How Much Procedurally Generated?

For those who don't know, Procedural Content Generation, Procedural Generation or PCG for short can be thought of as a fancy way of saying Randomized Content.

Here's a video that goes through the development of the first Elite game and how PCG played a huge role in it (educational for those who don't know) that video also made me understand why people care so much about Elite and why they often seem to perhaps over-appreciate/glorify it.

For a bit of background of why i started thinking about this...

Here's a video introduction for a game named Limit Theory (pre-alpha footage, game looks much better by now ofc but still has ways to go) that has been successfully funded through kickstarter.


Now after seeing that, Limit Theory, a game based completely around PCG, if i'm not mistaken even the textures will be generated through PCG, whereas former examples of PCG use has mostly been universe/terrain generation (for example: Minecraft) now that game generates the entirety of its content through PCG, quite interesting and i'm going to buy it as soon as it's out, that thing has huge potential! But learning about it quickly made me wonder how much of Elite Dangerous will rely on PCG.

So far the only information i have is that apart from the currently known universe discovered through our exploration (in real life), the galaxies will be generated through PCG with a total of 400 billion solar systems if i'm not mistaken, and the NPCs (inevitably) will mostly be generated through PCG as well.

But apart from this i don't know how PCG is being utilized in the game, i've seen an example from Limit Theory where Josh (lead developer) has done incredible things for example combining PCG with the AI, allowing the AI behavior patterns to be randomized and harder to predict/more dynamic. (<- more here)

So to repeat the question: How much is PCG being used within Elite Dangerous? As in, what will PCG be used for and what will it not be used for? i'm really interested to know this, especially while playing because i think it'll be cool to know what was hand-made and what was just generated by the game.
 
Last edited:
For those who don't know, Procedural Content Generation or Procedural Generation for short can be thought of as a fancy way of saying Randomized Content.

Here's a video that goes through the development of the first Elite game and how PCG played a huge role in it (educational for those who don't know) that video also made me understand why people care so much about Elite and why they often seem to perhaps over-appreciate/glorify it.

For a bit of background of why i started thinking about this...

Here's a video introduction for a game named Limit Theory (pre-alpha footage, game looks much better by now ofc but still has ways to go) that has been successfully funded through kickstarter.


Now after seeing that, Limit Theory, a game based completely around PCG, if i'm not mistaken even the textures will be generated through PCG, whereas former examples of PCG use has mostly been universe/terrain generation (for example: Minecraft) now that game generates the entirety of its content through PCG, quite interesting and i'm going to buy it as soon as it's out, that thing has huge potential! But learning about it quickly made me wonder how much of Elite Dangerous will rely on PCG.

So far the only information i have is that apart from the currently known universe discovered through our exploration (in real life), the galaxies will be generated through PCG with a total of 400 billion solar systems if i'm not mistaken, and the NPCs (inevitably) will mostly be generated through PCG as well.

But apart from this i don't know how PCG is being utilized in the game, i've seen an example from Limit Theory where Josh (lead developer) has done incredible things for example combining PCG with the AI, allowing the AI behavior patterns to be randomized and harder to predict/more dynamic. (<- more here)

So to repeat the question: How much is PCG being used within Elite Dangerous? As in, what will PCG be used for and what will it not be used for? i'm really interested to know this, especially while playing because i think it'll be cool to know what was hand-made and what was just generated by the game.

Procedural generation will be used for planets but under guidance of artists, meaning they make a "recipe" for how a planet should look.

Other than that nebulas i would guess, maybe missions will have a procedural nature to them.
 
A couple of quick opinions...

Procedural content isn't always the best way to get things done. With the best will in the world, procedural space-ships won't look nearly as good as hand-crafted space-spaceships. Also, procedural textures might be good for natural stuff, but again, a lot of things will look better with hand-crafted textures. The one major benefit PCG textures have over raster textures is that they take up barely any room, so they're good for mobile games where space is at a premium.

Also, I saw that Limit Theory AI video, and what Josh was describing would be better described as simulation. If I understand it correctly, it's nothing new, and Elite will have that to some degree or another. Cerainly with the evolving galaxy it will have it in spades.

I would be interested to know how deep the procedural stuff is going in Elite though. For instance, we know that the layout of spacestation interiors is procedurally generated, and you can get elements like parks. But what I'd like to know is whether the parks all going to look the same, or will they in turn be procedurally generated? And if the parks are procedurally generated, what about the individual trees, or benches, or fountains....
 
I think any life that we discover on planets (when planetary interaction comes in) will also be procedurally generated. Some of it will be 'handmade' (like the Thargoids, obviously, and any other sentient life that we may encounter that will be part of the back story), but on the whole I think procedural generation will be used for the vast majority of the rest.

Cities and colonies will possibly be built from PCG too?
 
all of it?

I think the vast majority of the game will be procedurally generated. Systems, planets, stations, ships, missions, all will have templates and the game will generate instances of each as required.

Everything DB and Michael have said has suggested to me that they're continuing with PG-everything, and tweaking the generators where required to get specific results where needed.
 
