes in the form of stealth. So. Time to get to work on figuring this out in a way that will prevent both our computers and Bashi from exploding, or turning into a giant squid of anger in Bashi's case.

To Be Discussed:
Sensors and there effect on deception
Deception through mimicary
Deception through behaivior (Moving stealthy)
Large scale deception (See: Moving of Alexandria Harbor in WWII)
Differentiation and confusion, AKA Taht Zebra thing
Deception through exploitive instincts (Frilled lizards and startled cats are good examples here)

This is a very important aspect that we should discuss. To add my two cents, I think that stealth in space should also be discussed. Anyone ever stopped to think how unrealistic are "sensors" that pretty much every sci-fi work features? What do they sense?
I think that on early space travel (when we stil are relatively "realistic" knowing where ships are and what planets are made off should be a major challenge, and it'd become easier as you progress through the tech tree into scientomagical mechanisms (gravitron sensors anyone? )

Djohaal wrote:

This is a very important aspect that we should discuss. To add my two cents, I think that stealth in space should also be discussed. Anyone ever stopped to think how unrealistic are "sensors" that pretty much every sci-fi work features? What do they sense?
I think that on early space travel (when we stil are relatively "realistic" knowing where ships are and what planets are made off should be a major challenge, and it'd become easier as you progress through the tech tree into scientomagical mechanisms (gravitron sensors anyone? )

Good point. This best detailed depiction of sensors in space I've found in Mass Effect, where they use what is essentially radio-telescopes to detect the massive amounts of radiation released by traveling ships. Hence the stealth systems essentially being massive heat-sinks.

Tenebrarum wrote:

Djohaal wrote:

This is a very important aspect that we should discuss. To add my two cents, I think that stealth in space should also be discussed. Anyone ever stopped to think how unrealistic are "sensors" that pretty much every sci-fi work features? What do they sense?
I think that on early space travel (when we stil are relatively "realistic" knowing where ships are and what planets are made off should be a major challenge, and it'd become easier as you progress through the tech tree into scientomagical mechanisms (gravitron sensors anyone? )

Good point. This best detailed depiction of sensors in space I've found in Mass Effect, where they use what is essentially radio-telescopes to detect the massive amounts of radiation released by traveling ships. Hence the stealth systems essentially being massive heat-sinks.

Yes but then there's lightspeed lag. You would only notice a massive ship moving near mars 30 mins after it really showed up. Unless we are taking the artistic license of making speed of light infinite for coding reasons so we don't end up with funky relativistic problems.

I beleive that up to mid spacefaring, planet and star analysis would require dedicated spaceships with spectroscopy equipment, until magicscan tech makes it pointless.

This would be an important aspect of space exploration, as probes and telescopes would get you only so far by knowing how a celestial body looks like, but not its composition. (actually telescopes can determine the composition of stuff, but you need a good image)

Djohaal wrote:

Yes but then there's lightspeed lag. You would only notice a massive ship moving near mars 30 mins after it really showed up. Unless we are taking the artistic license of making speed of light infinite for coding reasons so we don't end up with funky relativistic problems.

I beleive that up to mid spacefaring, planet and star analysis would require dedicated spaceships with spectroscopy equipment, until magicscan tech makes it pointless.

This would be an important aspect of space exploration, as probes and telescopes would get you only so far by knowing how a celestial body looks like, but not its composition. (actually telescopes can determine the composition of stuff, but you need a good image)

In any case, I think technological sensors might be suited to a different thread.

Tenebrarum wrote:

Djohaal wrote:

Yes but then there's lightspeed lag. You would only notice a massive ship moving near mars 30 mins after it really showed up. Unless we are taking the artistic license of making speed of light infinite for coding reasons so we don't end up with funky relativistic problems.

I beleive that up to mid spacefaring, planet and star analysis would require dedicated spaceships with spectroscopy equipment, until magicscan tech makes it pointless.

This would be an important aspect of space exploration, as probes and telescopes would get you only so far by knowing how a celestial body looks like, but not its composition. (actually telescopes can determine the composition of stuff, but you need a good image)

In any case, I think technological sensors might be suited to a different thread.

For organism based stealth I reckon we could copy the traditional stealth of some RPGs (oblivion for instance, or fallout and friends), but have the stealth "factor" be a function of several components. Padded feet for instance. Or your critter's color compared to the biome's main color.

