OpenSim crashes during ID-Analyses

[Matthias Kunze]

Hi at all,

i've the same problem like described in thread:
https://simtk.org/forum/new.php?forum_i ... roup_id=91

I follow all instructions, but without any positiv results.

I use OpenSim 2.2 for a simulation of hopping. I scaled the existing Gait2354 model for the Scale Tool (with own experimental data) and then I made IK Analysis (seems to be correct). The motion was stored in an *.mot file. I create a own file containing the ground reaction forces in the format which is required.

Now, whenever I try to do an Inverse Dynamic Analysis (scaled model, motion file from IK, grf in a *.mot file) OpenSim crashes within a few seconds after the "Run" without any kind of error-message.

Could anybody help me, please.

Thanks,
Matthias

[Matthias Kunze]

correct link the old thread:
https://simtk.org/forum/message.php?msg_id=3468

[Ayman Habib]

Hi Matthias,

Since I believe we shouldn't crash for any reason, and definitely not abort without informative error messages, please file a bug report and attach the model and data/setup files necessary to reproduce the crash.

If you run in the GUI and set
Edit->Preferences->Debug=3
then you'll get a stack trace in a dialog instead of a full crash. This may help explain the issue so you can work around it before the bug is resolved. In the past many of these issues had to do with the external forces.
Please let us know how your investigation goes.

Thanks,
-Ayman

[Matthias Kunze]

Hey Ayman,

thank for help.
I've send a bug report on the following link:

https://simtk.org/tracker/index.php?fun ... 1&atid=322

But the IK-trial-file and the GRF-file are too large as well. (10 seconds with 1000Hz)
So I upload them to our university server to the following link:

https://webmail.uni-jena.de/upload-data ... RF-mot.rar

Thanks a lot...

Matthias

[Ajay Seth]

This may not be your issue, but there is a bug in 2.2 that if the starting time is precisely 0 and the data begins at 0, that after filtering, the internal start time is slighlty negative (-1e-7) which causes an exception. A quick fix is to start at any time greater than 0. This will be fixed in 2.3!

[Ayman Habib]

Hi Matthias,

I managed to reproduce your problem. The problem is caused by the bad inertial properties of the toe segments. If you load and then save the model, or just adjust the inertia matrix of the toe_segments in the osim file to satisfy the more strict Simbody test introduced in version 2.2 that the diagonal entries of the inertia matrix satisfy the triangular inequality (e.g. .0001, .0001, .0001) then ID runs to completion.

I'll leave the bug open until a more robust error message/handling is introduced in version 2.3 but please let me know if that resolves your problem.

Cheers,
-Ayman

[Matthias Kunze]

Hi Ayman.

Thanks for the answer.
At first I want to mention that I find mistake which I've even made:
- the .osim-file wasn't placed in the same directory as the other files...so yesterday the ID runs (with all files in the same directory...)
- but not if the time range starts at 0.0 ... (what Ajay mentioned in the last message)

Now to your suggestion:
Do you mean I've to change the inertia values in the .osim-file (toe_segments) to this?

<inertia_xx> 0.0010000 </inertia_xx>
<inertia_yy> 0.0010000 </inertia_yy>
<inertia_zz> 0.0010000 </inertia_zz>

With and without this change, the ID runs.
I think the main problem was my directory mistake and the time range bug in version 2.2.
Or am I wrong?
Does the inertia changing cause other (more correct) ID results?

Regards,

Matthias

[Matthias Kunze]

Thanks Ajay,

the start time was one of the problems that causes the crash.

Regards,

Matthias

[Ayman Habib]

Hi Matthias,

I was running a debuggable environment which is slightly different, if ID runs for you then you should be fine (we correct the inertia problem and issue a warning when loading the model if we detect this situation). I expect the tool to either run or to abort if it runs into problem but not produce wrong answers.

The data files you sent me start at 6.0, do you have another data set that starts exactly at 0.?

The issue with exactly 0. time was not present in version 2.2, however, you have to have data (including GRF) that covers all your simulation/analysis. Because of padding/filtering it's advisable to step a little bit inside the data to avoid the end-effects of splining.

Hope this helps,
-Ayman

[Matthias Kunze]

Hi Ayman,

thanks for the notes.
The files in time range 0 -9 are available at the following link:

https://webmail.uni-jena.de/upload-data ... 0-9sec.rar

What do you mean with "data (including GRF) that covers all my simulation/analysis"?
I hope my recorded data satisfy all the conventions of OpenSim.

Thanks very much.

Regards,

Matthias