Questions about scaling and inverse kinematics

[Yongtao Lu]

I am working on a project about musculoskeletal (shoulder) modeling and using OpenSim in this project. From the website, we downloaded several musculoskeletall shoulder models, e.g.Dynamic Upper Limb and set up 6 markers on the models based on the experiment set-up. However, after scaling, the markers location is changed and not similar to the original setting and experimental setting. We tried to adjust the locations of the markers, but it still did not work. Total squared error = 0.00364992, marker error: RMS=0.024664.
Then I load the motion data collected from experiments to calculate the Inverse Kinematics. But the simulated motion is different from the experimental motion. The angle between the trunk and arm is obviously different.The errors in every frames are around 0.02.
Are the errors in scaling step and Inverse Kinematics step enough low? How low do you prefer us to control the tolerance?
What kind of faults could cause these problems?

[Dimitar Stanev]

If you scale the model correctly you will get less IK error obviously. Can you be more specific on the scaling outcome? Did you try to preview the .trc file and to see if there is some problem. Is the .trc correctly aligned in case that your model is grounded?

Best

[Yongtao Lu]

Hi Dimitar,

Thanks a lot for your reply. Indeed our scaling did not work well. After scaling, the locations of markers are moved and does not match those of experiment set-up. We tried to adjust the attachment locations (to the bone) of the markers, but could not get good results. Can you please give us some hints on how to solve the problems? Is this due to too few markers (we have only 6 markers in the model)? Is the initial attachment position of the markers important for the scaling process?

Best,
Yongtao

[Dimitar Stanev]

It is hard to tell. Can you post a fig, where both the model and the .trc are visible? Also, check whether the model is grounded.

[Yongtao Lu]

[img]

limb.png (35.01 KiB) Viewed 1621 times

[/img]
The right one is before scaling. The left one is after scaling. The error is 10e-4.
[img]

limbexp2.jpg (194.49 KiB) Viewed 1611 times

[/img]
It is the experimental example.
The marker setting changed obviously.
[img]

limb.png (35.01 KiB) Viewed 1621 times

[/img]
It is one of the frames of motion, IK computing after scaling. The angles of joints are weird. The error is 10e-3.

[Yongtao Lu]

by the way, what do you mean 'the model is grounded'? How to check/set-up this?

[Dimitar Stanev]

Hi,

Please show me the model and the .trc together. You can preview the .trc from File->Preview Experimental Data. What I mean is if your .trc file is displaced and if the model has its base fixed to the ground, then you can't perform IK correctly.

Best

[Yongtao Lu]

Hi,
Thanks for your explanation. But I don't understand how to fix the .trc to the model.
This is the Experimental Data and the model markers I set.[img]

trc.png (28.49 KiB) Viewed 1588 times

[/img]
I guess the experimental arm is not perfectly straight.
I set the markers by my eyeballing, which is based on the joints' shape. Because we set marker on the bone protrusion in experiment.
Is there any better method for setting markers and let it fixed to the ground ?

Looking forward to your reply.

[Dimitar Stanev]

Ok. It seems that if you unlock the base body (suppose thorax) as free joint then you will be able to perform IK. Otherwise you would have to translate and orient the .trc motion.

[jimmy d]

Relevant Post on Scaling and Marker Placement