[ODE] RE: [CsMain] Penetration depth/exact collision

[Darken, Chris]

It is also my understanding that exact penetration depth calculation for
arbitrary meshes is difficult.  See my previous posting for paper and code
references.  However, I am wondering what would happen if we fed a ODE a
lower bound on the penetration depth, say specifically the penetration depth
of any two triangles that get returned from a collision detector like RAPID.
If we're lucky the penalty we'd pay for the approximation would only be that
it takes an extra frame or two for an interpenetration to be removed, and
not even that if we make sure that objects aren't traveling fast enough for
a deep interpenetration to begin with.  I was hoping to experiment myself,
but haven't found time yet.  Any comments from those in the know?
Chris


FROM: Daniel Duhprey
DATE: 04/17/2002 15:16:21
SUBJECT: RE: [CsMain] Penetration depth/exact collision
On Tue, 16 Apr 2002, Anders Stenberg wrote:

-->could add this for me I'd be really thankful. Right now, I'm pretty 
-->stuck with the physics, but it'll probably not take too long finishing 
-->if someone would fix this little feature in the CD-system.

Making an arbitrary mesh depth penetration is unfortunately not a little 
feature... you may want to just limit physics to only the basic types used
by 
ODE and support ODE's collision detection... The i* interfaces CS provides 
would fit sorta neatly on top of ODE's CD.