Oscillations in joint torques

[Norbert Braun]

Hi all,

I just noticed a strange issue in cartwheel-3d: sometimes, the generated
joint torques seem to oscillate widely. A plot is at:

http://www.ikp.uni-koeln.de/~braun/8db0196bdfe6d4b3f5084933e2fb31fc/torqueplot.png

The plot was generated by outputting, in ODEWorld::advanceInTime(), the
current time and the joint torques (i.e. jts[j]->torque), once per
simulation step.

Has anyone else noticed this? Any idea about what might cause it? Or am
I merely misunderstanding something here?

Best regards,

Norbert

[Stelian Coros]

Hi Norbert,

This is probably a sign that some gains (PD or maybe virtual forces)
are too high relative to the size of the time step that is used.
Decreasing the former (especially the damping coefficients) or
increasing the latter should help.

Best,
Stelian

[Norbert Braun]

Hi Stelian,

On 2012-10-31 12:58, Stelian Coros wrote:
> This is probably a sign that some gains (PD or maybe virtual forces)
> are too high relative to the size of the time step that is used.
> Decreasing the former (especially the damping coefficients) or
> increasing the latter should help.

I tried playing with the PD gains and decreased the time step. While the
oscillations do change somewhat, I could not make them disappear.

Do you, or anyone else, happen to have a set of working parameters to
get me started? The graph I posted in the last mail uses the default
character and the default settings, so I am not doing anything exotic
here...

Thanks,

Norbert

[Stelian Coros]

Hi Norbert,

I am unfortunately using a different version of the control code, so
the same values that I am using would be doing something completely
different in your framework. Did you try scaling down these values by
a factor of 10 or so?

Best,
Stelian