simtk.org

Hi,
I just want to clarify a couple of things with tracking simulations for both torque and muscle driven simulations of walking.

1) The first question is can I track experimental ID joint moments for a torque-driven simulation? I have a 3D gait model with muscles on the lower extremity and coordinate actuators for the lumbar and arms. To modify the problem to be torque-driven I used ModOpRemoveMuscles() and then ModOpAddReserves() to add a torque to each coordinate that doesn't already have a coordinate actuator. Can I use the MocoControlTrackingGoal and reference experimental joint moments from ID to track ID moments and apply those to the reserve torques for each state? I know there is not a specific cost term to track ID moments, but could this be a way to do that with a torque-driven tracking simulation since the torques are the controls? Then should I be able to validate the output in the /forceset for the reserves be comparable against the reference ID moments?

2) The second question is what is the best way to compute or extract the joint moments from the states and grf tracked data from a muscle-driven simulation? Should I run that through ID using a model without the ground contact model? Or is there a way to do this without running ID?

Thank you for any help you can provide!

Kayla

Hi Kayla,

On point 1, I'm fairly sure I tried this (i.e. track torques from ID as 'controls') and it seemed to work OK. It sounds like a feasible and logical approach to me.

On point 2, could you get the moment arms of the muscles from an analysis and combine these with the forces to get the muscle moments and sum these at each joint coordinate? I recall taking a Moco solution back to inverse dynamics once and getting different results - specifically the Moco solution had very low residuals at the pelvis, but then the results out of ID suggested much higher values for these residuals, so I'm not sure it would be consistent.

Hope that helps.

Aaron

What Aaron described for (2) is how I've previously gotten joint moments from a Moco solution. If you have other actuators at the joint besides muscles, you'd want to add those too.

You can also get joint forces and moments from doing a Joint Reaction Analysis, but I'm not sure I'm doing that right or understanding correctly what the analysis is doing. For the constrained joint motions, the moments looked right to me (e.g. the frontal plane moment on the knee in my model with a 1-DoF knee looked like a typical KAM), but the moments on the actual DoF in the model didn't look right.

Ross

You can also get joint forces and moments from doing a Joint Reaction Analysis, but I'm not sure I'm doing that right or understanding correctly what the analysis is doing. For the constrained joint motions, the moments looked right to me (e.g. the frontal plane moment on the knee in my model with a 1-DoF knee looked like a typical KAM), but the moments on the actual DoF in the model didn't look right.

Hi,
That's because the coordinate systems used to express joints moments are different. Inverse Dynamics tool reports moments for coordinates (notice the knee joint moving axis in Rajagopal model), whereas Joint Reaction tool uses proximal/distal segment coordinate systems (parent/child). In Ref[1], the advantage of reporting joints moments in joints coordinate system is discussed, and in Ref[2], you can find why it's better to report KAM in segment coordinate system.

Ref:

[1] Schache et al. (2007). On the expression of joint moments during gait. Gait Posture. 25(3):440-452. doi:10.1016/j.gaitpost.2006.05.018

[2] Derrick et al. (2020). ISB recommendations on the reporting of intersegmental forces and moments during human motion analysis. J Biomech. 99:109533. doi:10.1016/j.jbiomech.2019.109533

Hope this helps,
-Mohammadreza