Inverse Kinematics Angle Calculation

[Radhakrishnan Vignesh]

Hi,

I have a question regarding IK:

If the total residual error or segmental residual error (residual error of markers in a segment) increases, but the joint angle error (compared to reference) decreases, why is the objective function having a greater residual error value? In other words, what can cause marker residual error to increase while producing better joint angles?

Are there any correlations betweem them?

Thanks

[Radhakrishnan Vignesh]

Hi,

Any luck with this? I have got results where the residual error decreases but the joint angle error increases. Can you tell me what may cause this? Does an increase in residual error affect the calculated velocities or accelerations more?

The model has been scaled and all model markers placed correctly.

Thanks
Vignesh

[Thomas Uchida]

I have got results where the residual error decreases but the joint angle error increases. Can you tell me what may cause this?

There are many possible reasons. For example, although it may seem like "all model markers placed correctly," small changes in marker registration can cause large changes in kinematics; the IK Tool uses marker weights which may need to be adjusted; there may be error in the data to which you are comparing; etc. The following webinar might be helpful: "OpenSim Webinar: Tips and Tricks for Data Collection, Scaling and Inverse Kinematics in OpenSim", https://youtu.be/ZG7wzvQC6eU.

[Radhakrishnan Vignesh]

Hi Thomas,

Thanks for your reply. But if the inverse kinematics equations adjusts angles such that the marker residuals are reduced, wouldnt it imply that there should be a direct relationship between marker residuals and joint angle errors (increase in residual error implying an increase in joint angle error and vice versa)?

Or am I missing something in the equation or working of IK?

Thanks

[Thomas Uchida]

But if the inverse kinematics equations adjusts angles such that the marker residuals are reduced, wouldnt it imply that there should be a direct relationship between marker residuals and joint angle errors (increase in residual error implying an increase in joint angle error and vice versa)?

If the model and data are perfect, then you should be able to get zero marker error and perfect joint angles (assuming a sufficient number of markers has been used). However, low marker errors with high joint angle errors is also possible. For example, in an extreme case, a single marker on the wrist can be tracked perfectly with an upper-limb model (assuming the marker remains within the model's workspace) but the pose of the arm is unlikely to be correct.

[Radhakrishnan Vignesh]

Thank you. So in your opinion, can we infer the accuracy of the kinematics from residual error (either global or total)? For example, I get an increase in residual error despite the joint error reducing (during swing) and vice versa in stance.

Is there no correlation or relationship between residual error and joint angle error using IK? And would this change if we use MOCO instead (since it ensures dynamic consistency?

Thanks

[Radhakrishnan Vignesh]

Sorry, adding to that question:

Were the guidelines for marker residual error stated in the inverse kinematic page, determined from literature? If not, how was the 2cm rms error decided upon?

Thanks