So to repeat the question: How much is PCG being used within Elite Dangerous? As in, what will PCG be used for and what will it not be used for? i'm really interested to know this, especially while playing because i think it'll be cool to know what was hand-made and what was just generated by the game.

Well, this is what I understand so far to be procedurally generated in ED (subject to known data, which in some cases overwrites the automatic procedural content, for example Earth and Sol system and local stars, known exoplanets etc):
  • The Galaxy
  • Star Systems
    • Celestial bodies including sun, planet, moon, asteroids and other celestial body combinations
    • Political, economical and tech level ratings
  • Planetary surfaces
  • Space stations
    • hand crafted space station modules using procedural combinations
  • Background galaxy simulation
    (Not sure if this is considered to be procedural, but a lot of algorithms in play to create dynamic events around the known galaxy)
  • NPCs
    • traffic
    • archetypes
  • Missions
  • Ship damage
  • Names
    • Star systems
    • planets
    • NPCs
  • Newsfeeds

Its also worth noting that some tier 1 NPCs, missions and events will be hand crafted and injected into the game (multiplayer version). Also all space craft will be hand crafted.

There are probably a lot of little largely unnoticeable tweaks and bits of content that may also be procedurally generated that will probably go largely unnoticable.
 
Last edited:
Welcome to the forums Dragotha! and thanks for your reply :)

A couple of quick opinions...

Procedural content isn't always the best way to get things done. With the best will in the world, procedural space-ships won't look nearly as good as hand-crafted space-spaceships. Also, procedural textures might be good for natural stuff, but again, a lot of things will look better with hand-crafted textures. The one major benefit PCG textures have over raster textures is that they take up barely any room, so they're good for mobile games where space is at a premium.

Also, I saw that Limit Theory AI video, and what Josh was describing would be better described as simulation. If I understand it correctly, it's nothing new, and Elite will have that to some degree or another. Cerainly with the evolving galaxy it will have it in spades.

I would be interested to know how deep the procedural stuff is going in Elite though. For instance, we know that the layout of spacestation interiors is procedurally generated, and you can get elements like parks. But what I'd like to know is whether the parks all going to look the same, or will they in turn be procedurally generated? And if the parks are procedurally generated, what about the individual trees, or benches, or fountains....

Of course there's no such thing as "always the best way" but to be noted is that PCG ships, although not as pretty as artist-designed ones, will offer a whole lot of variety, and to me that is an acceptable replacement for artistic finesse. I'm open to new things and as a gamer i think that's pretty important, Limit Theory is a game designed quite differently from most games we know today, i'm willing to give it it's chance and try it out, just as well as you i can see the potential flaws it may have or which parts might be better off not procedurally generated, but looking at it differently, thanks to the reliance on PCG, Limit Theory's development time is much shorter than otherwise (there's no 3D modelling and minimal if any texture work involved for example) thanks to that, Limit Theory is possible with a a tiny-sized development team and as low a budget as 200k $, which is quite a small amount for a game of that scale, this design method might grow more popular thanks to it's capabilities of shortening development time and reduced costs of manufacturing. Although i don't want to see this trend overused in console and PC games, it could make for some interesting mobile device games.

I'm not surprised his AI thing has been done before, good to know Elite has that too.

And yeah i'd like to know all that stuff too which is why i made this thread.

Thanks for that list bitrogue, it's helpful :D
 
Last edited:
Yeah, I have a great interest in, and really enjoy games with procedural content. Limit Theory is maybe a showcase for it even better than Elite in some ways. Though the best example is probably .kkrieger, which was entirely PCG even down to the sound effects and music, and fit an FPS into 96kb...
 
kkrieger is just mind blowing, clearly shows the power of proc.gen.
I didn't know about it, so thanks for the info!
 
I'd also like to point out that procedural generation isn't the same thing as randomly generating something, there are rules.

Indeed, that's why I chose the parks in space-stations as an example regarding what level of granularity the game will finally aim for. You could just use a few hand-crafted parks which get repeated over and over, or you could write a routine that generates different parks depending on various factors. Does it have a lake/does it have a water-feature/is there an open grassy area, what's the tree density/where are the entrances and exits/what route does the path take... And then you could have rules for the different furnishings, like benches in Federation parks are solid and angular, whereas benches in Imperial parks are curvy and ornate. You could even procedurally modify the base meshes for benches and lamp-posts... Does the Imperial Lamppost have a fluted column or a smooth column, and does it have a Corinthian/Doric/Ionic cap?

I've no real conception of how many man-hours it would takes to write even a basic auto-park generator, but I imagine quite a few. And then there are so many parts of the game, it would be hard to lavish detail on all of them without melting our CPUs and blowing the game's budget spectacularly.

It's a situation where I'd love to see innovative middleware solutions develop. Consumers are demanding bigger and more detailed open-worlds, even while these mega-budget games are now seen as failures for selling only 4 million copies. Something has to give. Speedtree has already shown there's a market both in games and movies for their foliage-generation tech. But what if there were middleware solutions for all sorts of stuff. Building generators, street furniture generators, bookshelf generators, mug generators, etc.

