Re: [Algorithms] Contact modeling (was: N-body processing)

[David B. <db...@bt...>]

> I didn't notice it, but you mentioned the face/face contact manifold,
which
> I do not detect. I only handle the four others, two of whom as degenerate
> cases. To make things clear I followed the Baraff's way, including how he
> deals with colliding & resting contacts.
>
> How do you detect the face/face case? It is useful? I guess it might be
> useful to handle the simple case of the cube falling vertically on the
> table...

Face/face contact detection isnt strictly necasery, things should work as
long as you detect all the other types of contact. The reason I detect
face/face contacts is that it can improve stability and provide more
sensible contact normals.

For example if you detect four point/point contacts for two cubes resting
exactly on top of each other, typicaly you would pick the average vertex
normals at the points(since a point/point contact is degenerate). A more
sensible choice is the normal of the contact face(top or bottom face of
cube).

The algorithm which I use to detect contact regions is basically a modified
polygon clipper which tracks features along the way... It outputs face/face
contacts just as easily as the other types.

David

http://www.dblack.btinternet.co.uk
ICQ #: 24402391
Mobile: (UK) 0778 7836188