Could Moco estimate muscle activity accurately without measured GRF?

[zhibo zhang]

Hi Moco Team:

I'm just wandering, could Moco estimate muscle activity accurately without measured GRF? We are trying to build musculoskeletal model for small animals which is very hard to measure the GRF directly. It would be ideal if Moco could estimate the muscle activity without prescribed GRF.

Best
Zhibo

[Nicholas Bianco]

Hi Zhibo,

If you want to solve a problem in Moco without experimental ground reaction forces, you can use a foot-ground contact model like SmoothSphereHalfSpaceForce. Note that when using a foot-ground contact model that it is important to create a good initial guess for your problem, otherwise it will most likely have trouble converging. As a starting point, I recommend applying strong torque actuators to all model degrees of freedom and put the model into an initial state where the contact spheres are not penetrating deeply into the ground (or even at all).

Best,
Nick

[zhibo zhang]

So basically, Moco cannot estimate GRF directly. It needs a GRF model, and some guessed initial value first, and the iterating through an optimization process to estimate the GRF, right? Does the estimated GRF in Figure 8 of Moco paper (Dembia, C. L., et al., 2020) also come this way?

[Nicholas Bianco]

Hi Zhibo,

Yes, that is correct.

-Nick