Maybe it's a bit early days for that fantasy to play out, but I wouldn't be surprised if the industry headed that way eventually. Companies already outsource asset creation, so if those asset creators find a way of churning out a high volume of product relatively cheaply (procedural generation), then they'll out-compete their hand-crafting competitors. Of course, you'd probably still outsource some special stuff to the hand-crafting studios, but...

Oh wait, I'm rambling.
 
Indeed, that's why I chose the parks in space-stations as an example regarding what level of granularity the game will finally aim for. You could just use a few hand-crafted parks which get repeated over and over, or you could write a routine that generates different parks depending on various factors. Does it have a lake/does it have a water-feature/is there an open grassy area, what's the tree density/where are the entrances and exits/what route does the path take... And then you could have rules for the different furnishings, like benches in Federation parks are solid and angular, whereas benches in Imperial parks are curvy and ornate. You could even procedurally modify the base meshes for benches and lamp-posts... Does the Imperial Lamppost have a fluted column or a smooth column, and does it have a Corinthian/Doric/Ionic cap?

I've no real conception of how many man-hours it would takes to write even a basic auto-park generator, but I imagine quite a few. And then there are so many parts of the game, it would be hard to lavish detail on all of them without melting our CPUs and blowing the game's budget spectacularly.

It's a situation where I'd love to see innovative middleware solutions develop. Consumers are demanding bigger and more detailed open-worlds, even while these mega-budget games are now seen as failures for selling only 4 million copies. Something has to give. Speedtree has already shown there's a market both in games and movies for their foliage-generation tech. But what if there were middleware solutions for all sorts of stuff. Building generators, street furniture generators, bookshelf generators, mug generators, etc.

Maybe it's a bit early days for that fantasy to play out, but I wouldn't be surprised if the industry headed that way eventually. Companies already outsource asset creation, so if those asset creators find a way of churning out a high volume of product relatively cheaply (procedural generation), then they'll out-compete their hand-crafting competitors. Of course, you'd probably still outsource some special stuff to the hand-crafting studios, but...

Oh wait, I'm rambling.
You are rambling but that just means you need to refine the rules of your procedural post generator.
 
I imagine the planetary terrain generation could be leveraged to generate the parks, just generate a square piece of terrain and introduce some procedural algorithms. :)
 
Kkrieger is nuts! why did i never hear about that game before! :eek: i have got to play that!

Slawkenbergius, you just revealed to me an interesting business opportunity of making a software that'll generate 3D models and possibly textures and some other things for developers procedurally which can then be exported and used as if they were hand crafted. A rapid development tool, i'm sure it'd sell well. If no one beats me to it in a few years i might give it a shot myself.

If not for rapid development it could speed up concept art creation quite a lot. (A lot of concept artists use 3D models)
 
Last edited:
Indeed, that's why I chose the parks in space-stations as an example regarding what level of granularity the game will finally aim for. You could just use a few hand-crafted parks which get repeated over and over, or you could write a routine that generates different parks depending on various factors. Does it have a lake/does it have a water-feature/is there an open grassy area, what's the tree density/where are the entrances and exits/what route does the path take... And then you could have rules for the different furnishings, like benches in Federation parks are solid and angular, whereas benches in Imperial parks are curvy and ornate. You could even procedurally modify the base meshes for benches and lamp-posts... Does the Imperial Lamppost have a fluted column or a smooth column, and does it have a Corinthian/Doric/Ionic cap?

I've no real conception of how many man-hours it would takes to write even a basic auto-park generator, but I imagine quite a few. And then there are so many parts of the game, it would be hard to lavish detail on all of them without melting our CPUs and blowing the game's budget spectacularly.

It's a situation where I'd love to see innovative middleware solutions develop. Consumers are demanding bigger and more detailed open-worlds, even while these mega-budget games are now seen as failures for selling only 4 million copies. Something has to give. Speedtree has already shown there's a market both in games and movies for their foliage-generation tech. But what if there were middleware solutions for all sorts of stuff. Building generators, street furniture generators, bookshelf generators, mug generators, etc.

Maybe it's a bit early days for that fantasy to play out, but I wouldn't be surprised if the industry headed that way eventually. Companies already outsource asset creation, so if those asset creators find a way of churning out a high volume of product relatively cheaply (procedural generation), then they'll out-compete their hand-crafting competitors. Of course, you'd probably still outsource some special stuff to the hand-crafting studios, but...

Oh wait, I'm rambling.

When FD did their IPO announcement the wrote that there where basically two things they would invest those extra funds into. Elite and the tools around the Cobra engine, which of course will directly benefit Elite. So I suspect that the plan is to build precisely these kinds of tools that later can be licensed out to other developers. Having a game that is more or less a simulation of an entire galaxy would be a pretty good "demo" for these, don´t you think? ;)
 
Back
Top Bottom