Arm26 CMC example: missing files

[nickbianco]

Steps to reproduce

Ran the CMC example on confluence with the arm26 model.

Expected result

Normal coordinate tracking behavior.

Actual result

Weird tracking result

In the second main section labeled "Changing the Desired Kinematics", the tutorial asks you to modify the inverse kinematics results by replacing the shoulder elevation column with the values from the elbow flexion column. Upon doing this, towards the end of the time window this happens:

I tried setting the shoulder elevation to half the elbow flexion values and that worked fine. But the original instructions should also work since shoulder elevation can range [-90, 180] degrees while elbow flexion is restricted to [0, 130].

Missing control constraints file.

In addition to the tracking error, the sample control constraints file that the confluence example references is missing from the Arm26 CMC subdirectory.

Environment and GUI version

Windows 10, Artifact 2018-11-02-e6f8d11e, AppVeyor build 1.0.1942

[aseth1]

@nickbianco can you run the example in 3.3 and verify that the same behavior is occurring. I saw similar behavior just running forward dynamics in 4.0 where the muscle points seemed to default to the origin when the state is NaN and that is a bug! But, the simulation was also returning NaN in 3.3 so the result may not be a bug although not the best example to be putting out there.

[nickbianco]

@aseth1 just tested on 3.3 and I get a similar error. Towards the end of the simulation the disappears, and the whole model disappears if you try to rotate it.

[aymanhab]

Thanks Nick, these are different ways to handle NaNs. In 3.3 we were allowed to set garbage transforms causing model to disappear, now we're limited to sensible rigid body transforms for bodies but path points are computed as difference between 2 Nans which end up as 0 🤗 the fix should be upstream rather than handling garbage and NaNs in state but that's not on the table for 4.0.

…

On Tue, Nov 6, 2018, 9:11 AM Nicholas Bianco ***@***.*** wrote: @aseth1 <https://github.com/aseth1> just tested on 3.3 and I get a similar error. Towards the end of the simulation the disappears, and the whole model disappears if you try to rotate it. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#1050 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ADyccCybWMdbQBy9pDSJyFVRQm7AE6kPks5uscLTgaJpZM4YPj2e> .

[nickbianco]

Alright. For now we could remove the section from the confluence example where the user edits the kinematics by hand.

[jenhicks]

Make sure the issue of drawing paths with NaNs is still open

[aymanhab]

File is also missing in 3.3 distribution

[aymanhab]

#762

[chrisdembia]

Should we move this issue out of the sprint then?

[nickbianco]

The actuator constraints are missing in 3.3, but it would be nice to include in the 4.0 distribution. The confluence tutorial would benefit from having the actuator constraints section.

[nickbianco]

I created the control constraints file if someone with a GUI environment wants to create a PR:
arm26_ControlConstraints.txt

[nickbianco]

@jimmyDunne do yo thang

[chrisdembia]

This is in progress only because we need to update the opensim-models submodule in opensim-gui.

[carmichaelong]

Looks like the file is correctly placed into /Models/Arm26/ComputedMuscleControl, and the example can be done.

It also looks like it is not in the reference folder (/Models/Arm26/OutputReference/ComputedMuscleControl), but all the other files for setup are in there. Should this file also be added to the corresponding CMC OutputReference folder?

[chrisdembia]

Should this file also be added to the corresponding CMC OutputReference folder?

Interesting... I don't think we should have a duplicate in the repo but maybe CMake should put a copy in OutputReference.

[carmichaelong]

Moving this to verified on windows and started a new issue with things to do beyond 4.0.