Saturday, May 04, 2019

[xoqqrdai] Procedurally generated full universe

We may be close to being able to procedurally generate a somewhat realistic entire static universe, e.g., the ultimate open-world game.

Minecraft as well as many other games already demonstrate procedurally generating an Earth-like world.  We assume that that degree of realism is good enough, because it's a game (though we would need to transform Minecraft onto a sphere or a cube).  We need to simulate other bodies in the solar system, though those don't seem to be too difficult and techniques to procedurally generate similar such bodies likely already exist.

To procedurally generate other star systems, we need a probability distribution over types of stars, and over possible layouts of stars and planetary bodies within a star system.

To keep things simple, we are procedurally generating a static universe, so we don't have to simulate orbital mechanics, or planetary, stellar, or galactic evolution.  Nothing moves (other than the player): everything is nailed down.  (What about day-night cycle?  Mobs come out at night.)

To procedurally generate a galaxy, we need a probability distribution over layouts of star systems in a galaxy, e.g., bright stars concentrated in spirals.  We also want to put gas and dust between star systems, because nebulae are pretty.  Astronomers have measured all these things.  We also want a model of the way things are, or can be, at the centers of galaxies, around its supermassive black hole.

Comets, rogue planets?

Trying to model how astronomical things like nebulae look from different angles, even for a static universe, might be difficult: naively, it would be a simulation of light interacting with vast amounts of diffuse material in space.

Do we want some instances of colliding galaxies?  If so, we'll need to procedurally generate things like starburst regions and active galactic nuclei.  Going back a bit, do we also want some instances of planetary bodies colliding (or having just collided)?  Although it's a static universe, it need not be a snapshot of its steady state, though the latter would be easier.

To procedurally generate collections of galaxies, we need a models of how galaxies are organized into clusters and superclusters, then into sheets and filaments and voids.  Although astronomers have observed and measured these structures, I haven't yet seen models which can generate synthetic examples of such structures.  It's probably not too difficult.

And then at the largest scale, the universe is uniform, so this is easy.

No comments:

Post a Comment