convert OpenSim 3.3 file to OpenSim 4

[vahid F]

Hi
I want to use my models and dataset in OpenSim 3.3 through OpenSim 4. So I converted my model and kinematic data using Convert files in OpenSim 4.
I ran CMC to calculate metabolic cost but my result in OpenSim 4 has a great difference with result in OpenSim 3.3 (as can be seen in the following figure). Is this difference reasonable?
Thank you

metabolic.PNG (33.4 KiB) Viewed 725 times

[Thomas Uchida]

Is this difference reasonable?

What is the source of the difference? Before jumping to metabolics, you might want to check joint angles, reserve actuators, muscle activations, etc.

[vahid F]

Thanks for your reply.

As can be seen in the following figure, the angle of "knee_angle_beta" has much difference in OpenSim 3.3 and 4 because it changed from Rotational to Coupled in the new version.

1.PNG (14.91 KiB) Viewed 413 times

I checked reserve actuators; their value is small and in the good region. But control value for some muscles such as "gastrocnemius muscle" reaches the maximum value and it is the source of difference between metabolic in version 3.3 and 4.
can you guide me on how this problem can be solved?

2.PNG (15.05 KiB) Viewed 413 times

Thank you in advance for any advice or tips

[metin bicer]

Hi,

I have a similar issue. Are you using Rajagopal's model (https://simtk.org/projects/full_body)?

In fact, for "knee_angle_r_beta" plot, the blue curve is just the scaled version of the red one in my case. The constant relationship between blue and red curves is pi*180. That means, the red curve (version 3.3) is in degrees, and the blue curve (version 4) is in radians. Although the inverse kinematics reports the angles in degrees, for this coupled motion "knee_angle_r_beta", it is expressed in radians. Is it the same in your case?

Using the ik output motion file of version 4, there is no visible problem in the simulation. However, static optimization results in a similar situation. Setting the knee_angle_beta of right and left legs in degrees (just multiply by 180/pi) does not solve the problem and resulted in very high patellofemoral joint dislocations.

I changed the muscle models from "Millard2012EquilibriumMuscle" to "Thelen2003Muscle" by keeping the attachment points and muscle parameters (max isometric force etc) the same. Now, the so results are better although it is different than the version 3 results due to different muscle models. I hope it helps in your case.

However, I still wonder the reason why it is not working with the original muscle model ("Millard2012EquilibriumMuscle").

Thanks

[vahid F]

Hi Metin

I'm also using the "Rajagopal's full_body" model. I found a linear relationship for "knee_angle_beta" between versions 3.3 and 4 (*180/pi), as you said. it seems that "knee_angle_beta" in the last version reported in degree but in the new version states in radian. Although I expected that "knee_angle_beta" has different value between the OpenSim 3.3 and 4 due to its change from Rotational to Coupled in the new version.

I will appreciate any guides that could help me to solve this problem.
Thanks
Vahid

[Ayman Habib]

Hi all,

The least error prone approach would be to get the model into version 4.0+ (either download the most recent version if it supports 4.0 or load the old model and save it in the new format). Ideally you have all the data/setup files to reproduce your results in the current version, so re-run on a clean slate. if not make sure to use a utility updatePre40KinematicsFilesFor40MotionType that we provided to convert older kinematics files, however this only takes care of coordinates and not other quantities that may have changed as a side-effect of the motion type change.

Hope this helps,
-Ayman