[ODE] Jitters and GeomTransforms

[matt stephure]

I have solved the problem with my robot. Turns out of the code that I
used to calculate Euler angles was incorrect, and the robot was
receiving improper information to its PID loop. Thanks for all your
help.

However, the documentation seems a little light on the issue of how
ODE creates their 4x3 rotation matricies. Does anyone know if it is in
column order, or row order? I typically just use 3x3 (Leaving out
entires, 3, 7 and 11) rotation matricies, and I think it's that
section of code that was wrong.

Thanks,


On Wed, 2 Feb 2005 09:55:57 -0700, matt stephure <stephure at gmail.com> wrote:
> I tried scaling up my robot, and the affect of the jitter was
> lessened, but not entirely so. I still think that part of my problem
> is the friction/collision callback setup I'm using. When I drop a
> large (10x10x5, mass=1) brick onto the PlaneGeom, I can detect a
> noticible jitter from that alone. Have you ever experienced this
> problem?
> 
> I'm debating moving to trimeshes if it solves this problem. The
> movement of my robot is not accurate with all the jittering, and I'm
> not sure how much longer I want to play the tweak-compile-test game :)
> 
> Thanks,
> 
> 
> On Wed, 2 Feb 2005 10:14:11 -0600, Andrew Schroeder <schroe2a at gmail.com> wrote:
> > I'm having simalar problems with geom Transforms.  I use one attached
> > to the same body as the head of my robot, and use it's position to
> > drive the location of a third-person camera.  This works really good,
> > the camera doens't affect the mass of the head, you can detect
> > collitions with the camera and stuff and do all the nice things you
> > want a third person camera to to.  However, I find that as my robot
> > moves around, the geom transform jerks around and it's motion is not
> > smooth like that of the body it's attached to.
> > One hunch is that Chris Calef is on the right track with the scaling
> > thing. Maybe when ode calculates the position of the geom transform
> > relative to the body small rounding errors cause the postion to be
> > slightly different each timestep causing jitter and even penetration
> > into the floor.  If this is correct then Chris is right that scaling
> > up the units would HELP, but not solve the problem.
> >
> > If you find any solutions or have more insight, I would be very interested...
> >
> > On Mon, 31 Jan 2005 11:50:37 -0700, matt stephure <stephure at gmail.com> wrote:
> > > Hi all,
> > >
> > > I'm working on a multiple robot simulator project. I'm currently
> > > having the following problem with ODE.
> > >
> > > I'm using a spring/dampening friction style (CFM/ERP calculated along
> > > the fly, along with the dContactApprox1). My solid plain geoms when
> > > dropped sit nice and still on the Plane.
> > >
> > > However, my transforms (created, iirc, correctly) have a jitter. When
> > > I built my robot up the jitter becomes more and more extreme.
> > >
> > > I had less of this problem, or it was less noticable, when creating
> > > large geoms. Most of the parts I'm making now are on the order of
> > > 50cms or so.
> > >
> > > I've tried adjusting the global CFM/ERP and I haven't had much
> > > success. I've added in BodyDampening and played around with those
> > > values, also without much success. Additionally, increasing the step
> > > size does help, but never fully correctly the problem.
> > >
> > > I'm willing to live with some jitter, but I can't have my robots
> > > blowing themselves apart after a while.
> > >
> > > Any insight would be great. Excellent work with ODE in general, I do
> > > not have a physics background, but I have been able to develop fairly
> > > complex simulations -- kudos. If it is a problem with the
> > > 'squishiness' of the plane, how can I make it more solid? My kp is
> > > already quite large.
> > >
> > > Thanks for your time,
> > >
> > > Other variables of interest:
> > >
> > > Max contacts: 10
> > > Step size: 0.002 .. 0.02
> > > Kd = 0
> > > Kp = Large (100000)
> > > Dampening = -0.001
> > >
> > > --
> > > | Matt Stephure
> > > | stephure at gmail.com
> > > +----------------------------------------------------------
> > > _______________________________________________
> > > ODE mailing list
> > > ODE at q12.org
> > > http://q12.org/mailman/listinfo/ode
> > >
> >
> 
> --
> | Matt Stephure
> | Software Developer
> | Alberta Ingenuity Centre for Machine Learning
> | stephure at gmail.com
> +----------------------------------------------------------
> 


-- 
| Matt Stephure
| Software Developer
| Alberta Ingenuity Centre for Machine Learning
| stephure at gmail.com
+----------------------------------------------------------