simtk.org

Hi all,

I am trying to run CMC on Arnoldetal2010_2Legs.osim model with own experimental gait data (scaling, Ik and RRA worked fine, and I had quite small residuals after running RRA twice), I got warning message in computing initial values and almost in every timestep.

e.g.
CMC.computeControls: t = 0.18
WARN- small force range for patlig_r (0.008113 to 0.00814743)
WARN- small force range for patlig_l (0.0114803 to 0.0114827)
WARN- small force range for pect_l (0 to 0)


I had to increase reserve and residual optimal forces to avoid optimization failure at initial time, and it took really long time to only computing initial values (about 20mins). The optimization finally failed at about 45% (not sure how long that took, few hours at least).

What can I do about the warning message? I suspected it affected the computing time.

Thank you in advance!

Ruoli

Hi Ruoli,

I'd suggest you contact the model author and describe the activity you're trying to simulate so as to figure if the model is appropriate for the activity, how long it takes to run and what adjustments need to be made to the model if any. I've seen cases where the warning messages occur but CMC runs to completion without problems.

Please let us know if you continue to have questions and good luck.
-Ayman

Hi Ayman,

Thank you for your suggestions.

The CMC run to completion when I increased few 'weights' in the CMC_task.xml, but the warning message still appeared in every computing step. I will consult authors for this part.

I have done a further induced acceleration analysis, and the superposition generally (e.g. COM) looks good, however, compared to the experimental accelerations, the total induced accelerations were much bumpy. Is that common? The controls after CMC were kind of bumpy also. Is it the reason? How can I improve it?

Thank you in advance!

Ruoli

"WARN- small force range" just means that at that time point in the simulation, the force produced by the muscle is not affected by the activation. In other words, no matter if the activation is 0 or 1, the force produced by the muscle is about the same magnitude.

This can happen for two reasons (that I know of): (1) The fiber length is in a place on the force length curve where the muscle cannot produce much force, or (2) The fiber has a high shortening velocity, such that it cannot produce much force.

- Sam

Hi Ruoli,
I don't think this is what is affecting your compute time. This message appears in patlig because that muscle's FMax is 0 (it's a hack from when there weren't ligaments in OpenSim). It's true that
there's a small force and that's appropriate. I'm not sure why it's appearing in pectineus, except that that small muscle is quite sensitive to scaling. This results in what can best be described as
"chatter" in the muscle's length and velocity, which could put you in the ranges like Sam suggested. I have observed slow down simulations. Since, at the end of the day, it doesn't contribute much
to the motion you're trying to model, I'd try taking it out completely and see if things improve. You can either take it out of the .osim file completely or switch the isDisabled tag to true.

Finally, this is a slow model in CMC. Your computer is newer and faster than mine, but two hours is probably not unreasonable, based on the computation times I was seeing in version 2.2.1. They were even longer. This is because this model has a lot of wrapping surfaces and they really slow things down in CMC, especially when you have multiple simulations or are iterating on control constraints. This is an ongoing area of development for the team, but in the mean time, I've
worked around it by setting up batches using matlab scripts and leaving them to run over night.

I hope this helps. My scripts for batch processing are over here: https://simtk.org/home/modelerskitchen

-Edith