Append/Prescribe vs. Tracking Ground Reaction Forces

[Mikayla Schneider]

Hello! I am working on a Moco tracking problem with experimental data of transtibial amputees walking with powered prostheses. We are trying to track the joint angles, GRFs, and prosthesis torque.

I was wondering if anyone had any thoughts on appending/prescribing vs. tracking the ground reaction forces. Since we are not quite sure if all the pieces (kinematics, GRFs, and prosthesis torque) are dynamically consistent, my thinking is that it is good to allow the GRFs to vary a bit in order to better match the kinematics and prosthesis torque (which generally does seem to be the case). However, the challenge is that the optimizer sometimes takes advantage of the GRFs. When I append/prescribe the GRFs, the overall motion is more smooth and consistent but I'm not sure if this is the best approach? Are there any major downsides/concerns I should be aware of in appending/prescribing the GRFs exactly?

Thanks in advance for any help!

[Nicholas Bianco]

Hi Mikayla,

Prescribing GRFs makes sense if your optimized motion will remain relatively close to the reference kinematics trajectory. If you're minimizing (or excluding) residual forces at the pelvis, the optimized trajectories will technically be dynamically consistent. But as you start to deviate farther away from the reference trajectory, prescribing the experimental GRFs will become less valid. If you're expecting large adaptations in response to your powered prostheses, then you might consider using a foot-ground contact model (e.g., SmoothSphereHalfSpaceForce) to predict GRFs.

Best,
Nick

[Mikayla Schneider]

That makes sense. Thank you, Nick!