Hello... again :rolleyes:

This week I´m going to take a step back from my project (I´m the guy of the infinite-procedural-chunk-tessellated-ugly textured based terrain , yep its a thing :huh: ) and think a bit about it. As any programmer I love the idea of inifinite and procedural, yep that sounds nice but watching recent failures, like No Man´s Sky, makes my think about it.

I´ve been reading some papers related with terrains used in games and other techs, for example this paper from CD Projekt: http://twvideo01.ubm-us.net/o1/vault/GDC2014/Presentations/Gollent_Marcin_Landscape_Creation_and.pdf , and most of the time they use external tools like world machine,terragen and a large etc. They also use some kind of "proceduralness" to place objects on the terrain or to add extra detail.

So my question for you, dear reader is: What techs are being used in games?

NOTE: I´m doing this terrain stuff because I´m at my last year at the university and one of the assesments (the big one , if you know what I mean :cool: ) is a personal project. If you have any ideas related with this, please I´m open to any suggestion/opinion/comment/blame?.

See you!

The link you just provided is pretty close to the gold standard at the moment. Realtime procedural stuff works well for simple things, but can get too complex too quickly to deliver results that might match things like world machine and etc.

If you're looking for a university type research project to do I might suggest Neural Networks. Have already seen them used to approximate normally highly complex things like fluid sim:

Can easily see the same thing being applicable for generating complex, good looking terrain at runtime. Though the current problem of NNs being hard to debug would certainly crop up.

The link you just provided is pretty close to the gold standard at the moment. Realtime procedural stuff works well for simple things, but can get too complex too quickly to deliver results that might match things like world machine and etc.

If you're looking for a university type research project to do I might suggest Neural Networks. Have already seen them used to approximate normally highly complex things like fluid sim:

Hey! I didnt know that you could use NNs for all that stuff, I will check it thanks!

Some links and text on the subject here http://www.gamedevpensieve.com/graphics/3d/landscape .

Thanks for that , thats a lot of interesting info.

Some procedural terrain generation turned into chunk LoD with most detailed version including tessellation, that's about it.

Even though it helps, it's not perfect by any means, writing the rules for generation/item placement is still trial and error and a change in terrain requires a complete regeneration, and since terrain is usually critical to gameplay you may have changes even late in dev...

I was thinking also in QuadTrees, that could also be interesting.

There is this library: http://nothings.org/gamedev/herringbone/

In the first link (2011 paper) there used to be a link to a similar but very similar algorithm for hex tiles, a pdf from Sucker Punch.

Infamous used something to do with hex tiles and constraints to generate its map offline (ie. developers did it once, not each game like in Minecraft or whatever else).