simtk.org

Hello All,

I feel as though this is a silly question because I have seen the answer to it right in the OpenSim documentation, but my results don't appear to match that representation.

I know that for IK, the documentation states that the maximum marker and coordinate errors are given in meters and degrees. However, using the Gait2392 model and my collected data, I run IK and associate the motion with the corresponding .trc data file to visually compare and though my maximum errors are saying ~0.25m (Frame 1177 (t=11.7283): total squared error = 0.154099, marker error: RMS=0.0981387, max=0.24699 (L.Thigh.Upper)) which is 25 cm which obviously is unacceptable, when I associate the motion file to try and troubleshoot why there error is so great, the data only appears to be off by a couple of centimeters at most and certainly not 25 cm! (attached image is the last frame of data corresponding to the results I included above).

Am I misunderstanding either the documentation or the association of motion data?

Thanks!
Kathleen

Maybe at some point the marker position is way off. You can try to increase the task weight for this marker and see if the max error improves.

the data only appears to be off by a couple of centimeters at most and certainly not 25 cm

The projection of the model into the port view can be misleading some times.

Thanks for the suggestions. I played around with the weights a bit and one of the two following things happens: the gait gets really wonky looking, or the gait pattern doesn't change much visibly and the max error either stays about the same or goes up.

I also checked to make sure my scaling didn't do something weird where my skeleton was enormous or scaled inappropriately (my subject is smaller than the skeleton and everything appeared to scale down appropriately.

I've also attached a front view of the same frame reported on earlier (last frame). As for your comment about the view being misleading, does this help clarify?

The maximum marker error is literally just the distance between the OpenSim marker and the actual marker, correct? To try and troubleshoot some more, I put a new temporary marker in the GUI and aligned it with the associated marker on the thigh. I then took the coordinates of each and found the distance between the two and got ~2.4 cm (which doesn't match up in magnitude or actual number as I thought may have been the case).

I should also note that the gait cycle looks visually appropriate and the knee flexion angle looks as I would expect.