[ODE] Jitter at the end of a chain of bodies

Tim Sorrells tsorrells at breakawayltd.com
Wed Sep 20 15:18:44 MST 2006


I'm simulating a tube with a chain of bodies attached with universal
joints.  I've had success with a chain of 15 objects.  However when I
put 21 objects end-to-end I get a substantial amount of jitter.  It
starts about 5 bodies back and gets progressively worse culminating in
the worst jitter at the end body.  The jitter for all the end bodies is
in sync, that they all change shape the same way, just to a greater
magnitude toward the end.

 

I'm using the inertial dampening code posted by Jon Watte.  My app uses
single precision floats internally to render the objects, however I
tried switching ODE from single to double precision, but that didn't
help.  Switching from dWorldQuickStep (100 iterations) to dWorldStep had
no effect either.  I run with a constant timestep of 0.05 and I tried
lowering it to 0.01, but it also had no effect.  My bodies aren't
colliding with anything.  I've tried using a fixed joint at the end, but
that doesn't change the amount of jitter.  Changing ERP and CFM didn't
change much.  I did notice that the jitter doesn't affect the
auto-disable feature and once disabled the jitter disappears.  If I take
the matrices from ODE and re-othogonalize them the jitter is reduced to
a barely noticeable (and quite acceptable) level.  But I'd rather not
have to do that to every ODE body in the interest of speed, so I'm
wondering what else I might try to help?

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://q12.org/pipermail/ode/attachments/20060920/f002e438/attachment.htm


More information about the ODE mailing list