RRA - excessive Fy & Fz residuals near take-off

[Nicos Haralabidis]

Dear users,

I am currently using the Rajagopal et al. (2016) model, and I have collected three-dimensional kinematics and kinetics from the block phase of a sprint. External kinetics have been collected for each of the arms and legs separately, and applied to the respective hand and calcaneus segments.

I have managed to perform RRA, and my average residuals after torso COM and kinematics adjustments following a single iteration are promising:

Fx = -34 N
Fy = 16 N
Fz = -5 N
Mx = -1 Nm
My = 6 Nm
Mz = -1 Nm


However, my residuals do not look as promising from inspecting their values during the trial.

RRA_residuals.PNG (152.63 KiB) Viewed 346 times

(1) I was therefore wondering if anyone had experienced similar issues with regards to the residuals during take-off? If so, how did you troubleshoot it?

(2) I am also surprised by the magnitude of the Fx, Fy and Fz residuals - as in my RRA actuators file their optimal force is set to 100, but they clearly exceed this. How is this possible?


I look forward to your replies!

Kind regards,

Nicos Haralabidis

[Kenneth Smale]

Hi Nicos,

I've also gone through RRA and had issues with the residuals being very high. I was looking at jump landing and my residuals were high prior to landing (i.e. any GRF being read by the force plates).

How do your rotational and translational errors look? Do they also satisfy the suggestions outlined by the online documentation and Hicks 2015 paper?

What really helped me was running this RRA optimizer from Dr. Michael Samaan. It performs RRA through a Matlab API and optimizes the weights of each coordinate to reduce the error. https://simtk.org/projects/rra-opt-gui

[Nicos Haralabidis]

Hi Kenneth,

Thanks for your prompt reply!

In relation to your first point, are you suggesting if the force platform is measuring low forces when the participant is not on the platform this causes spikes in the data? Should I therefore set the force values to 0 when the participant is not on the platform? Does the ID or RRA tool have a threshold for when I can apply the forces to the segments?

I have not explicitly checked the positional errors yet as I wanted to reduce the residuals. I have compared the .trc file associated with the particular trial and they have been manipulated, but not unrealistically from a qualitative perspective. I shall check the article you suggested, thanks!

Thanks for sharing this, I shall definitely check it out!

Kind regards,

Nicos Haralabidis

[Kenneth Smale]

Hi Nicos,

Yes, I had issues when there were no GRF present to account for the movement/accelerations of the bodies in space. These left very large "hand of God" residual forces at the pelvis but through the RRA optimizer I provided a link to, they were reduced down below the recommendations.

I really recommend checking the residual simultaneously with the new kinematics. If you overlook the kinematics and only focus on the residual forces/moments, you can have major kinematic errors that go unnoticed especially if you are working through an API and not directly in the GUI where you can see the animation. (I'm not sure if this is your case).

RRA actually outputs a nice pErr.sto file with the results that help you visualize the errors between the original and RRA kinematics. Also plotting the original .mot file coordinates against the RRA q.sto coordinates gives a nice visual representation of the errors. The online OpenSim error tolerances are 2cm for translations and 2deg for rotations.

Good luck!

[Nicos Haralabidis]

Hi again Kenneth,

I have generated the ID setup files and external loads through the Matlab OpenSim API, but I have ran the tools through the OpenSim GUI. The latter two points you make I will definitely take on board - really good advice! Thanks a lot for your input!

Kind regards,

Nicos Haralabidis