The Future is Procedural

[Shamus Young]

The Future is Procedural

Procedurally generated worlds are offering new possibilities to gamers, but are they set to overtake the handcrafted worlds of old?

Read Full Article

 

[The_root_of_all_evil]

There was the short lived MMO Auto Assault that was a great Mad Max environment, but it suffered from the simple idea of trying to encorporate long range weapons, high speed but keep it within a players frame of reference.

In the end it just turned into "hold down fire button while ramming something".

A procedural MMO....oooh... I can almost wish for such a thing. Hunting alone while going back to the city to sell up, it would be heaven.

 

[j0z]

I am not sure that procedurally generated terrain and environments could ever take off and be as good as those created by humans. The algorithms would have to get very advanced to make it look really good. But if they could do it, and some games do, it would allow such huge worlds.

http://www.infinity-universe.com/Infinity/

Looks like it will be cool, and it is all completely procedurally generated.

 

[zoozilla]

Procedurally generated environments could definitely work for certain games.

Could it work with a Half-Life or some similarly story-oriented game? No way.

Part of what makes games like Half-Life great is the amount of detail put into the worlds - those small details are what make the world feel alive, and really what allows the player to get sucked into the game world.

Procedurally generated environments are the moment are decidedly bland. There's really no personality there, and when telling a story you need the setting to be a major character and convey important information to the player about the world. FUEL isn't trying to tell a story or anything, so however many gazillion miles of dirt works just fine. You don't need much of a setting to screw around in an ATV.

 

[Kilo24]

An article about the future being procedural generation and no mention of Dwarf Fortress? That is procedural generation at its deepest.

We've had a few games rely on procedural generation. They're rare (less so in ASCII games) but they've been around for a while (Elite being one early example.) The challenge is in making a sophisticated enough generator to keep players interested. Making terrain procedurally generated well is very easy in comparison to doing so for characters, social factions, or gameplay mechanics.

 

[Zwebbie]

I very much hope that procedurally generated content won't overtake hand-crafted worlds.

A 14,000km2, but guaranteed that you won't find anything interesting, because a computer won't hide easter eggs, little stories, challenges or whatever else you might find in a hand-crafted world. It all falls apart as soon as you think about gameplay.

GTA's huge maps - San Andreas in particular - were praised at the time. But the fun wasn't so much in being able to drive for hours on end, as much as that you'd find all kinds of stuff. Trail off the road and you might find a neat car, weapon, a location with a story or easter egg. That's neat. The roads themselves carefully balance between towns, mountains, tunnels, hilly landscapes and flat plains, so that you'll never look at the same environment for more than a couple of minutes at a time. San Fierro's road cleverly comes out of a tunnel so that you see the city rise up as out of nothing.

Take FEAR - I know it's odd to call this an example of good map design, but hear me out - FEAR was praised for its AI. In reality, it's not so much the AI that is clever, as the level design. The soldiers will jump from cover to cover and the designers carefully placed this cover so that the soldiers move to a location that is likely to be the player's flank. Not only is this a great example of hand-made map design, it's also a case where hand-made mapping triumphs over procedurally generated AI.

Which isn't to say that I'm entirely against procedural generation - STALKER spawns troops all by itself with great effect and L4D's Director is great too - but I think seeing it as a viable alternative to hand-made worlds is severely underestimating the cleverness that went into making the maps and the gameplay that results from it. Thankfully RAGE is trying to make it a whole lot easier for artists to make unique environments; that's the future!

I still enjoy that 5 kilometer walk to the city and back I make a couple of times a week. It's a short trip, but there's so much to see!

 

[Licaon_Kter]

@Shamus Young: did you see Eric Bruneton rendering inspired by Arthur C. Clarkes' "Rendezvous with Rama" ?
video: http://www.youtube.com/watch?v=zBIQCm54dfY&fmt=18
site: http://ebruneton.free.fr/rama3/rama.html

 

[Nerf Ninja]

And yet I still can't find this game to buy on the PC. I know it exists I've played on my friends copy.

 

[The Random One]

GTA's huge maps - San Andreas in particular - were praised at the time. But the fun wasn't so much in being able to drive for hours on end, as much as that you'd find all kinds of stuff. Trail off the road and you might find a neat car, weapon, a location with a story or easter egg. That's neat. The roads themselves carefully balance between towns, mountains, tunnels, hilly landscapes and flat plains, so that you'll never look at the same environment for more than a couple of minutes at a time. San Fierro's road cleverly comes out of a tunnel so that you see the city rise up as out of nothing.

True - and that's why GTA:SA map is much more interesting than the eternally repeating drab homes of its much less successful counterpart with a larger gamezone, True Crime: Streets of LA. But procedural content could still generate those things, and leave procedural content to the drab stretches of road that need to link those places together.

Also, please see: NetHack, Dwarf Fortress, that will be all.

 

[Shamus Young]

@Shamus Young: did you see Eric Bruneton rendering inspired by Arthur C. Clarkes' "Rendezvous with Rama" ?
video: http://www.youtube.com/watch?v=zBIQCm54dfY&fmt=18
site: http://ebruneton.free.fr/rama3/rama.html

That was excellent. Thanks so much for sharing.

 

[Licaon_Kter]

That was excellent. Thanks so much for sharing.

Sorry for the delay, i meant to post it, in regard to your PixelCity, since May, when i first saw it my self.

 

[Doug]

I am not sure that procedurally generated terrain and environments could ever take off and be as good as those created by humans. The algorithms would have to get very advanced to make it look really good. But if they could do it, and some games do, it would allow such huge worlds.

http://www.infinity-universe.com/Infinity/

Looks like it will be cool, and it is all completely procedurally generated.

