Scaling

[Ary Silvano]

Hi,
My name is Jia Cheng Xu from Swedish National Road and Transport Research Institute (VTI). We have experimental data from volunteer tests subject to acceleration/braking pulses. After scaling, the resulting scaled models have different stance compared to the volunteers' standing postures (compared with experimental photos that is), e.g., longer femur, over-flexed knees (increased knee angle).

Questions:
In general the below is asking if the error values are required to follow the recommendation despite if anthropometry and kinematics look good when compared to experimental pictures and videos.

1. Is it a must that the marker errors are under the recommended threshold (maximum 2-4 cm), during scaling and IK? If the marker placement and anthropometry look good so to speak?

2. With respect to scaling and IK:
If the scaled model has similar anthropometry when compared to experimental photos, but the markers are placement weirdly; are we allowed to manually place those that are misplaced i.e. is this a reasonable and acceptable procedure?

If the motion looks similar to experimental videos, but the errors are too large, will the resulting kinematic output be reliable? How is the IK results affected by errors above the threshold?

We have tried with increasing weights for markers that are misplaced and it improves misplaced markers a bit but are still quite misplaced. We have thought of running the scale tool iteratively by applying it to the saved scale model and repeating the scaling process but with no luck. We figure that if the scaling errors are too large then IK will get affected and in turn affect ID, resulting in OpenSim output that are not reliable.

[jimmy d]

There is a paper and Webinar on verification and validation that you should spend some time going through.

[Olivier Vandebroucke]

Thanks for posting these, I hadn't found them earlier and because of that, it was rather difficult to get SimTK to do what I wanted it to...