Page 1 of 1

Finding muscle activations reaching task

Posted: Tue Dec 08, 2020 4:40 pm
by lysea.m
Hello,

I would like to generate a reaching motion in an upper limb model and find the muscle activations that create this motion.

Can I use SCONE to maybe do this by defining an objective function of minimizing the error between two points (end effector & target) as well as activation levels and jerk?

I've tried to import the arm26.osim model and MoBL_ARMs model into the Scone GUI and have used the example script structure from Tutorial 1a with the same PieceWiseConstant controller but with BodyMeasure - defining the min & max & penalty (I'm unsure what these should be). However, nothing seems to happen when I try to optimise the scenario it either freezes or the program stops responding.

Just wondering if this this a suitable problem for SCONE or if there is some other functions I should be using? Or perhaps there is an example similar to this problem you could share. Thanks for your help! :D

Lysea

Re: Finding muscle activations reaching task

Posted: Wed Dec 09, 2020 8:41 am
by tgeijten
Hello Lysea,

You should definitely be able to do this in SCONE. The first thing to check is if your model simulates correctly. You can check this by pressing Ctrl + E (or via the menu Scenario -> Evaluate). If this takes a very long time, there may be an issue with your model.

Let me know if this helps!

Re: Finding muscle activations reaching task

Posted: Wed Dec 09, 2020 1:46 pm
by lysea.m
Hi Thomas,

Thanks for your reply. Yes it does evaluate, when I use the arm26 model it takes ~1second but does a strange backward movement and when I use the MoBL_ARMS model it takes ~1.5minutes it moves slightly forward.

The models looks fine in the GUI but I did have to change the geometry files. Would this have an effect on the optimiser?

Lysea
MoBL_ARMS.PNG
MoBL_ARMS.PNG (112.25 KiB) Viewed 209 times
arm_26.PNG
arm_26.PNG (97.82 KiB) Viewed 209 times

Re: Finding muscle activations reaching task

Posted: Tue Jan 26, 2021 9:16 am
by tgeijten
Sorry for the late response -- I missed your post.

I hope it's still relevant, but the geometry files have no effect on the performance. Seems like the arm26 model is suitable for forward dynamics simulation and should optimize well, while the MoBL_ARMS model has some issues.