Differences of ID and IK in OpenSim 3.3 and 4.0

[Wei Wang]

Hi all,

I found the inverse dynamic result from the OpenSim 3.3 GUI and OpenSim 4.0 GUI are different when I use them to analyze the exactly same model and IK file(generated from OpenSim 3.3 GUI). The values are slightly different (e.g., around 0-2Nm).

Moreover, when I use OpenSim 4.0 to calculate the inverse kinematics using the same TRC file as in OpenSim 3.3, the results are also different. (For some spinal joints, varied like 10 degrees).

This thing only happens to my model (a complex model but without constraints and force elements). For other models, like Gait2392, the IK and ID results from the two versions are the same.

Anyone knows if there is any difference in the inverse dynamic and inverse kinematics solver in the two OpenSim versions?

Thank you and kind regards,
Wei

[Dimitar Stanev]

The problem may be due to some change in the configuration of the tool. Without providing the data to reproduce the problem it is hard to tell.

[Wei Wang]

Hi Dimitar,

Thanks for your reply! I have attached the model and the marker data. Could you check this for me?

Could I also ask for your understanding that after the post, the data would be removed due to data protection?

Wei

[Dimitar Stanev]

Small differences are also observed. Can you tell me which coordinates have the largest difference in your simulation?

I would expect this, since the model has many degrees of freedom and the two versions of OpenSim can use different optimization settings, thus they would result in different solutions that satisfy the motion.

[Wei Wang]

Hi Dimitar,

The largest IK difference was mainly at the spine level, especial for L5_S1_ FE (around 9 degrees).

Wei

[Dimitar Stanev]

I found 5deg difference. As I told you, this is to be expected since the model has many degrees of freedom and bodies. The optimizer finds a different solution, either because it starts with different settings or different initial conditions. The tracking error is low, so the algorithms finds an aceptable solution that tracks the motion well. The problem may probably require more markers to find a unique solution.

[Wei Wang]

Hi Dimitar,Thanks for your explanation. I got it.