[ODE] Contacts filtering/merging

[Marco Grubert]

Thanks Jeffrey, I did not realized that the Tri collider was changed to use
last frame position. Passing this in helped a lot. Now I just need to find
out why the normals are flipped the wrong way.

- Marco

----- Original Message ----- 
From: "Jeffrey Smith" <jeffreys at Softimage.com>
To: "Marco Grubert" <mgrubert at conitec.net>; <ode at q12.org>
Sent: Wednesday, May 19, 2004 07:10
Subject: RE: [ODE] Contacts filtering/merging


And then Marco Grubert was all like:
> As a follow-up: are there any examples of Trimesh-Trimesh collision
> detection? I am getting some weird normals/depths once I change from
> Trimesh-Box to Trimesh-Trimesh, causing the body to whirl around.

Are you passing in the last frame's transformation matrix properly?
What size timestep are you using?  Are you stacking trimesh objects
on one another?

I heartily recommend filtering and merging contact data that comes back
from dCollideTTL, too.  You could try sorting them by contact depth,
and then marging any contacts with similar depth, normal and location.
As a first pass that should fix some of the collision asymmetry that
results from asymmetric triangulations.

-jeff
-=o virus found by mail at conitec =-