Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Normandy]

The world is just awesome. Perlin noise heightmaps just don't cut it justice. Everything from plate tectonics to climate to bedrock composition affects topography.

Originally I was going to make my own generator, but time constraints have forced me otherwise. However, you can still contribute! Post interesting terrain generation projects that you are undertaking here, for us to admire and comment on.

Old OP:

Spoiler (click to show/hide)

Phase I - Plate Tectonics: Part 1
Plate Tectonics is more important to the generation of the world than you think. Not only does it create mountains and rifts, it drives nearly all crustal movement. Without it, our landscape would be flat and featureless. I will be implementing two main drivers behind plate tectonics: magma convection currents and plate density differences. Let's begin with magma convection currents.

Temperature differences in magma cause plumes of hot magma to rise in some locations, cool, then fall in other locations. This drives some movement of the crust above. It's not really known what specifically causes these temperature differences, so I'll start off with a simple perlin noise map of magma temperature:



I'm not going to establish a specific temperature scale just yet, but the yellow magma is hotter while the orange magma is cooler. Simple enough. Next, we need to actually establish the convection currents. We'll keep it simple for now, and keep magma velocity proportional to the temperature gradient. We end up with something looking like this:


The hue of the color represents the direction the magma is moving in, while the value (i.e. brightness) represents the speed of the magma. Use the color wheel below for reference. For example, a bright red spot represents magma quickly moving to the right, while a dark green spot represents magma moving slowly to the upper-left. Note that when the colors spin in the same direction as the color circle the magma is diverging, and when they spin in the opposite direction the magma is converging. You can match up these spots of convergence and divergence on the magma temperature map above.


Using this, we can begin to construct a map of the stresses on the plates above. These are only the stresses contributed by the magma convection; more stresses will be added once the plates actually begin moving. Another important thing to note is that the movement of the plates themselves can affect the magma convection current below. These effects will be implemented over the next few days. The current map of stresses generated by this static convection:

Red is compression stress (pushed in), green is tension stress (pulled apart), and blue is shear stress (pulled side-to-side). Put simplistically, compression creates mountains, while tension and shear creates rifts. Visual artifacts aside (I'm not quite sure what causes the blocks yet), already you can see some interesting terrain shaping up. These stresses will form the basis of the next step, actually breaking the crust into plates.

[Hugehead]

This is really neat, will you release it as a developer's resource or do you intend to keep it for yourself?

[Captain Mayday]

I have seriously been looking for a terrain generator of this nature for years. I'm deeply interested in seeing how this goes.

Edit: I notice this isn't seamless. Are you looking at getting seamless maps working once you have overall functionality up and running?

[Akhier the Dragon hearted]

   Quite an interesting terrain generator you have
shaping up here. I wish you luck on this venture
because the way its looking the generator will be
amazing in its realness as compared to life.

[Virex]

I've looked into noise generation for plate tectonics before, using Teragen 2, though without taking convection into account. To generate the plates I started off with distorted voronoi noise. I used the hue of the flat-shaded cell variants to determine the average height of the plate and the distance gradient voronoi variant to mask where the mountains and troughs should be.

[nbonaparte]

I am very interested in this, posting to watch.

[Max White]

Your block effects are interesting, any clues as to the cause? I mean if I had to take a stab in the dark, I might guess that to speed up the generator you multi threaded it, giving each 'block' it's own thread, but didn't give any way for them to interact, leading to less activity on the edges on such blocks.

On that note, it would be in your bets intrest to make the edges interact, so that the left conencts to the right, and the top to the bottem, giving some strange hyper dimentional sphere.

[Shades]

Your block effects are interesting, any clues as to the cause? I mean if I had to take a stab in the dark, I might guess that to speed up the generator you multi threaded it, giving each 'block' it's own thread, but didn't give any way for them to interact, leading to less activity on the edges on such blocks.

On that note, it would be in your bets intrest to make the edges interact, so that the left conencts to the right, and the top to the bottem, giving some strange hyper dimentional sphere.

From a performance point of view he'd be best off treating the blocks as tasks with a multi-threaded api like tbb. That way that can still interact and you only have as many threads as you do cores but can infinitely (within memory limits) extend then world, rather than a thread for each block. But your probably right that it's an interaction disconnect cause the tile look.

Looks like an interesting project, watching to see how it goes

[Rose]

This looks rather interesting indeed. following this.

[Normandy]

Actually, the blocks are visual artifacts generated by perlin noise. Perlin noise looks seamless, but in reality it has subtle artifacts when any coordinate is close to an integer. I'll probably change the noise generation or find some way to remove these artifacts, but it's not of particular interest to me now, since eventually the magma temperatures will be affected by other factors which (should) get rid of these artifacts.

I won't be generating plates through noise; I'm first going to see if I can manage to get them down through the stress maps. The mathematics are a bit complex, so I'm doing it all on paper first; however I've got the theory behind it:

The movement of a plate should be equal to the sum of the forces on it; ignoring friction between plates and gravity for a moment, let's look solely at the force contributed by magma convection currents. In a highly viscous fluid, drag is directly proportional to velocity, so that works out well. Using a simple flood-fill algorithm over a plate to calculate this total force is pretty easy then. You can then use this total force to accelerate the plate accordingly.

Once the plates begin moving, there will be a disparity between the motion of the plate and the motion of the magma underneath. For now we'll assume the plates do not change the flow of the magma (they do, but that's rather complicated right now). The difference between the motion of the plate and the magma underneath creates stress. Once stress reaches a certain threshold, the plate begins to deform and rift (i.e. separate), forming new plates.

This model depends on rigid plates that don't have significant deformations or velocity deviations relative to their total velocity. I think I can fix for the deformations by adding in a deformation factor somewhere in there, but I'm currently stuck on the actual process of separate plates forming. Since the rift has not completely separated the two plates yet, a flood-fill algorithm will overstep this boundary. However, the velocities of the plates on either side of the forming rift are bound to be in complete opposite directions.

Anyways, next update will probably add rectangular map wrapping to make some math easier. Look forward to it.

EDIT:
Also, if anyone has a java function for 4d perlin noise, I could sure use it.

[Armok]

the plates need to rotate to.

[Nadaka]

Your block effects are interesting, any clues as to the cause? I mean if I had to take a stab in the dark, I might guess that to speed up the generator you multi threaded it, giving each 'block' it's own thread, but didn't give any way for them to interact, leading to less activity on the edges on such blocks.

On that note, it would be in your bets intrest to make the edges interact, so that the left conencts to the right, and the top to the bottem, giving some strange hyper dimentional sphere.

Your strange multidimensional sphere is actually a simple torus. Spheres in higher dimensions are much weirder than that.

[G-Flex]

Your strange multidimensional sphere is actually a simple torus. Spheres in higher dimensions are much weirder than that.

It's not even a simple torus, actually. It's toroidal, but with an odd distortion going on.

Let me put it this way: Imagine folding a square piece of paper like that. Fold it into a tube, then attempt to fold that tube into a torus by connecting the remaining edges. You can't! The outer circumference of the torus is greater than the inner circumference, which doesn't work here for obvious reasons. In other words, your typical RPG map isn't only a torus, it's a distorted toroidal shape that doesn't make much sense in real-world terms, as the area of some arbitrary section of the torus does not always equal the same amount of area on the map itself.

[Nadaka]

Right of course, I was assuming that people would understand that there is some distortion any time you attempt wrap a flat plain onto a curved surface. You can still wrap that sheet of paper into a torus without distortion if you allow the thickness between the inner and outer surface to be 0.

[Rose]

wrap a sheet of paper around a sphere. OMG! you can't!