Is there any method to reduce RRA FY value?

Provide easy-to-use, extensible software for modeling, simulating, controlling, and analyzing the neuromusculoskeletal system.
POST REPLY
User avatar
Clarise Choi
Posts: 8
Joined: Wed Jul 04, 2018 9:09 pm

Is there any method to reduce RRA FY value?

Post by Clarise Choi » Sun May 16, 2021 9:06 am

Since I use 2392 model for walking.
some of the model has large FY value after running RRA
Is there any way that I could solve this problem?

I hope to solve it for better result!
Thanks

Tags:

User avatar
Aaron Fox
Posts: 291
Joined: Sun Aug 06, 2017 10:54 pm

Re: Is there any method to reduce RRA FY value?

Post by Aaron Fox » Sun May 16, 2021 6:07 pm

Hi Clarise,

I've noticed a number of studies running multiple iterations of RRA with the adjusted model and kinematics to progressively reduce residuals (i.e. run RRA, update your model and kinematics, re-run RRA, repeat until desired). There's probably a ceiling effect with this though. I find that if I'm not getting the desired level of residuals after 2 or 3 iterations then there is potentially an issue with the scaling or experimental data that needs addressing before RRA.

Aaron

User avatar
Clarise Choi
Posts: 8
Joined: Wed Jul 04, 2018 9:09 pm

Re: Is there any method to reduce RRA FY value?

Post by Clarise Choi » Mon May 17, 2021 2:56 am

Thank you for advice...

But I don't get the method that update model and kinematics.
How can I update with my experimental data?

thank you

User avatar
Aaron Fox
Posts: 291
Joined: Sun Aug 06, 2017 10:54 pm

Re: Is there any method to reduce RRA FY value?

Post by Aaron Fox » Mon May 17, 2021 3:12 am

Hi Clarise,

One of the outputs of RRA will be a new set of kinematics (it will have '_q' to end the filename I think). Those become your 'updated' kinematics. RRA will also output an 'adjusted' model if you tell it to, with the centre of mass of one of the main bodies (typically the torso) altered based on the RRA results. The one thing RRA doesn't do automatically is update the mass of the bodies in the model from the changes recommended by RRA - so you need to do this manually in some way. There are some good accessory MATLAB functions included with the Rajagopal et al. gait model that can help do this.

So after the first time you run RRA, you get your updated model and kinematics - and you also still have your experimental GRFs. If you run RRA again with the updated kinematics and model + the experimental GRFs, and potentially repeat this multiple times, then theoretically this iterative process will help reduce the residuals.

Hope that makes sense.

Aaron

User avatar
Clarise Choi
Posts: 8
Joined: Wed Jul 04, 2018 9:09 pm

Re: Is there any method to reduce RRA FY value?

Post by Clarise Choi » Mon May 17, 2021 8:01 am

Thank you so much for the explanation.
As my understanding, in 1st RRA process check the adjust model -> run RRA - > rerun with adjusted model and input at Desired kinematics with RRA_q.sto file..
Is it right? (I'm still working on the MATLAB code for weight)
Somehow my only problem FY getting larger and also FX get larger...
And if FY value reduced then, the model walk with abnormally..
Did I misunderstand your explanation?:cry: :cry: (Any advice?)

Thank you again

User avatar
Aaron Fox
Posts: 291
Joined: Sun Aug 06, 2017 10:54 pm

Re: Is there any method to reduce RRA FY value?

Post by Aaron Fox » Mon May 17, 2021 5:25 pm

Hi Clarise,

I think you've understood the process. RRA is difficult though as there are also the different task weights for the various coordinates to consider too (i.e. changing how heavily you weight different coordinates will also affect your solution and how much you can reduce the residuals).

I find FY always seems to be the biggest issue and this has a lot to do with the mass of the model not aligning to the forces produced by the model - and the largest magnitude of force is in that vertical direction. Going back and making sure your scaling is good can help with this.

Another option to consider are the new Moco tracking simulations. You can aim for a dynamically consistent simulation by having no residuals and allowing the kinematics to change to a certain extent based on different weights applied to the coordinates. I'm starting to shift to a process of running one RRA iteration to get an OK solution with residuals, and then shifting across the the Moco tracking simulations to solve for a solution with no residuals. It's obviously a new tool to learn but could be useful.

Aaron

User avatar
Clarise Choi
Posts: 8
Joined: Wed Jul 04, 2018 9:09 pm

Re: Is there any method to reduce RRA FY value?

Post by Clarise Choi » Mon May 17, 2021 6:44 pm

Thank you!
If the difficulty keep occurs, then I should try to learn new thing..
Sorry for keep asking about RRA..
I could not make adjust model for some reason..
It keeps failing when I check the adjust model..
Is any situation like this happened with your RRA process?

Thank you

User avatar
Aaron Fox
Posts: 291
Joined: Sun Aug 06, 2017 10:54 pm

Re: Is there any method to reduce RRA FY value?

Post by Aaron Fox » Mon May 17, 2021 7:15 pm

Hi Clarise,

Not sure exactly what's occurring with respect to the adjusted model failing - I know it's a checkbox you have to select and add some settings to in the RRA process. My advice would be to start with the RRA example(s) that come packaged with the gait models in OpenSim, as you know these work. Rhen you can just adapt this RRA example to your own data and models, or compare it to what you're currently doing to try and identify the error.

Aaron

POST REPLY