Query related to UMocoD Project

[Karthick Ganesan]

Dear Ross Miller,
Could you please share the fiber type ratios of ercspn, recabd and fdb muscles? I went through the references. But I could get only some of their values and not sure of the exact values used in the model. I need them for metabolic cost computations.
Also could you please mention the relationship used for calculating muscle activation time constants? I couldn't get it right from the reference.
Thanks & Regards,
Karthick.

[Christopher Dembia]

Karthick,

If coming by such data proves difficult, you can perform a sensitivity analysis to assess how much the fiber ratio affects your final results. It's possible that the fiber ratio has a small effect on your results.

[Karthick Ganesan]

Thanks Chris. If I don't get the exact values, I will do as you suggested.

[Ross Miller]

Hi Karthick,

The figure here has all the FT values for the muscles in that model. You can also back them out of the value of <activation_time_constant> (I call this tau1):

tau1 = (5+0.2*m*(1-FT)^2)/1000
m = muscle mass = Fo*Lo*rho/sig
Fo = maximum isometric force (N)
Lo = optimal CE length
rho = muscle density (I used 1059.7 kg/m^3)
sig = specific tension (I used 620,000 N/m^2 from the Rajagopal model)

The reference for the time constants is Winters & Stark (1985) where they define them as functions of FT (actually ST=1-FT) and muscle fiber mass.
https://ieeexplore.ieee.org/abstract/do ... SJkKW5AijA

Generally I find these values don't effect results that much for walking mechanics because in OpenSim's muscle models the value of FT is not coupled to the curvature of the force-velocity curve, although it will affect metabolic cost a lot in some energetics models.

Hope this helps,
Ross

Screen Shot 2020-08-12 at 6.08.52 PM.png (277.37 KiB) Viewed 986 times

[Karthick Ganesan]

Dear Ross Miller,
Thank you very much for sharing all the values and for the detailed answer.

Karthick.

[Karthick Ganesan]

Dear Ross Miller,
I went through the publication associated with the experimental data used as reference trajectory in the simulations. It seems like the ankle trajectory is not the mean trajectory from the paper. It appears to be above the mean. Does this corresponds to some specific trial or only the ankle angle is different from mean for some reason? As I loosen the weight on tracking, I wanted to plot the angles along with standard deviation and see that they are within the standard deviation.
I digitized the mean ankle angle from the paper, smoothed using moving average and tried it in the simulations. It did not converge within 2000 iterations. May be due to digitization or approximating the left leg ankle angle or something else.
Thanks,
Karthick.

[Ross Miller]

Hi Karthick,

The ankle data in the paper (Miller et al., 2014, MSSE) were "normalized" to the standing calibration trial (aligning all the segment axes with the global axes when the subject was in their "neutral" standing posture). This is not how OpenSim defines/computes joint angles, so for the ankle angle I shift the baseline of the data so that the model's ankle angle around heel-strike is similar for the data around heel-strike. For the hip and knee I don't find this is necessary but for the ankle it makes a big difference in tracking simulations. Without this I will often get weird looking ankle angles and patterns of ankle muscle activity in the model.

Ross

[Karthick Ganesan]

Oh. Thanks Ross Miller for the explanation. That was really helpful.
Karthick.

[Karthick Ganesan]

Dear Ross Miller,
In the 'goWalk_v2.m' code, it was mentioned that "Starting from the initial guess provided and running this code as-is, the optimization converged after 100 iterations". When I ran this code as is with the provided initial guess, it converged after 364 iterations. Does the hardware(processor/RAM) or software (OS/Matlab version, etc) could change the number of iterations? Or 'after 100 iterations' could mean more than 100 iterations?
I understand that hardware configuration will affect the time of each iteration.
Thanks,
Karthick.

[Ross Miller]

Hi Karthick,

I would imagine such things could vary depending on hardware and software, but I don't know for sure.

Ross