RRA ends with incomplete results

[Alex MacIntosh]

Hello,

I am using the upper extremity model in OpenSim Version 3.1, and applying an external force at the finger tip. When I try to use the RRA tool, it runs for 0.02s (of the requested 9s) and then finishes.

Through the command line, with verbose printing on, it completes without an error message and outputs an incomplete states.STO. In the GUI, with the Debug preference set to 3, it returns "Tool execution failed or canceled by user. Output files were not written.", along with an incomplete states.STO file.

I have read that the upper extremity model is not particularly apt for dynamic assessments, and have made some adjustments to the model: I have updated the model to have mass and inertial properties, and have also changed the ground_thorax weld joint, to be more similar to the ground_pelvis joint in the gait2354 model (i.e. a custom joint with 6 DOF).

I have also tried to replace the Schutte1993Muscle_Deprecated muscle model with Thelen2003Muscle. They both run for the same amount of time, however with the Thelen2003Muscle the RRA tool ends without any particular message, and with the Schutte1993Muscle_Deprecated the message reads:

"Schutte1993Muscle_Deprecated::computeForce, muscle DELT2 force < 0 at time = 0" (and the same for multiple muscles).

I am able to run ID, however the results are not what was expected. (There is an increase in the joint moment throughout the trial, the cause of which I am not sure of yet).

I have thought that perhaps RRA is not appropriate for this. My overall goal is to get joint moments that match the experimental results, and ultimately muscle contributions that create those joint moments during the fingertip loading.

If anyone has any suggestions it would be a great help.
I would be happy to provide my work to help troubleshoot as well.

Thanks

Alex

[Ayman Habib]

Hi Alex,

A couple quick notes:
1. I'm not sure why you're using RRA to begin with. We use it in gait/running situations to account for errors in measurements by adjusting mass properties to make f=ma. I'm not sure why you're using it for fingertip simulation. Regardless, a successful run of RRA (if you specify an adjusted model) should result in a new model with modified mass properties.
2. Muscles are (internally) removed from the model while running RRA so changing Muscle types should have no effect on RRA behavior.
3. It is true that most upper extremity model OpenSim models floating around have been developed a while ago and were not designed with Dynamics in mind.

I'd suggest you start with a simple model and build up the complexity only as needed for your problem and let us know if you run into unexpected results. This will make it easier for you and us to pinpoint problems and/or explain results without dealing with a complex/unknown model.

Hope this helps,
-Ayman

[Alex MacIntosh]

Hi Ayman,

Thanks so much for your quick response!

I must apologize for the ambiguity. In reference to replacing the muscle types, I have tried this for when I run Static Optimization, and CMC (which are also not producing expected results).

Initially, I was hoping to use RRA to reduce the residual forces and moments I was getting from Inverse Dynamics. However, if RRA is more tailored to gait type scenarios than I will not use the tool.

You are right, I have been decreasing the parameters in the Upper Extremity model, trying to tease out what the issue is, but it will probably be more effective to take a bottoms up approach for this simulation.

I will definitely let you know how it goes, or if I run into unexpected results.

Thanks again!

Alex

[Alex MacIntosh]

Hello,

Thank you for your suggestion Ayman, it was helpful indeed. I have simplified the model to only include the 4 muscles that control the MCP joint of the index finger. From this, I have been able to run Static Optimization and receive reasonable results. However, when I try to run CMC for the same motion only the initial states gets computed, and then the tool fails (without an error message other than that it failed and output files were not written).

I have gone through the suggestions in the CMC guide, but it remains unsuccessful:
http://simtk-confluence.stanford.edu:80 ... +Practices
(some of these were however helpful in running the SO)

Any suggestions would be really appreciated.
I would be happy to provide my work to help troubleshoot as well.

Thanks,

Alex

[Ayman Habib]

Hi Alex,

What muscle model are you using? It would be extremely helpful if you file a bug report and attach all necessary files to reproduce the problem so that we can help you troubleshoot, since we should at least give a meaningful error message if we can't solve.

Best regards,
-Ayman

[Alex MacIntosh]

Ayman,

I am using the Thelen2003Muscle model.

I have submitted a bug report as per your request (ID# 1882-CMC computes initial result then closes without error).

Thanks so much for your help.

Alex