"Stanford VA upper limb model" : Problem with tools.

[Nicolas Eydieux]

Dear OpenSim community,

For my research project, I am working on the Stanford VA upper limb model in order to study the manual wheelchair propulsion. I have had not any problem in editing bodies' inertial parameters, scaling the model and running inverse kinematics. I even had pretty good results.

However, when I try to launch a static optimization or a RRA or CMC analysis, OpenSim crashes just before the beginning of computation. Actually, the loading bar stays at 0%, it is impossible to cancel the process and the "Messages window" shows the following lines, for example with static optimization :
"Creating states from motion storage
No external loads will be applied (external loads file not specified).
Executing the analyses from 0.01 to 1..."

I am quite confident in .xml files that I use for these analyses, so I really don't understand where is the problem. Maybe the model is not adapted to these tools ?

Thanks for your helps,

Best regards,
Nicolas Eydieux.

[jimmy d]

Hi Nicholas,
The older Simm models always have significant issues for running simulations. Massless bodies, old muscle models, Simm wrapping and Simm constraints all result in OpenSim crashes. The model you are using requires some significant upgrades to be able to run muscle driven simulations.

A group that did get a version of the model to run in Static optimization did the following;

"Upper-limb geometry and muscle paths are unchanged. Changes to the original model include;

- Muscle models were updated to Thelen (2003) type muscle definitions.
- Maximum contraction velocity of all muscles changed from 0 to 10
- Removed massless bodies
- Removed or changed some wrapping surfaces "

Once you edit your model I would suggest doing inverse dynamics and Static optimization. If you cant get those to work you wont be able to get RRA and CMC to run.

cheers,
-james

[Nicolas Eydieux]

Hi James,

First, thank you very much for your quick answer.
As you suggest, I will upgrade this model and I will keep you informed.

Thanks again for your advice,
Nicolas.

[Nicolas Eydieux]

Hi James,

I have updated the "Stanford VA upper limb" model as you suggested. I have even been obliged to modify the articular kinematics which seems to be impossible to reproduce without massless bodies...The Inverse Kinematics calculation is clearly faster but despite these modifications OpenSim crashes during the calculation of Static Optimization (at only a few percents).

Perhaps OpenSim does not support the different sort of WrapObjects, which are WrapEllipsoid, WrapCylinder,WrapSphere and WrapTorus...

Thanks per advance,
Nicolas.

[jimmy d]

OpenSim does support those wrap objects, however there may be an issue in how the wrapping surfaces are interacting with the muscles.

What muscle model are you using? the latest Thelen muscle model?

Other muscle models may cause a crash easily if they are on weak areas of their force length curve.

-james

[Nicolas Eydieux]

Hi James,


I use the Thelen2003 muscle model which is used in the arm.osim model without any problem...

Interaction between muscles and wrapping surfaces are defined differently in muscle <PathWrapSet>, depending on which muscle is concerned. As an example :
<PathWrapSet>
<objects>
<PathWrap>
<wrap_object>TMIN</wrap_object>
<method>hybrid</method>
<range> 1 2</range>
</PathWrap>
<PathWrap>
<wrap_object>INFSP_and_TMIN_hum_head</wrap_object>
<method>hybrid</method>
<range> 1 2</range>
</PathWrap>
<PathWrap>
<wrap_object>TMINhum</wrap_object>
<range> 1 2</range>
</PathWrap>
</objects>
<groups />
</PathWrapSet>


Despite I have reused the definition of wrapping surfaces in the arm.osim model (an example in the following xml code), I may have troubles of definition...

<WrapSphere name="LAT_TMAJ2hh">
<xyz_body_rotation> -0.127235 0.276111 0.0197222</xyz_body_rotation>
<translation> -0.0016 0.0092 0.0052</translation>
<active>true</active>
<quadrant>-z</quadrant>
<VisibleObject>
<!--Set of geometry files and associated attributes, allow .vtp, .stl, .obj-->
<GeometrySet>
<objects />
<groups />
</GeometrySet>
<!--Three scale factors for display purposes: scaleX scaleY scaleZ-->
<scale_factors> 1 1 1</scale_factors>
<!--transform relative to owner specified as 3 rotations (rad) followed by 3 translations rX rY rZ tx ty tz-->
<transform> -0.127235 0.276111 0.0197222 -0.0016 0.0092 0.0052</transform>
<!--Whether to show a coordinate frame-->
<show_axes>false</show_axes>
<!--Display Pref. 0:Hide 1:Wire 3:Flat 4:Shaded Can be overriden for individual geometries-->
<display_preference>0</display_preference>
</VisibleObject>
<radius>0.03</radius>
</WrapSphere>

I'm quite suprise that Static Optimization continu not to work after all these upgrading...
If you have any idea I will be very glad to manage to make this model running.

Thanks,
Nicolas

[Nicolas Eydieux]

Hi James

Here is my upgraded model. It is composed of Thelen2003 muscles and just of few massless bodies which describe shoulder movement and which are allowed because of massful bodies connected outboard (according to Michael Sherman). I just change extension from .osim to .xml to be able to attach it.
In addition, I send you the simple motion I have to use to run StaticOptimization.

I hope that with these files you will be able to better understand my problem.
Thanks per advance,
Nicolas.