Page 1 of 1

Excitation signal computation

Posted: Mon Oct 28, 2013 9:02 am
by alisy
Hi
My name is Ali, and I am a PhD student in robotics.

I have been reading the papers related to Thelen 2003 Muscle Model to understand better the algorithm of the activation and the musculotendon contraction dynamics. I have also been reading the papers related to CMC algorithm in order to understand the inverse problem of computing the excitation profiles. My problem is related to those topics, I'd like to compute the muscle excitation profile given a muscolotendon force (a force generated by a muscolo-tendon actuator). It's like a half way of CMC algorithm if I understood it well. How can I solve this inverse problem?

Thanks
Ali

Re: Excitation signal computation

Posted: Sun Nov 03, 2013 11:08 pm
by aseth
OpenSim does not currently have a tool for determining the muscle excitation directly from muscle force profiles. You are correct that this is part of the CMC algorithm, but it it computes the desired force using a variant of static optimization (to solve the muscle redundancy problem at a given instant in time) and does not expect them as given quantities. If you know the force that you want a muscle to produce and you want to solve for the excitation that would produce that force profile you can try to do something similar to the backsolve in CMC, where it solves for the excitations at the current time step to produce a target force at some dT (CMC time window) in the future from static optimization (with the current musculotendon length and velocity fixed). Alternatively, by evaluating the force-length and force-velocity effects from the current muscle kinematics (given the joint kinematics and target muscle force), you could solve for the activation that would produce the target muscle force and invert the activation dynamics to determine the input excitation (control). You should also note in CMC, that the muscle forces estimated by its static optimization are thrown away once the excitation (control) is computed. The control is applied to the forward dynamics model and muscle and multibody dynamics are integrated simultaneously and these produce the actual activation, muscle length and muscle force trajectories.