[ODE] Convex Geometry

Harald Schmid harald.schmid at gmail.com
Thu Apr 14 18:58:36 MST 2005


It's a virtual holobench (stereo projection and shutter glasses for
true 3D) which is connected to several trackers, and force feedback
gloves.

Now that you mentioned it: We have set a fixed ratio between collision
detection loops and simulation loops.

We have 500 Hz for the collision detection engine (Solid) and (I
think) about 50-100Hz for the simulation engine (ODE).

Harald

On 4/13/05, Pierre Renaux <prenaux at alias25.org> wrote:
> 
> Did you try to lower the number of updates per seconds (say 60 per
> seconds) ?
> 
> Five hundreds sounds like a lot... Also what kind of application is it ?
> 
> Thanks,
> - Pierre
> 
> >Surprisingly, one contact point is enough for our application. I
> >actually thought that this would be the major obstacle, however it is
> >not.
> >
> >Here's my interpretation: If two objects collide, that will take place
> >over 1/10 of a second. Since collision detection runs 500 cycles per
> >second, I get 50 penetration depth vectors. That's a lot of
> >information to work with.
> >
> >But that's only guessing. Fact is, collision detection and handling works well.
> >
> >Harald
> >
> >
> >
> >On 4/13/05, Pierre Renaux <prenaux at alias25.org> wrote:
> >
> >
> >>Hi,
> >>
> >>If you got only a penetration depth how to you convert that to the
> >>several contact points required by ODE ? Or is that enought to convert
> >>to one contact point ?
> >>
> >>- Pierre
> >>
> >>
> >>
> >>>I have combined ODE with SOLID, which uses GJK. It works perfectly
> >>>well, although I have only the pentration depth to work with.
> >>>
> >>>A convex object is a convex set. And a convex set is defined as
> >>>
> >>>B is a convex set <=> for every x,y in B and any s in [0,1] it is true
> >>>that z = s*x + (1-s)*y is in B.
> >>>
> >>>Or in words: A set is convex if for any two points in the set the line
> >>>connecting the two points is also contained in the set.
> >>>
> >>>I don't know if there is such a thing as a concave set.
> >>>
> >>>Harald
> >>>
> >>>On Apr 12, 2005 7:24 PM, Jeffrey Smith <jeffreys at softimage.com> wrote:
> >>>
> >>>
> >>>
> >>>
> >>>>Stephen say wrote:
> >>>>
> >>>>
> >>>>
> >>>>
> >>>>>Just wondering if anyone has ever experiemented using a trimesh to simulate
> >>>>>a convex geometry (like a bucket for example).
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>A bucket is actually conCAVE. Convex objects are locally spherical everywhere
> >>>>(i.e. no cavities, dips or valleys).
> >>>>
> >>>>But to answer your question, trimesh vs. primitive collisions are relatively stable,
> >>>>but trimesh/trimesh collisions remain problematic.  An implementation of GJK/EPA
> >>>>would fit well into ODE, if you're looking for a "fun" project.
> >>>>
> >>>>-jeff
> >>>>
> >>>>_______________________________________________
> >>>>ODE mailing list
> >>>>ODE at q12.org
> >>>>http://q12.org/mailman/listinfo/ode
> >>>>
> >>>>
> >>>>
> >>>>
> >>>>
> >>>
> >>>
> >>>
> >>>
> >>
> >>
> >
> >
> >
> >
> 
> 


-- 
-----------------------------------------------------------
Harald Schmid
-Student-
Jochbergweg 5
85748 Garching

Email: mail at schmid-harald.de
www: http://www.schmid-harald.de
Mobil: +49 (179) 9113493



More information about the ODE mailing list