Water in games?
Author
I have really been thinking about how they do water in games lately. In my game/engine I would like to have an island surrounded in water, and perhaps an inland lake. I have been thinking of the various ways of actually drawing the water.
I think the best method by far would be to use large patches of triangles, kind of like a heightmap, but for water. The water could have waves, which would look a lot better than a giant textured plane. The problem with using patches of triangles for water is that they would need to interact with eachother. Suppose the one chunk of water had a wave in it, and that wave was going to splash into the next chunk of water. How would this be handled?
I recently had the chance to see a game called ''Jak and Daxter'' for the PS2. I think they did a pretty good job of the water in the very early levels. The water had waves, and wasn''t square against the shore. How is it that water is done in games like this, and like Halo?
er i think to make this problem not an issue you could just not use patches just use one big patch and put it in a quadtree or something.
-David-
-David-
Gamasutra had an article on this, i practically passed-out when reading the math. And i think it''s very hard (read impossible) to do realtime.
Article
----------------------------
"Quotes are cool" - Me
Check out my game Asteroidz, it rules !!!
Article
----------------------------
"Quotes are cool" - Me
Check out my game Asteroidz, it rules !!!
gamasutra''s article was for crazily realistic water - the screenshots looked like photographs, almost.
There is a kind of simple method that Baldur''s Gate and some other games use. They represent the water as a heightfield, as you suggested, and then in every frame, they go through each segment of water and move it towards the average of the segments around it. To make a wave, all you have to do is depress one square and let go. The next frame, the segments surrounding it will go down and the original segment will go up. The next frame, the third ring of segments will go down, and the inner segments will behave in the same way they did before - eventually the wave will start to repeat, and you get a ripple.
To make a straight wave, just depress a whole line of segments to begin with, etc. This works pretty well for modelling characters walking through ponds and such, but it has some limitations - waves can never crest, for example, and you''ll have to use a separate particle engine or something if you want splashes.
There is a kind of simple method that Baldur''s Gate and some other games use. They represent the water as a heightfield, as you suggested, and then in every frame, they go through each segment of water and move it towards the average of the segments around it. To make a wave, all you have to do is depress one square and let go. The next frame, the segments surrounding it will go down and the original segment will go up. The next frame, the third ring of segments will go down, and the inner segments will behave in the same way they did before - eventually the wave will start to repeat, and you get a ripple.
To make a straight wave, just depress a whole line of segments to begin with, etc. This works pretty well for modelling characters walking through ponds and such, but it has some limitations - waves can never crest, for example, and you''ll have to use a separate particle engine or something if you want splashes.
Author
Yeah, I have seen that article before, but I never took a very deep look into it (sorry for the pun
). It is pretty crazy looking water. I think i will do something simple like specular lighting with a sphere map. If I wanted something a little more crazy looking I would go for a bump map and gloss mapping, perhaps with a cube map thrown in.
The problem is connecting the patches. David20321 had an interesting suggestion by using one giant patch with a quadtree or something. The problem with this is that it would be hard to determine where the water was under the terrain, and when to render it. It could end up taking alot of time telling whether the water was under the terrain or if it was above.
Moe''s Site
). It is pretty crazy looking water. I think i will do something simple like specular lighting with a sphere map. If I wanted something a little more crazy looking I would go for a bump map and gloss mapping, perhaps with a cube map thrown in.The problem is connecting the patches. David20321 had an interesting suggestion by using one giant patch with a quadtree or something. The problem with this is that it would be hard to determine where the water was under the terrain, and when to render it. It could end up taking alot of time telling whether the water was under the terrain or if it was above.
Moe''s Site
I think you shouldn''t store every vertex of the water, but rather generate them from a "wave" algorithm. Just give a starting point and height, and move through the "square" from that point on. And just use rectengular pathces ^_^
-Maarten Leeuwrik
=================
[Illiad] J.R.R TOLKIEN ACHIEVES 40,000 RPM IN GRAVE
-Maarten Leeuwrik
=================
[Illiad] J.R.R TOLKIEN ACHIEVES 40,000 RPM IN GRAVE
I''ve also been looking into the water thing. Besides the (really nasty) article on gamasutra, you may want to check out the past programming contests on flipcode - there was one devoted to water effects. Also search on www.siggraph.org. Some really good stuff there.
---
Matt Benic
"Programming today is a race between software engineers striving to build
bigger and better idiot-proof programs, and the Universe trying to produce
bigger and better idiots. So far, the Universe is winning."
---
Matt Benic
"Programming today is a race between software engineers striving to build
bigger and better idiot-proof programs, and the Universe trying to produce
bigger and better idiots. So far, the Universe is winning."
Author
Yeah, I remember the contest at flipcode. I will have to check that out in greater detail and perhaps fire of a few e-mails to the coders of the entries.
Moe''s Site
Moe''s Site
The article on Gamasutra isn''t that evil. It''s pretty neat, if you get the concepts. And you don''t have to implement every bit they described, since it is a collection of different wave synthesis algorithms, just pick the one you need. I tried the first one, the FFT. The math isn''t that bad, it''s just more or less standard fourier transforms with a frequency/phase generator. And the results are extremely good, you can do some shortcuts (replace some of their expensive equations with simpler and faster ones) and it will still look very good.
I also tried to write a 2D Navier Stokes solver. It works rather nice (very realistic interaction with objects), but unfortunately it is rather unstable. Making it stable with the method they suggested (adaptive relaxation) dropped the FPS from 52 to 3...
The shallow wave thing in the article is somewhat overkill IMO. I didn''t try to implement it.
I also tried to write a 2D Navier Stokes solver. It works rather nice (very realistic interaction with objects), but unfortunately it is rather unstable. Making it stable with the method they suggested (adaptive relaxation) dropped the FPS from 52 to 3...
The shallow wave thing in the article is somewhat overkill IMO. I didn''t try to implement it.
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement
