simtk.org

Hi everybody,
I'm studying a side step cutting jump on one leg with a “Gait2392_Simbody” without the torso
and the left leg, and the mtp angle locked.
I saw that the Gait2392 model can be used also for jumps simulations or anyway for movements
with a higher dynamic than the normal gait, so I think it's right for me.
After the Inverse Kinematics and the Inverse Dynamics that worked well, I tried to run the
Static Optimization in order to calculate the muscle forces. The Static Optimization works
for almost the half of the time but then it always gives errors starting from the same frame
until the end:

____________________________________________________________________________________________
SimTK Exception thrown at InteriorPointOptimizer.cpp:261:
Optimizer failed: Ipopt: Maximum iterations exceeded (status -1)
OPTIMIZATION FAILED...

StaticOptimization.record: WARN- The optimizer could not find a solution at time = 0.353333

The model appears too weak for static optimization.
Try increasing the strength and/or range of the following force(s):
glut_med2_r approaching upper bound of 1
glut_med3_r approaching upper bound of 1
semimem_r approaching upper bound of 1
bifemlh_r approaching upper bound of 1
glut_max1_r approaching upper bound of 1
glut_max2_r approaching upper bound of 1
glut_max3_r approaching upper bound of 1
peri_r approaching upper bound of 1
vas_int_r approaching upper bound of 1
vas_lat_r approaching upper bound of 1
med_gas_r approaching upper bound of 1
lat_gas_r approaching upper bound of 1
soleus_r approaching upper bound of 1
tib_post_r approaching upper bound of 1
____________________________________________________________________________________________

I tried to do many different things:

1) I tried to lock the subtalar angle, a good approximation for the gait analysis, not for my
case, but I just wanted to see if it could work, but the problem was the same.

2) I tried to increase the maximum force of the residual actuators applied to the pelvis
(multiplying by 2, 5, 10, 100, 1000) but I couldn't solve the problem.

3) I tried to add other residual actuators to the joints (1 subtalar, 1 ankle, 1 knee, 3 hip)
with different magnitude but the errors are always occurring.

4) I tried to increase the maximum isometric force of all the muscles with a multiplier factor
of 2,5 and the errors appears some frames after, but they still appear.

5) I tried to filter the kinematics on my own without using the filtering mode of OpenSim, with
a Butterworth filter of the 3rd order at 6Hz, but I couldn't solve the problem.

6) I tried not to use the relation between muscles force-lenght-velocity but I couldn't solve
the problem.


I saw that the instant in which the error occurs in the original model corresponds to the instant
of maximum flexion angle of the knee and maximum flexion moment of the knee, and this value it's very high
(around 150 Nm).
I checked if in that instant there is any singularity in the kinematics or in the ground reactions
data, but it doesn't seem.

Does anyone can help me please?
Regards

Dennis Da Corte.

Hello everybody,
I solved the problem multiplying the maximum isometric force of all the muscles by a factor of 4 and increasing a little the maximum force of the residual actuators to the pelvis..the values of the maximum isometric force doesn't seem reasonable, but in this way the static optimization works.
Anyway the results seem reasonable, except the value of the maximum force of the soleus, that in particular for 2 tests reaches 6000 N.
Do you think that this value can be really reached?
Regards

Dennis.

Hi,

That is the issue with increasing the muscle force by some factor (arbitrary or incremental).

Have you examined whether allometric scaling of your model to fit your subject has affected the muscles normalized operating range? The approach we take is to optimize the fiber and tendon slack length to preserve the normal operating range for muscle. This will ensure that during your modelled tasks the muscle is operating in a range where it is at least capable of producing substantial forces (and hence moments) needed to sum to the external torques. Remember the cost function SO uses, so if your muscle is in a state where it can only produce large force with a massive activation then such a solution is unlikely given a minimization of activation objective.

Also worth looking at the passive forces. If you amp up the strength but do not consider other muscle properties (lengths, etc...) you may be generating non-physiological passive forces. So the solver converges but your results are not possible for humans. Not good.

More generally, have you considered whether static optimization is appropriate for a task such as side step cutting?

DJS

Hi Saxbyd,
thank you for the reply!
I tried to optimize the optimal fiber length in the hypothesis of rigid tendon, so making the Muscles Analysis I find the lenght m+t (muscle + tendon) frame by frame and from this I put away a constant contribute that is the tendon slack lenght (constant and taken from the original model), than I took the pennation angle at the optimal fiber length (considered constant) and I calculated the length of the muscles as:

lm=(lmt-lt)/cos(alpha)

frame by frame.
Than, since I wanted to be working around the optimal operating range for the muscles, I calculated the optimal fiber length as a mean of all the values lm calculated frame by frame.
I know that this process has a lot of approximations respect to the model used by OpenSim but it seems working well for a reasonable increase of the maximum isometric forces of the muscles (2,5 times the original ones).

To see the passive forces I launched the muscle analysis, but the values I obtain for the passive forces are constant..all the results of the muscle analysis seem incorrect out of the moment arms ant the length files. Am I wrong in launching it?

Thank you for your help.

Regards

Dennis.

Can anybody help me with calculation of optimal force. Is that a assumption. I can not figure it out.

Thank you in advance