#procedural planets

Continuous World Generation in 'No Man's Sky'

GDC publically releases the videos of some sessions every year and this year one of those is Innes McKendrick’s talk about the generation pipeline in No Man’s Sky.

There’s a lot of interesting bits in the talk, so I’ll just highlight a few things that appeal to me personally.

First off, the emphasis on augmenting artists rather than replacing them. The engine is agnostic about where the content comes from, making no distinction between generative content and hand-authored content. As Innes says, the goal was to enable "our artists to produce more, rather than replacing them with an algorithm that doesn't quite do the same job."

This is an important point, and something I’ve tried to emphasize here: sometimes a generative approach isn’t the best approach; choose the method that fits. And giving artists more tools is a perfectly valid approach to using generative methods.

Another point is the use of evaluation tools to test the generator and its output, so that they could track changes from build to build. This is exactly the kind of data that is useful to automate, and it makes the jobs of the artists and the programmers easier.

The talk also has a ton of technical details that are useful to consider if you’re wanting to work on your own continuous world generation. Not to mention voxels, interesting non-planar spaces, and building a world at scale.

Starflight (1986)

I found my record-album copy of Starflight! (While I was looking for something else, naturally. It was inside the Age of Empires collector’s edition box for some reason.)

Staring development before the release of Elite, Starflight had a rather different philosophy for procedural generation. You see, instead of generating a few trillion galaxies, it generates about 800 planets. Since the game came on two 360K floppy disks, this was rather a challenge.

It uses a fixed seed, so that every game has the same planets. It also remembers your actions, so that minerals you collected would still be gone if you returned.

Where it really stands out from its contemporaries is the story content; with a lot of surprisingly deep interactions. Combined with the on-planet gameplay, which is the part that uses the most procedural generation, this has been incredibly influential influential: the planetary exploration against the backdrop of an epic sandbox story is the prototype for everything from Star Control II to Mass Effect to the Forgotten Beasts in Dwarf Fortress.

As the continuation of the Elite procedural generation legacy, Elite: Dangerous is of obvious interest. While I have some things to say specifically about the game’s current state, here’s a look behind the scenes at the planned future development of planetary surfaces in the upcoming Horizons season. Much of it is specific to this game, though if you’re working on your own planet generator you’ll undoubtedly get some inspiration.

Here’s the videos themselves, without the audio or the streaming glitches:

Frontier: First Encounters (1995)

Explore a galaxy of procedurally generated stars and planets!

The third Elite game (legal disputes notwithstanding) was very similar to Frontier: Elite 2. It updated the graphics with procedural texture mapping and it adds more of a plot structure, should you choose to explore that. It also added a few new bugs. Despite those, it’s become a reference implementation for Frontier/Elite style gameplay.

The world of Elite is still fundamentally a lonely one, full-motion-video NPCs notwithstanding. But there are an immense number of planets to potentially visit, and while most of them don’t have a whole lot of extrinsic rewards for doing so, the physics of landing your spaceship at any arbitrary location keeps each planet as a destination, not just something to look at.

The planetary rendering code makes use of progressive tessellation to handle the level of detail changes required to deal with the differences in scale. Due to the tech limitations at the time (on the 386) this creates very visible popping as you fly closer to a planet. This can occasionally make things tricky when you try to land. 

The way this exposes the underlying algorithm also has an interesting humanizing effect: the game feels more alive because of this technical limitation. Observing the inner workings of the game makes it easier to understand what is going on, involving the player in the process and facilitating the ergodic agency that is at the heart of the game’s appeal.

Once you are on the ground, the major gameplay element that interacts with arbitrary parts of the surface is the deployable mining machine, which can be used nearly anywhere on the planet. (You get fined if you try to mine near a populated spaceport) While it’s not the most engaging part of the game, it does tie intimately into the minute details of the generator.

If you’re going to design a vast generative landscape, that’s something worth keeping in mind: having some kind of interaction that depends on the details of your generator. Try to give your generator a feel that goes beyond just simple visual looks. It can be an extrinsic reward (as with the mining) or intrinsic (such as the feel of trying to land on the surface) but think beyond the immediate, material appearance and try to give the generator a life beyond its surface representation.

First Encounter's major intrinsic reward, like with most exploration games, is sightseeing. Because the physics of the planetary orbits are modeled in detail, you can go stargazing: turn on the time compression and watch the stars and planets spin overhead. And while the planets aren’t populated with very many structures or vegetation, they do manage to convey an atmospheric bird’s eye view of countless alien worlds.