Djohaal wrote:

For organism based stealth I reckon we could copy the traditional stealth of some RPGs (oblivion for instance, or fallout and friends), but have the stealth "factor" be a function of several components. Padded feet for instance. Or your critter's color compared to the biome's main color.

It's more complicated than that though. What about mimicary? What about different sensors? What about zebras?

Tenebrarum wrote:

Djohaal wrote:

For organism based stealth I reckon we could copy the traditional stealth of some RPGs (oblivion for instance, or fallout and friends), but have the stealth "factor" be a function of several components. Padded feet for instance. Or your critter's color compared to the biome's main color.

It's more complicated than that though. What about mimicary? What about different sensors? What about zebras?

Yes the concept would have to grow outwards into such elements. Mimicry would be specially complex because having the computer "measure" how much you look like with a meanier species might be difficult. It can't derive and compare parametric data from the skeleton because someone can easily make a similar shape from a radically different skeleton.
Different sensors would also be accounted for, although it doesn't seem to be as complex.
For zebras and such, I beleive we could make it give contrasted stripes and shapes a individual detection bonus if you are close to other individuals. (eg, a predator is more likely to shift focus to another critter, getting confused)

Also poisonous blood/ vile tasting flesh and warning colors.

I think that we have to think about what makes the senses sense the most. Sight is most reactive to colour, shape and movement, to different extents depending on the type of eye. So having the colour that is sensed be similar to the colour of the surroundings would help. Moving slowly would make it less likely to be seen, too. Shape is tricky. If you have a similar shape to something that is common in your surroundings you aren't noticed? That may be hard to code... With sound, quieter is better. Smell is basically unhideable unless if you roll in something, but even that is only temporary. Feel? Less is better to avoid being felt. Can't hide electrosense, but you can stop moving and thus producing them (except for those your organs give off). And echolocation would be a matter of shape yet again.

As for things outside of sensory camouflage, i have no idea. Having a sudden change in shape should give a shock to an enemy, like with frilled-neck lizards and certain butterflies and moths. But what is this shock, how do we calculate it and does it have any lasting effects? I think mimicry would be covered with the shape thing (it all hinges on that, doesn't it?).

US_of_Alaska wrote:

I think that we have to think about what makes the senses sense the most. Sight is most reactive to colour, shape and movement, to different extents depending on the type of eye. So having the colour that is sensed be similar to the colour of the surroundings would help. Moving slowly would make it less likely to be seen, too. Shape is tricky. If you have a similar shape to something that is common in your surroundings you aren't noticed? That may be hard to code... With sound, quieter is better. Smell is basically unhideable unless if you roll in something, but even that is only temporary. Feel? Less is better to avoid being felt. Can't hide electrosense, but you can stop moving and thus producing them (except for those your organs give off). And echolocation would be a matter of shape yet again.

As for things outside of sensory camouflage, i have no idea. Having a sudden change in shape should give a shock to an enemy, like with frilled-neck lizards and certain butterflies and moths. But what is this shock, how do we calculate it and does it have any lasting effects? I think mimicry would be covered with the shape thing (it all hinges on that, doesn't it?).

It might be possible to compare geometries, but then again, I've never seen any sort of geometry comparing system that worked (see "models shaped like this" On the sketchup 3d warehouse. It makes no sense)

What if it grabbed the creatures in orthostatic position, and did three shots of them in frontal, lateral and horizontal planes, and then compared the 2d images between each other?

Djohaal wrote:

What if it grabbed the creatures in orthostatic position, and did three shots of them in frontal, lateral and horizontal planes, and then compared the 2d images between each other?

It'd probably be a simpler method, (although that's coming from someone with little to no programming experience) but would it be simple (and fast) enough for the context we're using? I'm not sure if I want to wait 30 seconds before I camoflauge myself. By that time, the predator I'm hiding from would've already eaten me.

I'm not even sure if this is possible, though. But I suspect this feature will start somewhat slow and get faster over time.

I like the suggestion, though.

Once again, anything I suggest here would pretty much be...Yeah, I have nothing to suggest (and if I do have anything someone here probably already said it). Now I just feel like a critic...

