Hi all,
I am investigating the early acceleration phase in sprinting.
However, I am not able to reduce the residuals to the recommended range without distorting the kinematics.
I use the gait2392 model and start with the actuator weights from the file "gait2392_RRA_Task.xml".
Depending on the deviation of the kinematics I adjust the weights. I also adjust the body mass as recommended.
Does anyone have a tip what I could do?
Thank you so much!
Best regards,
Daniel
RRA - How to reduce residuals without distorting kinematics?
-
Daniel Streif
- Posts: 1
- Joined: Tue Oct 22, 2019 5:56 am
-
Israel Luis
- Posts: 11
- Joined: Thu Oct 24, 2019 3:21 am
Re: RRA - How to reduce residuals without distorting kinematics?
Hi Daniel,
I am no expert but I have some comments to say about it:
- RRA changes kinematics by definition. So I think the question is "how much kinematic difference is allowed". There is a publication that talks about it, if you google: "Is my simulation good enough? OPENSIM" I think they mentioned that around 10-15% of the ground reaction force is admissable in the residuals. Have you checked this?
- There will be more error because of the nature of the explosive motion you are analyzing. So I think that aiming for a 2 cm error is not realistic (if you are trying to do that). You have to see what are the admissible values for such a motion previously reported.
- 2392 is a model developed for walking, not running, or even sprinting, so you would have to adjust the MTU parameters or use a model that can handle such high forces when using a Muscle Redundancy Solver (such as MOCO or CMC).
Finally, I have been in the same place, changing weights to get the "best results". But this is also affected by your scaling and IK weight values. "Good enough" results should be feasible if you carefully scale your model. I would argue that if you have experimental data of walking, start there first (since kinematics of that motion is well known), so then you know that your scaling is good enough, and then move to sprint
I am no expert but I have some comments to say about it:
- RRA changes kinematics by definition. So I think the question is "how much kinematic difference is allowed". There is a publication that talks about it, if you google: "Is my simulation good enough? OPENSIM" I think they mentioned that around 10-15% of the ground reaction force is admissable in the residuals. Have you checked this?
- There will be more error because of the nature of the explosive motion you are analyzing. So I think that aiming for a 2 cm error is not realistic (if you are trying to do that). You have to see what are the admissible values for such a motion previously reported.
- 2392 is a model developed for walking, not running, or even sprinting, so you would have to adjust the MTU parameters or use a model that can handle such high forces when using a Muscle Redundancy Solver (such as MOCO or CMC).
Finally, I have been in the same place, changing weights to get the "best results". But this is also affected by your scaling and IK weight values. "Good enough" results should be feasible if you carefully scale your model. I would argue that if you have experimental data of walking, start there first (since kinematics of that motion is well known), so then you know that your scaling is good enough, and then move to sprint
Re: RRA - How to reduce residuals without distorting kinematics?
Hi Daniel,
You could also consider trying this tool out for your RRA step:
https://simtk.org/projects/rra-opt-gui/
From memory it allows you to apply various weights of important for kinematics, residuals etc., and runs an optimisation that identifies task weights that minimise the residuals and changes in kinematics. I believe this groups published a couple of papers on this too (one here: https://pubmed.ncbi.nlm.nih.gov/27145118/).
Aaron
You could also consider trying this tool out for your RRA step:
https://simtk.org/projects/rra-opt-gui/
From memory it allows you to apply various weights of important for kinematics, residuals etc., and runs an optimisation that identifies task weights that minimise the residuals and changes in kinematics. I believe this groups published a couple of papers on this too (one here: https://pubmed.ncbi.nlm.nih.gov/27145118/).
Aaron