Inverse Dynamics results off the charts

[Koen Peters]

Hi,

I have been working on a simple (no muscles) full body model. Segment properties are based on anthropometric data, a variety of ball- and pinjoints was used aswell as coordinate limiting forces (based on OpenSim tutorials).

After executing scaling and IK, which both look fine. I encountered some unexpected results from the Inverse Dynamics tool. The results from ID run in the millions (Nm), except for the pelvis, elbow joint and wrist joint. I've checked for a mistake in units (meter/milimeter), double checked my scaling and IK but couldn't find anything which explains te ID results.

Any tips as to where I might look, or things I could've missed which can fix the problem are greatly appriciated.

Kind regards,
Koen Peters

[Stephen Riutta]

Hi Koen,

Here are three things I would check:

1) If you're applying ground reaction forces to the model during inverse dynamics, you may want to check that those values are reasonable, and are in the correct units.
2) Are you using OpenSim's built-in filtering when you're doing inverse dynamics? If the IK tool had a difficult time finding an IK solution with your model and experimental data, it may cause instantaneous "jerks" in the motion. These would show up as large spikes in your inverse dynamics torques.
3) Is the mass of the model reasonable? You could check that with the "Scale" tool.

Hope that helps,

Stephen

[Koen Peters]

Hello Stephen,

Thank you for your reply.
1) The input file for Inverse Dynamics is an .xml file, do you have any suggestions how I can verify its contents? (Opening the file in notepad shows code). I also have a grf.mot file to associate to motion data, values in this file seem reasonable (700 - 800 N in total, while standing still)
2) Yes, OpenSim's built-in filtering is being used during Inverse Dynamics. IK resutls seem smooth, however ID results do seem to contain 2 large spikes (see image). I have tried adjusting filter settings during data processing (prior to running IK), unfortunately results remain similar. The filter currently used in processing data is a 2nd order zero phase shift filter, cut-off frequency = 13.25 Hz, sample frequency = 200 Hz.
3) The unscaled model wheighs 70.6 kg, the scaled model wheighs 74 kg.

Again, thank you for your reply, any other suggestions are welcome as well.

Kind regards,
Koen

[jimmy d]

Koen,

This problems typically occurs when there is added devices/forces attached to model. Bushings, springs and coordinate limiting forces (and even muscles) will be included in the ID calculation unless specified.

There is a 'do not include' list that can be added to ID. Internally we do this with muscle so that their force contributions are not seen. You will need to add this list to take into account your coordinate limiting forces.

Cheers,
-James

[Koen Peters]

Hello James,

Thank you very much for your reply, excluding the Coordinate Limiting Forces from ID solved the problem.
Up to this point I used the analyze tools as provided by the OpenSim GUI. I was unaware of these extended posibilities (e.g. create an ID setup file) with the Matlab-OpenSim Interfaces.

Others who might look to solve similar problems, the Matlab-OpenSim Interfaces can be found under projects on the home page of SimTK

Thanks again,
Koen