I''m having a hard time getting to grips with how to start with a cliff-face generating algorythm. My situation is pretty simple: 1) I have a height-mapped terrain. 2) Each point on a height-map has a height property and a cliff-height value (0-10). 2.1) For each tile (2x2 collection of points) on the height-map that has the same cliff-height value, the geometry is created as a normal quad. 2.2) For each tile that has differing cliff-heights, I need to create cliff faces. So, the following area 00 00 would be a normal quad with vertices generated by the height-field. This area 01 00 would be created as a flatish bit at the bottom and a cliff segment in the top-right corner. My problem is that I have no idea how to go about actually creating the cliff-face. My general algorythm goes something like this: :Assume we are here because we have a cliff: 1) Check all four points to find the lowest cliff-height 2) With the lowest cliff-height, generate a standard quad using either the height of the point if it is at the lowest cliffheight, or the Height - (MyCliffHeight - LowestCliffHeight) if you are any other height. (This gives me a base quad) 3) Find the next highest group of cliff-heights and create a cliff-face for them as well as a cliff-top for that area. The problem is, how can I figure out the outline for the continuous cliff area? View from above, 01 00 Would be something like a box |_ (a single continuous edge) 01 10 would be more like // (two edges) At the moment I am hard-coding the 12 possible patterns and then using pattern-matching to find out which is the appropriate cliff-face to use, but this is nasty and inflexible. Anyone have any ideas?

OK, but the problem with that is that the cliff is forced to have an incline -> you can not do sheer cliff faces.

For example

0----1-2----3
| | | |
| | | |
4----5-6----7

where the cliff between 1-2, 5-6 becomes near-vertical because the horizontal distance is contracted.

So, my plan is to work on adding geometric complexity to a single tile, therebye giving me far more control (my cliff can take on different characteristics other than just texture). If I define the control points for a cliff tile like this (all other tiles being just two polys):

0-------a------1--------------2--------------3
| | | |
| b | | |
d c e f | |
| g | | |
| | | |
4-------d------5--------------6--------------7

the lettered points give me additional control points from which I can draw vertical cliffs. A corner cliff would be
0-------a------1--------------2--------------3
|=======| | | |
|=======b | | |
d=====c/ e f | |
| g | | |
| | | |
4-------d------5--------------6--------------7

A top cliff

0-------a------1--------------2--------------3
|==============| | |
|=======b======| | |
d-----c/ \e----f | |
| g | | |
| | | |
4-------d------5--------------6--------------7

and a diagonal cliff would be

0-------a------1--------------2--------------3
|=======| | | |
|=======b | | |
d-----c =\e----f | |
| \g======| | |
| |======| | |
4-------d------5--------------6--------------7

If I introduce a randomness factor to the location of the control points, I can do walls (rand=0) or craggy chalk cliffs (rand=1).

I'm only showing a single cliff tile here. Obviously, all the adjacent tiles with differing cliff heights would also have cliff faces, like this:

0-------a------1--------------2--------------3
|=======| | | |
|=======b | | |
d-----c =\e----f------\ | |
| \g======|======| | |
| |======|======| | |
4-------d------5--------------6--------------7
0-------a------1--------------2--------------3
| |======|======| | |
| b======|======| | |
d \e----f------/ | |
| | | |
| | | | |
4-------d------5--------------6--------------7


[edited by - SoaringTortoise on October 15, 2003 8:01:51 AM]

Sorry for the messed up ascii art. Forgot about vs char [32].

Yah. I''ve been using the Warcraft III editor as a reference project, and they seem to be using mesh tiles for their cliffs (with some fancy coding to pin them to the terrain VB). I tried going the same way and couldn''t get the alignment to work right, so I switched to the sub-division method which works quite well.

By being able to effectively cover a cliff-height using a horizontal distance 1/4 the length of a normal tile, the cliff faces appear far more vertical.
[from side]
Normal
\
o\
oo\
ooo\
oooo\

Subdivided
__
00|
00|
000|
000|_

Plus, by assigning weights to each vertex in the subdivided tile I can get different characteristics -- Convex, concave, linear and near-vertical -- by simply changing a couple of a parameters. Another parameter can control "vertex zones" to introduce a degree of randomness to the cliff. That is, instead of saying point is at 0,0 you can say point is in the range of -0.1,-0.1 and 0.1,0.1.

All in all, a pretty neat solution. Of course, it does end up using 32 polys instead of 2, but since cliff faces are comparatively rare it''s an acceptable performance hit.