mgrubert at conitec.net
Mon Apr 25 20:03:10 MST 2005
Could someone familiar with the Trimesher-Sphere collider please explain the
contact calculation ?
In particular, I think that the contact->depth value is wrong and the
contact->normal calculation seems odd:
GetContactData() calculates the closest feature distance between a sphere
and a triangle and also returns the triangle normal. Let's assume the sphere
contains one of the triangle vertices and the sphere center is in the
triangle plane, but the triangle center is not inside the triangle. In that
case the calculated distance is _not_ the distance along the
triangle/contact normal but instead the distance along a normalized vector
from that vertex to the sphere center. As long as the triangle center can be
projected onto the triangle itself the calculation is correct, but it seems
to be off in the general case.
In dCollideSTL all contacts are merged by default. For this all normals are
scaled by the contact depth, added together and then the following
calculations are performed:
// Remember to divide in square space.
Contact->depth = dSqrt(dDOT(Contact->normal, Contact->normal) /
Why is the division done in "square space" as opposed to dividing the
magnitude (not squared!) by OutTriCount ?
More information about the ODE