I've got a few for optical sensors:
Fading with the terrain: We can simply raytrace the ground or whatever is the creature living on, get it's texture's average color value and compare it with average color of the organism. For flying organisms, we could get color of the planet's atmosphere.
Sneaking: Get the organism's height and speed. The lower, the better.
Mimicry/Zebra confusion: Quite difficult, if not to impossible, to implement into the game. I think we should just include a bunch of different "camo" patterns, such as stripes, spots, or similar.

From these factors, a coefficient would be made, wich would then be compared against the looking organism's eye "quality" (with added penalty for distance or bad weather conditions).


Now, that is eyes. Alaska already did a good job on the rest, so is it all?

Djohaal wrote:

What if it grabbed the creatures in orthostatic position, and did three shots of them in frontal, lateral and horizontal planes, and then compared the 2d images between each other?

Don't forget that every organism evolves, and each time it evolves, it will all have to be recalculated. It would be okay for a few species, but there will be quite a lot of them (around a hundred, it will fluctate as species will split and die off).

Commander Keen wrote:

I've got a few for optical sensors:
Fading with the terrain: We can simply raytrace the ground or whatever is the creature living on, get it's texture's average color value and compare it with average color of the organism. For flying organisms, we could get color of the planet's atmosphere.
Sneaking: Get the organism's height and speed. The lower, the better.
Mimicry/Zebra confusion: Quite difficult, if not to impossible, to implement into the game. I think we should just include a bunch of different "camo" patterns, such as stripes, spots, or similar.

From these factors, a coefficient would be made, wich would then be compared against the looking organism's eye "quality" (with added penalty for distance or bad weather conditions).


Now, that is eyes. Alaska already did a good job on the rest, so is it all?

Djohaal wrote:

What if it grabbed the creatures in orthostatic position, and did three shots of them in frontal, lateral and horizontal planes, and then compared the 2d images between each other?

Don't forget that every organism evolves, and each time it evolves, it will all have to be recalculated. It would be okay for a few species, but there will be quite a lot of them (around a hundred, it will fluctate as species will split and die off).

Yes mimcry would become a problem, unless it used very low res shots, but then weirdness could ensue.

We could refine this problem to only include comparison between Organisms that the onlooking Organism actually encounters.

For Batesian Mimicry (in terms of colouration and pattern), since the skin pattern system that we are using for the OE (As far as I remember) supports a uniform system of colouration and patterns. We would only have to compare the difference between the two Organism's data (but I have no idea how it would be programmed) based on the other Organism's eyesight (for example, if colour or shape is most important).

Mimicary shots could be taken before each generation. They could use wireframe models instead of full pictures. We must remember that it doesn't have to be Front A vs. Front B. The wings of certain moths are designed to look like the faces of certain owls.

That is the problem. Computers are incredbly dumb when it comes to analyzing patterns. We'd need to write very advanced (and processing expensive) algorythms to do anything useful...

Djohaal wrote:

That is the problem. Computers are incredbly dumb when it comes to analyzing patterns. We'd need to write very advanced (and processing expensive) algorythms to do anything useful...

I wouldn't imagine. With wireframe models you could calculate the degree of overlap. The more overlap, the better mimicary.

Tenebrarum wrote:

Djohaal wrote:

That is the problem. Computers are incredbly dumb when it comes to analyzing patterns. We'd need to write very advanced (and processing expensive) algorythms to do anything useful...

I wouldn't imagine. With wireframe models you could calculate the degree of overlap. The more overlap, the better mimicary.

Trust me doing that is way more complicated than it sounds...

Djohaal wrote:

What if it grabbed the creatures in orthostatic position, and did three shots of them in frontal, lateral and horizontal planes, and then compared the 2d images between each other?

Too cpmplicated, and it ould only work if a creature was being looked at exactly along one of those axes.
I think we're on to something with certain patterns giving different levels of stealth. In the OE's texturing panel, certain texture brushes could correspond with advantages in stealth.

~sciocont wrote:

Djohaal wrote:

What if it grabbed the creatures in orthostatic position, and did three shots of them in frontal, lateral and horizontal planes, and then compared the 2d images between each other?

Too cpmplicated, and it ould only work if a creature was being looked at exactly along one of those axes.
I think we're on to something with certain patterns giving different levels of stealth. In the OE's texturing panel, certain texture brushes could correspond with advantages in stealth.

Biome cloaking wouldn't be hard to do. It'd have to pick your texture's average color and compare to the average color of the biome's base texture, shouldn't be very complicated.