simtk.org

Hi everyone,

I am using "MoBL-ARMSDynamicUpperLimb-latest" shoulder model and trying to simulate the abduction motion. The model has total 50 muscles. When I ran the static optimization, it solved with very small 'Performance' and 'Constraint violation' values. But I got very noisy muscle forces. I did filtering first the trajectories, then ran IK, and used the output of IK for static optimization, but there was still noise, in the muscles forces (attached figure).

I also tried other ways. I filtered the outpur of IK, and then fitted the cubic spline for the angle values. Then, input these angle values for static optimization, but the muscle forces are still very noisy. I checked the inverse dynamics results. The generated moment for the motion looks smooth.

Does any one has idea what are the mistakes I am doing? Am I doing anything wrong in the actuators file? or is the noisy kinematics/angles?

Any kind of suggestions/help is appreciated.

Thanks,

The result is not surprising. You have 50 unknown parameters and the optimization finds a solution of the under-constraint problem. It does not mean that the solution is physiologically correct, since there are many (infinite) solutions to the problem. Moreover, keep in mind that SO solves the problem at each time instance independently from the previous instances. I would add that the muscle acts as a low pass filter, so despite that you are getting this noisy signal, it would be translated to a more "smoother" force.

Hi Dimitar Stanev,

Thanks for your reply and suggestion.

Could you please explain what do you mean by saying "I would add that the muscle acts as a low pass filter, so despite that you are getting this noisy signal, it would be translated to a more "smoother" force."

How can I do this? Do you mean to add low pass filter for each muscle forces after the SO (i.e. to pass all the 50 muscle forces obtained after SO and just get smooth)? Or, you mean something to add in the main file i.e. to edit muscle in xml file before SO?

Also, if I smooth the muscle forces specifically after the SO, how to verify that muscle forces and joint reaction analysis results are right? From published similar results, and EMG data, right?

I really appreciate your help.
Thanks,

I mean that muscle activation and contraction dynamics behaves as low pass filter anyway. That is that the muscle can't produce force instantaneously.

Don't try to smooth the result since you will not satisfy the constraint that R * f_muscle = tau (R: moment arm, tau: joint required torques). In anyway SO approximates the muscle force when solving the inverse problem. The best way to validate your results is to compare with the envelope of the EMG recordings.