Should we believe the motion resulted from RRA?

[Simon Jeng]

Hi all,

I am reviewing the tutorial of static optimization https://simtk-confluence.stanford.edu:8 ... timization. In this tutorial, Study 2 compares the output muscle activations and residual forces from unfiltered, filtered, and RRA motions. The unfiltered motion leads to the worst result. The filtered motion is better, and the RRA-induced motion is the best. A question is proposed at the end of this Study, which asks "Which kinematic input should be used for the Static Optimization analysis? Explain your reason(s).". Basically, I have the same confusion and the following paragraph is my thought.

We may consider the motion measured by the gold standards (e.g. motion capture system) to be real (although involving some errors). RRA improves results (dynamic consistent), but it alters the measured motion, and the difference is larger than the filtered motion. It may be reasonable to think that RRA reduces the reality of the movement. So,come back to the above question, the filtered motion is a better choice than RRA-induced and unfiltered motion. Is that correct?

Thanks,
Simon

[Mohammadreza Rezaie]

Hi, Personally, I do not use RRA; because you need a full body model, appropriate coordinates' tracking weight and residual actuators for COM. Also, 3 force plates are required to capture all external forces of a stance phase during walking. Due to these challenges, I try to improve the dynamics inconsistencies by spending more time on Scaling/IK. You can find recommended thresholds for residual forces/moments here: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4321112/

Residual forces are<5% of the magnitude of net external force (peak and RMS). Residual moments are<1% COM height times the magnitude of the net external force (peak and RMS).

One alternative is to use OpenSim Moco (markers + GRF tracking) which likely requires a couple of hours to be converged. If RRA is your way, don't miss RRA Best Practices and Troubleshooting.
Hope this helps.