Using Voxels, Particles, Atoms, and Compounds for Physics and Graphics

Using voxels,particles,atoms,and compounds for physics and graphics would be easier than using polygons as voxels are volumetric thus physics calculations use less resources,can easily use a grid, and make every thing use the compound system for materials. Procedural Generation would be easier to code,be more advanced,and use less resources. There is the Atomontage game engine which is interesting because it uses voxels and atoms(which can convert into different types of atoms.) as its main tools for physics and graphics and optimizes things very efficiently by the use of a simple AI based controller for managing the modules to perform the best in a particular situation on a particular machine. I think we should contact him about using the voxel part of the physics and graphics engine and the AI controller or have him help us make a custom ones for thrive. The developer Branislav Síleš has some contacts a email: sites@atomontage.com and skype: Branislav Siles

Hey Ton12

Quote :

would be easier than using polygons as voxels are volumetric

I'm not sold on this, I'm having a hard time finding useful articles/tutorials on how all of this works, it all seems like a not yet ready attempt at being the next-gen rendering techniques, which isn't very useful for us.

Quote :

thus physics calculations use less resources

Physics are unlikely to include especially expensive calculations for us from what I can foresee.

Quote :

Procedural Generation would be easier to code,be more advanced,and use less resources.

This I agree with, and will almost certainly be a part of the game, but it doesn't necessitate a change of technology.

Quote :

There is the Atomontage game engine

It's a graphics engine, or rather could become one. They don't have anything other than videos and images to show at the moment and there doesn't seem to have been much of any progress in the past year or two (a few occasional tweets was all I could find). Also the idea that we can ask for sourcecode for something like this isn't realistic. They're looking for investors and investors will want a return on their investment, hence the graphics engine will be at a license and they can't just give it away, but even if he would/could, it would be a half-finished framework at best with a very unsure prospect of any advantages.

I like voxels, I really do. However, I had a bunch of reasons why going all-out voxel would be a bad idea, but then I deleted my diatribe since I reread the OP and it hit me that the methods the Atomontage engine uses probably cover much of that.

Anyway, a summary of my issues with pure-voxel engines:

• Managing areas of different resolution can be icky -- especially with systems that change over time, since resolution will have to change over time too. That means lots of memory copying, generally (though I can think of one cheap way to double res with minimal memory movement, it would lead to much pointer-chasing).
  
• The simplest solution to that problem is to keep everything at the highest res you'll need. For memory reasons, this is a very bad idea.
  
• Rendering the voxel data will most likely require us to mesh everything anyway. Unless, by the nature of the sim, we can be sure it either changes a lot (thus, necessitating heavy mesh modification regardless of whether we use voxels or not) or it doesn't change at all (thus allowing us to only run the meshing once). This includes systems whose geometry doesn't morph except in catastrophic cases -- like a spaceship which, every now and then, gets a chunk blown out of it. Things will also work out fine if the state of the system changes just a little bit, which would allow us to tweak the positions of the mesh vertices directly instead of remeshing entirely (coincidentally, what we're planning to do with microbe membranes).
  
• There are alternatives to meshing, of course -- not least of which is voxel cone raytracing, which we could see on some Xbone games if it works. We could use clouds of particles to render voxels, with which we'd lose out on surface solidity, which is exactly what we'd want for the diffuse systems we might use voxels for (fire, smoke, clouds -- none of which we'd actually bother with a voxel sim for though :P).
  

Of course, Atomontage seems to use voxels only for static geometry. That is, the voxel data is only generated and stored for as long as it takes to mesh everything. So, they're pretty much doing what we'll be doing with the cell membranes. I feel like I'm missing something here though -- what are these atoms you speak of?