True, the problem I find with procedurally generated items is that everything feels very samey - everything lacks uniqueness and variety, which was one of Spore's problem (although, trust me, there where more, and EA/Maxis could have done a better job without making it 'too smart' for the mass market).

That said, these days, with people wanting larger game worlds, we'll probably have little or no choice but to go with this method. The problem I do have is that, although the worlds a bigger, the content and variety is the same. Although I can picture a system, for example, like Mount and Blade, but better;

Define a map size, the locations and rough sizes of some cities, and the rough height map (resolution of say 10 to 200+ metres). Procedurally generate a landscape, and city scapes - realistic sized cities, for once, not 30 houses (aka Oblivion). Include in these procedurally generated and/or randomized NPCs, with an economic network between them all; have the cities belonging to factions, and then from there, we get a complex world of trade, war, and diplomacy from a relativity small amount of data (although the algorithms would be most complex).

EDIT:
Come to think of it, if you combined procedurally generation with genetic algorithms, some of the leg work in developing the algorithms to produce 'good' maps could be reduced. I did experiment with it alot when I was younger, although I only ever applied it to 2D colour patterns - still, a form of selective spawning could produce very good patterns with clear 'species'.

 

[Doug]

@Shamus Young: did you see Eric Bruneton rendering inspired by Arthur C. Clarkes' "Rendezvous with Rama" ?
video: http://www.youtube.com/watch?v=zBIQCm54dfY&fmt=18
site: http://ebruneton.free.fr/rama3/rama.html

Very impressive!

I'm curious as to what data he feed into it, if any, to generate the cities and roads.

 

[Dev Null]

Well in a way its still money spent on graphics; its just money spent on more, smarter graphics, instead of just more detailed...

 

[axia777]

Procedural graphics will never beat custom graphics ever. I don't care what the argument is. It is just never going to happen. Cheaper is not better. I am a 3d modeler and a texture artist. No computer generated crap is ever going to beat what an artist can do as long as that artist has the skills. What procedural graphics are good for are things like trees and bushes. The grunt work no artist wants to do or should be made to do. But for environments and characters procedural is never going to cut it better than the artists.

 

[Galenor]

I really enjoy Procedural technology. The love I have for it stems from a desire of seeing games where you enter a few nuggets of data into a machine, and it makes the level for you. It's such a smart way of making the workload a lot easier for future content.

This is a great example of that, how FUEL makes itself, in a way. It also allows each player to have their own crafted world, unlike anybody elses. Or, if you're bored with your world, make a new one.

Dwarf Fortress, as mentioned, had this. For those not in the know, it had an engine that automatically made a world for your dwarves to reside in. It takes a staggering amount of factors into accord; the years going by, the mountain ranges and marshes, even the amount of heroes dying and relics left behind over the years. The process takes 2 minutes max, and you can see the map changing and forming as the game calculates rivers and trees.

I'm also a huge fan of Left4Dead's zombie spawning engine. You feed the game a map, and it gives you a fully functioning level. It truly has made custom content for the game a lot easier than having a done-by-hand zombie spawner method.

Along these lines, I have a huge crush on 'ARhrrrr' - an Augmented Reality zombie shooting game. You feed the game a map printed out on a big piece of paper, and it automatically generates the level.
On the paper.
Have a look, and remember to pick up your jaw when you leave: http://www.youtube.com/watch?v=cNu4CluFOcw

 

[Ushario]

I couldn't help but think of .kkrieger as soon as you mentioned procedural generation, I have seen your own work and I was impressed. .kkrieger is simply stunning. I was first introduced to the .kkrieger project and procedural generation of content earlier this year.

While I have long known about procedurally generating content such as in the Diablo series or, in one of my longest played games, Nethack .kkrieger really got me interested in using procedural content.

Just like you I see no reason to spend the time creating a world like Cyrodil from a blank canvas. Its possible to save hundreds of hours of work by creating a generator and tools to use it, Bethesda is already half way there with their editor so why not take the next step?

Your point is obvious and rings true, having procedurally generated content that the artists can then modify has two major benefits. Content creation would be faster, which means cheaper and it would provide more game space to the customer. It sounds like a win-win situation to me.

Now I'm off to play some NetHack, its time for a new Wizard, my last one had a bad run in with Team Ant! =(

 

[civver]

Agree with the main idea. Procedural generation will create compelling worlds with less time and effort.

 

[Spleenbag]

Along these lines, I have a huge crush on 'ARhrrrr' - an Augmented Reality zombie shooting game. You feed the game a map printed out on a big piece of paper, and it automatically generates the level.
On the paper.
Have a look, and remember to pick up your jaw when you leave: http://www.youtube.com/watch?v=cNu4CluFOcw

*drools* It's hard to take my jaw up off the floor from that one, mate. My God, that's some cool technology.

 

[KDR_11k]

A problem I see is that the amount of thought that goes into the primary parts of a level is MASSIVE, I don't think it's feasible to program a computer to do that well. Sure, it can easily generate the parts of a level noone cares about but think about it, what are those parts? Filler. The ease of procedural/random (in the end it's the same, just that static pocedural uses a fixed seed) content generation leads to the huge temptation of adding a lot of filler. I've seen it happen especially when Japanese devs get their hands on a random dungeon generator, instead of needing to design intricate levels themselves they'll just throw X random levels in and increase X for every dungeon. Since the random stuff is meant to be disposable filler between the highlights this leads to the game getting more and more filler to the point where the player gets bored.

Also please don't point at kkrieger, the procedural generation used in that was only to save disk space. Everything in it was made by hand but instead of directly painting anything in photoshop they used a special application that built the textures out of manually selected and configured filter stages. It's not meant to save effort.