[ODE] Collision detection for arbitrary polytopes ?

[Pierre Terdiman]

And I'll take this opportunity to say that Opcode 1.2 has faster sphere-mesh
tests (but still unreleased, no time - old song, I know).

Two things provided huge speedups (in case someone can't wait and want to
implement them) :

- sphere-mesh queries can early exit as soon as the sphere totally contains
a box of a node (hence we know it also contains all boxes of all children
and we can directly dump the list of related triangles)

-  temporal coherence can help by using a slightly larger sphere for the
query, and keeping track of touched triangles. Next frame, if the previously
used larger sphere still contains current sphere, no need to do the query,
we can reuse the list of triangles from the previous frame to compute the
collision response. We'll get more triangles than strictly necessary, but
overall it runs faster.

It also works for other volumes of course, not only spheres.

Pierre

----- Original Message -----

From: "J. Perkins" <jason@379.com>
Cc: <ode@q12.org>
Sent: Thursday 13 June 2002 15:55
Subject: Re: [ODE] Collision detection for arbitrary polytopes ?


> Nate W wrote:
> > http://q12.org/cgi-bin/cvsweb.cgi/ode/contrib/tri-collider/
> >
> > I am using it to do collision detection against height-mapped terrain,
and
> > it's working nicely.
>
> I'll take this opportunity to say that OPCODE 1.1 makes this
> integration with ODE a lot easier. I had no trouble getting
> a sphere-mesh test working within a day. If I ever get the
> chance to add tests for the other primitives I will package
> it up as a contrib, but for now you can get the source from
> my project at:
>
> http://www.379.com/f4/
>
> Look here:
>
> F4/Services/Collider/Src/mesh.cpp
>
> (Thanks to Pierre for the lib and his help getting it working).
>
> Jason
>
>
>
> _______________________________________________
> ODE mailing list
> ODE@q12.org
> http://q12.org/mailman/listinfo/ode