Page 1 of 1
Model does not move to match experimental data
Posted: Wed Jan 20, 2021 9:06 am
by pablo_delgado
Good morning,
I am trying to scaled a model and then apply inverse kinematics to it with some experimental data that I have. However, after the scaling my model does not move to match the experimental data (please look at the image). It looks like the model is locked/fixed at the origin, if I offset the model, the coordinates of my experimental data also move with it.
Any idea of what the issue might be? I would really appreciate the help.
Regards,
Re: Model does not move to match experimental data
Posted: Wed Jan 20, 2021 10:24 am
by tkuchida
Perhaps some of the coordinates are locked. Please see the "Coordinate Controls and Poses" page in the User's Guide:
https://simtk-confluence.stanford.edu/d ... +and+Poses.
Re: Model does not move to match experimental data
Posted: Wed Jan 20, 2021 1:56 pm
by pablo_delgado
Thank you very much for the information, however, the coordinates are unlocked. Please look at the image.
Re: Model does not move to match experimental data
Posted: Wed Jan 20, 2021 2:06 pm
by tkuchida
In that case, unfortunately, it isn't obvious what the issue is. Perhaps the Scale Tool setup file has not been configured correctly, perhaps the experimental data are in the wrong units, perhaps the measurements have not been correctly defined, perhaps the weights are incorrect, etc. There is a tutorial that runs through the scaling process ("Tutorial 3 - Scaling, Inverse Kinematics, and Inverse Dynamics",
https://simtk-confluence.stanford.edu/d ... e+Dynamics); you could go through the tutorial and compare with the data and strategy you are using. The following webinar may also be useful: "OpenSim Webinar: Tips and Tricks for Data Collection, Scaling and Inverse Kinematics in OpenSim",
https://www.youtube.com/watch?v=ZG7wzvQC6eU.
Re: Model does not move to match experimental data
Posted: Wed Jan 20, 2021 4:00 pm
by pablo_delgado
tkuchida wrote: ↑Wed Jan 20, 2021 2:06 pm
In that case, unfortunately, it isn't obvious what the issue is. Perhaps the Scale Tool setup file has not been configured correctly, perhaps the experimental data are in the wrong units, perhaps the measurements have not been correctly defined, perhaps the weights are incorrect, etc. There is a tutorial that runs through the scaling process ("Tutorial 3 - Scaling, Inverse Kinematics, and Inverse Dynamics",
https://simtk-confluence.stanford.edu/d ... e+Dynamics); you could go through the tutorial and compare with the data and strategy you are using. The following webinar may also be useful: "OpenSim Webinar: Tips and Tricks for Data Collection, Scaling and Inverse Kinematics in OpenSim",
https://www.youtube.com/watch?v=ZG7wzvQC6eU.
Thank you very much for the information. I have actually gone through both, the tutorial and the webinar. They are very helpful, however, the problem is that the model does not move around to match the experimental data, like if it was locked at the origin. I have also used the "preview experimental data option" to check the units and orientation of my data and they look correct. Also, I unfixed the markers, then when I scale the model, the markers move to the markers from the experimental data, however, the body of the model does not move. I think that the problem is due to the markers, I went through en example using the already defined markers on the model and it moves to match the experimental data. However, if I save the markers as a file, and delete the markers from the model, then add the markers using the scaling tool the model exhibits the same issue that I am facing using my markers.
Re: Model does not move to match experimental data
Posted: Wed Jan 20, 2021 6:57 pm
by aafox
Hi Pablo,
I've had similar problems to this with gait models during scaling when the limits on the coordinates don't allow the model to move to the position described by the markers. For example - I had a static calibration trial where a participant faced in the -X direction, but the axial rotation of the pelvis was limited to -90 to 90 degrees. Due to this, the model couldn't do a complete turn to face the direction the static trial represented.
It looks like the model you are using doesn't have coordinates describing the motion of the torso relative to the lab coordinate system (or perhaps these are welded), and that your experimental markers are in fact much higher than the model. If you are experiencing a similar situation to me - perhaps it's that the model cannot shift the torso up to match your experimental data given that the torso may be welded in place?
Hopefully that makes sense.
Aaron
Re: Model does not move to match experimental data
Posted: Thu Jan 21, 2021 7:16 am
by pablo_delgado
aafox wrote: ↑Wed Jan 20, 2021 6:57 pm
Hi Pablo,
I've had similar problems to this with gait models during scaling when the limits on the coordinates don't allow the model to move to the position described by the markers. For example - I had a static calibration trial where a participant faced in the -X direction, but the axial rotation of the pelvis was limited to -90 to 90 degrees. Due to this, the model couldn't do a complete turn to face the direction the static trial represented.
It looks like the model you are using doesn't have coordinates describing the motion of the torso relative to the lab coordinate system (or perhaps these are welded), and that your experimental markers are in fact much higher than the model. If you are experiencing a similar situation to me - perhaps it's that the model cannot shift the torso up to match your experimental data given that the torso may be welded in place?
Hopefully that makes sense.
Aaron
Hello Aaron,
Thank you very much, it does make sense. Actually, the Torso does not have a describing motion relative to the lab coordinate system, I try to look at its coordinates in the coordinate panel. The same thing is happening to me with two different models. Look at the images below, this model is MoBL_ARMS_Upper_Limb_Model_OpenSim, using the original marker and the data provided with the model, I can do the scaling without any problem (the experimental data is near the model). Do you have any suggestion on how I can let the models to move freely in order to match my experimental data?
Re: Model does not move to match experimental data
Posted: Thu Jan 21, 2021 3:36 pm
by aafox
Hi Pablo,
I can't remember how the MOBL model is constructed with respect to the torso, but I suspect like most models the torso is "joined" to the ground body of the model (and then all other joints flow on from the torso). The type of joint used for this (e.g. a weld joint, meaning the torso cannot move with respect to the ground) may not be what you require - and hence need to be edited.
A lot of the gait models will give you an example of the joint that's required - in that in these cases the pelvis is allowed to move freely in 6 DoF with respect to the ground, which it sounds like what you might need your torso to do. Alternatively, you could check out a recent project I conducted (
https://simtk.org/projects/gh-caps-sims) which contains a torso and upper limb model. This model has the kinematic foundation of the MOBL arms model, but uses the musculoskeletal geometry of
Wu's shoulder model. In my analysis I locked the torso coordinates, but I think you could easily unlock these to achieve what you need.
Aaron
Re: Model does not move to match experimental data
Posted: Thu Jan 21, 2021 6:30 pm
by pablo_delgado
aafox wrote: ↑Thu Jan 21, 2021 3:36 pm
Hi Pablo,
I can't remember how the MOBL model is constructed with respect to the torso, but I suspect like most models the torso is "joined" to the ground body of the model (and then all other joints flow on from the torso). The type of joint used for this (e.g. a weld joint, meaning the torso cannot move with respect to the ground) may not be what you require - and hence need to be edited.
A lot of the gait models will give you an example of the joint that's required - in that in these cases the pelvis is allowed to move freely in 6 DoF with respect to the ground, which it sounds like what you might need your torso to do. Alternatively, you could check out a recent project I conducted (
https://simtk.org/projects/gh-caps-sims) which contains a torso and upper limb model. This model has the kinematic foundation of the MOBL arms model, but uses the musculoskeletal geometry of
Wu's shoulder model. In my analysis I locked the torso coordinates, but I think you could easily unlock these to achieve what you need.
Aaron
Dear Aaron,
Thank you very much, thanks to you now I know what is my problem. I will definitely look at your project, I just need to scale the model with my experimental data, then I will lock those joints in order to apply IK.
Regards,