is this technologically feasable?
Quote:Original post by Codeka
The problem with procedurally generated worlds is that they're boring. The thing about the real world and what makes it actually interesting is that it's not random - there are many processes involved in making mountains and rivers and fjords and so on, none of which are entirely random.
Well procedural and random aren't the same thing... for instance, you could make a mountain by taking a height field and randomly populating each entry. That would make a really crappy mountain.
A better algorithm might use fractal generation (not fractal as in mandlebrot) and give more realistic mountains.
Both are procedural, and heck both could be based on random numbers, but the quality of the 2 algorithms is pretty different.
You could procedurally generate an entire game with the right algorithms... you could even procedurally generate storyline by taking common plot lines, having the program choose which ones to use, splicing them together, and filling in sockets for sub plot lines and characters.
I have the same complaint you have but I think procedurally generated content can be done a lot better, because after all... this world and all the people in it are the result of a more complex algorithm involving physics and strict genetic rules :P
Quote:Original post by Durakken
It's not just like the real world. And haven't you ever heard of Role play?
Sure, but typically people role play fantasies, i.e. stuff they can't do in the real world. Limiting resources like that might make it a bit too realistic. Realism isn't always good.
Author
Quote:Original post by RoboguyQuote:Original post by Durakken
It's not just like the real world. And haven't you ever heard of Role play?
Sure, but typically people role play fantasies, i.e. stuff they can't do in the real world. Limiting resources like that might make it a bit too realistic. Realism isn't always good.
It matters how it is done and also there is such a thing as recycling...
Quote:Original post by Codeka
What you could perhaps do, Durakken, is rather than impose an arbitrary boundary to your world, just make the world outside of you main "area" randomly-generated. That way, a player will be able to travel for days in one direction if they really wanted to, just seeing randomly-generated rocks and trees, but nobody is ever actually going to bother.
Maybe in a single player game, but in a MMO the "frontier" would find its uses: meeting with some privacy and arranging battles in the middle of nowhere are only the two most obvious ones.
The "main area" would inevitably expand into the random zone like a sprawling city, if it isn't already unusually sparse and diffuse thanks to the vast spaces.
Quote:This is precisely why nobody uses true voxel worlds - the storage requirements are enormous.
Depends on how you store your voxels. Lossless compression schemes as the ones images use work just fine.
The simplest lossless encoding consists in replacing series of adjacent voxels with the same value by a number and a single voxel.
Of course that one isn't very nice to compute quickly the value of the pixel at a given coordinate, so you have to do smarter things.
Quote:The problem with procedurally generated worlds is that they're boring.
They wouldn't be boring if the players were allowed to alter it by building their own houses, castles, cities, could cut off trees, dig into the ground, etc.
Basically having a fully constructible and destructible world.
Quote:Original post by loufoqueStill enormous. You can use all the fancy compression techniques you like, but eventually some idiot is going to come along and mine your game world into a close approximation of swiss cheese, and you will be back to the worst case.Quote:This is precisely why nobody uses true voxel worlds - the storage requirements are enormous.Depends on how you store your voxels. Lossless compression schemes as the ones images use work just fine.
Just look at some of the levels Dwarf Fortress players have created...
Quote:To my mind, procedural worlds have two main benefits: a) cheap content creation, and b) low storage requirements.Quote:The problem with procedurally generated worlds is that they're boring.They wouldn't be boring if the players were allowed to alter it by building their own houses, castles, cities, could cut off trees, dig into the ground, etc.
Basically having a fully constructible and destructible world.
Allowing users to alter the world greatly enhances point a), at the expense of point b), since we have to now store this massive amount of generated content.
Quote:Still enormous. You can use all the fancy compression techniques you like, but eventually some idiot is going to come along and mine your game world into a close approximation of swiss cheese, and you will be back to the worst case.
Some parts would be constrained by physics (you cannot have stuff levitating in the air) or game mechanics (you cannot dig into any material without the appropriate strength) but that is true indeed, the user could alter the world to render the compression mostly useless.
Looking back at the number though, 3 billion voxels, i'm not sure this is any serious problem for a server. I definitely wouldn't say it's enormous.
I agree 3 billion nodes isn't beyond today machines, using smart compression schemes only keeping active nodes in memory you can probably fit a cubic mile with a 5-6 gigs of memory.
This reminds me of the isometric games of the past, they used a volumetric representation as well ( in their case iso tiles ). In particular a game which did this well was X-Com Apocalypse. They had simple physics where tiles unconnected to the ground would collapse, of course we now have the computational power to do much more realistic physics.
Instead of just raw voxel, you can use a simple primitive voxel tile set which will allow you more creative freedom. Couple this with procedural world generation and you've gots the making of a whole new complex world.
-ddn
This reminds me of the isometric games of the past, they used a volumetric representation as well ( in their case iso tiles ). In particular a game which did this well was X-Com Apocalypse. They had simple physics where tiles unconnected to the ground would collapse, of course we now have the computational power to do much more realistic physics.
Instead of just raw voxel, you can use a simple primitive voxel tile set which will allow you more creative freedom. Couple this with procedural world generation and you've gots the making of a whole new complex world.
-ddn
Quote:Original post by ddn3The issue at that point is rendering, rather than storage. A cubic mile is certainly plausible from a storage standpoint, but the cost of tessellating and rendering a voxel landscape that size using current techniques is, I fear, beyond our available processing power.
I agree 3 billion nodes isn't beyond today machines, using smart compression schemes only keeping active nodes in memory you can probably fit a cubic mile with a 5-6 gigs of memory.
And don't forget, that standing on the ground, the horizon is about 3 miles away, giving a visible area of 30 m^2, so a cubic mile really doesn't an entire planet make.
Part of the issue is the state of the technology involved - voxel rendering hasn't been popular in games for a long time, and thus the techniques haven't improved at the same rate as polygonal approaches. Eskil Steenberg's Love is one of the few voxel-like games that I am aware of in development currently, and it is a 3-level heightmap, rather than actual voxels.
PolxVox has a nice demonstration of actual voxels using OGRE, but I am pretty sure nowhere close to a cubic mile of them.
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement
