[ODE] unstable objects

[Todd Wasson]

I'm having some issues with object instability in a program I'm working
on, using ODE.  I've got a model of a warehouse and a table.  The table
is constructed of five geom boxes; a top and four legs.  When resting on
the ground, which is just modeled as a large box, it appears unstable
and jitters around a bit.

I first suspected that collision detection might be the culprit.  After
some investigation, it turns out that I'm seeing some aberrent
edge-on-edge collisions (the edge of a table leg and the edge of the
large box it rests on, several meters away) that certainly shouldn't be
happening (due to the fact that the colliding edges aren't even near one
another) but they don't seem to be the issue.  I've culled them out with
some hard coded bounding boxes to see what happens, leaving only some
very nice and clean collisions at the feet of the table, but still, it
jitters.  Even with the collisions left in, the penetration depths are
on the order of a hundredth of a millimeter, and probably aren't
resulting in any significant forces, or motion, anyway.

So, here is some basic data that may prove to be useful: the positions
on the collisions and physical objects are not near to the origin;
they're at about 130,000 x 26,000 x 10 meters (units).  Centering them
about the origin for computational purposes is possible, but since any
loss of precision will have already occurred, it's probably not going to
make a difference.  Also, the jitters are non-constant: sometimes, the
table appears more restless than others.  If it's at rest and it gets
nudged, often times it'll end up jittering more than it was originally.

Oh, and I'm using ODE's collision detection, and hooking it into the
physics modeling in the standard way, i.e. via contact joints.

So, if anyone has any clue as to what's going on, or can provide any
insight, or if you need more information, please don't hesitate to say
something.

Thanks.


Todd