simtk.org

Hi everyone,

I'm working on a study about balance and I want to know if there is a way to predict the GRF.
In most cases people create contact models between the feet and the ground, but it seems to be complicated. I've seen some studies that try to predict the GRF during squats using simplified single leg models in OpenSim, but that's not an option for me. Others try to add weld constraints to the feet, but this doesn't give good results because OpenSim doesn't work well with closed-loop kinematic chains. I have read that MOCO can solve closed-loop kinematic chains, is this possible?

I've tried adding weld constraints between my model's feet and the ground and then using the "squat to stand" tutorial to solve a torque-driven predictive solution (adding the appropriate coordinate actuators and changing the start and end time position bounds to "ground_pelvis coordinates = 0"), but the problem don't converge and stop after 500 iterations even when I increase the solver tolerances from 1e-4 to 1e-2. I've done the same but with only one constrained foot and it works (the solution isn't pretty but it's a solution).

Am I missing something? If it's possible to solve this kind of problem, how can I get the forces on the feet constraints?

Thanks for your help
Adrian

Hi Adrian,

While foot-ground contact models present their own challenges, I think they're currently your best option for predicting GRFs with Moco. It's technically possible to use kinematic constraints between the feet and the ground to model contact, but even foer simple problems like standing, these constraints can be very tricky to deal with in practice. Before fully committing to using kinematic constraints, I'd recommend giving a foot-ground contact model (e.g., SmoothSphereHalfSpaceForce) a try first.

Best,
Nick

Hi Nick,

Thank you for the help, I will try to reconsider the problem as obviously using constraints does not seem to be a good way to proceed.

Best,
Adrian