Static Optimization -balancing actuator contributions

[Ryan Chhiba]

Hi,

I am trying to run Static Optimization with an upper extremity model to simulate a power grip and it continuously fails. I have tried adding in reserve actuators at the recommended levels (0.1N - 1N) but they continue to fail. If I increase them, the reserves contribute too much and the activations are minimal.
- Are there any recommendations on how to systematically determine the needed actuator contributions?
- Should I start at one joint and work from there, or should I find conditions that the simulation runs and then try and reduce them incrementally? For context: they are static trials, with external forces placed at the fingers, and the min/max controls are set to (1/-1)

Any recommendations would be appreciated. Thank you

[Carmichael Ong]

It looks like the maximum force/moment that your reserve actuators can produce might be very small. The total force/moment is determined by (control value) * (reserve actuator force value). So, these actuators are generating 1 N or 1 N-m (depending on if it's a linear coordinate or a rotational coordinate) at most.

You can increase the min/max controls much larger to first get the simulation to run, and then reduce it incrementally. Since the reserve actuator force/moment values are set to a low number (0.1 or 1), the optimizer will still tend to choose to use muscles rather than the reserve actuators.

[Ryan Chhiba]

Thank you,

To confirm as well. the contribution is determined by comparing the moments computed through inverse dynamics to the torques from the force.sto files outputted from the Static Optimization?

[Carmichael Ong]

Static optimization uses a least-squares approach because there are more actuators than degrees of freedom (see https://opensimconfluence.atlassian.net ... tion+Works). The equations there don't quite capture what the reserve actuators are contributing, but their control values are considered the same as the "activation" values for the muscles in this least-squares problem.