Confusing with Joint Reaction Analysis

[Zhen Yang]

Hi Matt, Ayman and everyone,

I try to use model Gait2392_Simbody to obtain hip joint reaction force (size and direction) using Joint Reaction Analysis.

Before I ran Joint Reaction, I have run through scaling, IK, ID, RRA and CMC.
When I ran Joint Reaction, I tried in 3 different ways, each of them generate different results. So I got really confused with Joint Reaction Analysis. Please help me to understand!

These are 3 ways I have done:
1) I used solution files from CMC, i.e. controls.xml and controls.sto in the Input "Controls" and "States", I checked the box "Solve for equilibrium for actuator states", then I appended gait2392_CMC_Actuators.xml to "Additional force set files", and added my external loads from lab to "External loads specification file", and I added JointReaction under Analyses tab.

Problem is I don't know what this analysis is doing.

2) I used solution files from CMC again, this time, I used states.sto file, which include both muscle activation and kinematics of the model, didn't check the box "Solve for equilibrium for actuator states", and I didn't append any actuators.
I used external loads and added JointReaction under Analyses tab.

Is it right to not append actuators? If not right, which file should I append?

When I validated my result here, the torques and force around pelvis are zeros, also torques around hip joint are zeros. So the model is fine, but reaction forces are a little bit too high as I expected.

3) I went through Matts webinar "Estimating joint loads in OpenSim", So I did first static optimization, using the model, solution from IK, external load, appended "gait2392_RRA_Actuators.xml".

After static optimization, I performed JointReaction, using the same model, IK solution, external load and actuators. I added also forces from Static Optimization to "force_file" under in JointReaction analyses.

Problem here is that when I tried to validate the result, torques around hip joint are not zero, but varies for example between -2.5Nm and 12Nm, I know that with a good result, the values should be zeros, but are they still good, even with these variations (forces and torques around pelvis are zeros. Reaction forces around hip joint are as I expected)?

Was it right to use "gait2392_RRA_Actuators.xml" as residual actuators?

Please help me, your answers are very important.

Zhen

[Zhen Yang]

Hi,

what I have understood that, I need a model, IK solution, external loads and residual actuators to run a static optimization analyze. But it seems that I don't need residual actuators because my trials are static (standing with different hip abduction angles). Because I have tried using with and without gait2392_RRA_Actuators.xml, and the results are identical. Am I right about no need of residual actuators or did I use the wrong file?

Thanks,
Zhen

[Zhen Yang]

Hi again,
in method 3) as I used static optimization in analyze tool, I got this error message

"time = 1 Performance =197.352 Constraint violation = 6.10963e-010" and the error messages appear every step until the end of simulation time.

I know this could depend on running simulation without a filter, but when I run again with filter on, the message remain.

I also tried to improve solution of my inverse dynamics, but the problem once again remains.

So I wonder if any one knows the reason to this problem and what is the solution of it.

Thanks,
Zhen

[John Zbyzec]

I hope you get an answer, because I am also wondering about what is preferable. Three methods with different results sounds worrying. What factors should we look at to know if a method is reliable? Which method do other prefer and why? How good indication of error is the nonzero torques around the pelvis?

[Ayman Habib]

Hi Zhen, John and all,

JointReaction analysis results will definitely depend on the forces applied. Muscle forces that result from SO are different from those resulting from CMC so that's not surprising, you get different results because you're solving a different problem. The flag to compute equilibrium for muscles means that the muscles are allowed to settle after their states are initialized from inputs so the resulting muscle states and forces are expected to change as well (vs. the flag unchecked).

Hope this explains,
-Ayman

[Zhen Yang]

Hi Ayman,

thanks for your reply. You mentioned that SO and CMC are solving different problems when obtaining muscle forces. I wonder which problems these two methods are solving respectively, in order to be able to apply one of them to my trials. My master thesis is about calculating hip contact force vectors for standing in different hip abduction angles, barefoot.

Regarding flag to compute equilibrium for muscles, I wonder when would I need to check/uncheck the flag?

Regards,
Zhen

[Ayman Habib]

Hi Zhen,

StaticOptimization and CMC both solve the distribution problem to compute muscle forces that produce specific joint torques. The main difference is that SO is solved frame by frame making some assumptions about the muscle (documentation here http://simtk-confluence.stanford.edu:80 ... tion+Works) while CMC solves same distribution problem while integrating to guarantee continuity and account for muscle states that can't change abruptly (documented here http://simtk-confluence.stanford.edu:80 ... +CMC+Works).

If you have full states (including muscle states) you should turn off the flag to compute equilibrium as your states will not be used as is, while if you don't have the full state then you'd turn the flag on.

Hope this helps,
-Ayman

[Zhen Yang]

Hi Ayman!

Thanks for all for inputs so far. I think I understand better now.
But I have another question about Analyze Tool. Under "Actuators and External loads" pane, I wonder if I should append actuators "Gait2392_CMC_Actuators" to model's force set or should I replace model's force set with the actuator file that I added. What I have understood, if I replace model's force set with the file, then all forces in the model will be removed and only those in the additional force set files will be applied.

And I need muscle forces from CMC in order to compute joint loads, and I checked "append to model's force set" in CMC Tool, so my guess is that I should append additional force set files here too. Is this right?

Zhen

[Ayman Habib]

Hi Zhen,

According to the documentation of the Joint Reaction analysis
http://simtk-confluence.stanford.edu:80 ... s+Analysis

<forces_file>: If a file name is provided, the applied forces for all actuators will be constructed from the forces_file instead of from the states. This option should be used to calculate joint loads from static optimization results. Please check the page above for notes related to how various actuator types are treated.

Hope this helps,
-Ayman

[April McPherson]

Hi Zhen,

I am also currently attempting to make sense of the results I am getting with joint reaction analysis. However, I am slightly confused reading your post here. Could you possibly explain why you feel getting torque values of zero at the hip joint is considered a valid model? I may be wrong in my thinking, but isn't there torque present at the hip joint?

Thanks for the clarification.
-April