RRA and CMC

[Friedl De Groote]

I'm trying to generate a muscle-actuated simulation from my own experimental data in OpenSim. I therefore follow the workflow described in "Generating a Muscle-Actuated Simulation in OpenSim". I'm able to scale my model, run the inverse kinematics tool to solve for joint angles and to run the inverse dynamics tool to solve for joint moments. Joint angles and joint moments look ok.

I'm unable to run the Computed Muscle Control tool. Using the Computed Muscle Control tool to reduce residuals results in an immediate clossure of OpenSim or in an animation of the model that is totally different from the inverse kinetics solution. When running the computed muscle control tool to solve for muscle excitations, I see the model animate (in a reasonable way) till the progress bar shows about 8% progress, thereafter OpenSim closes. I compared my input files to those provided in the examples but was unable to find what I'm doing differently. I get the same problem when using two different datasets.

Can somebody help me?
Thanks in advance,
Friedl

[Daniel Leib]

Hi Friedl,

I don't have a definitive answer, but I'm having the same sort of problem myself and think it is related to the integrator settings. Changing the settings will influence where the CMC calculations fail, but I haven't found any settings where it will complete.

Just to test out if it was me, I used example data provided and tried to run CMC for a full file rather than a small segment. Same results. I also know it's not a problem with system resources, as the simulations are being run on a brand new Dell Precision 690.

Any help from the OpenSIM team would be greatly appreciated!

[Ayman Habib]

Dear Daniel,

Can you please send me the details of how you reproduced the problem in OpenSim 1.0 with our own model and data files so that I can investigate?

Thanks,
-Ayman

[Daniel Leib]

Sure, no problem. After reading a post below, however, I wonder if it was because I was changing the settings through the GUI and not editing the XML files directly. I'll try that on Monday, but for now, here's what happened. I decided to post this here so others who were having the same type of problem could compare what happens.

I followed the tutorial for generating a muscle-actuated simulation using the folder Gait2354_Simbody up to the IK analysis. Since the tutorial only computes a small part of a stride and I wanted to see how a full stride looks, I widened the data range from time-points 0.4 to 1.6 to time-points 0 to 4. The IK ran with no problems. I then followed the tutorial up to the CMC analysis, but again widened the data range from the defaults to 0.0 to 4.0. It failed at 13% of the simulation a few times, so I started tinkering with the integrator settings. I widened the time steps, shortened them, changed both error tolerances both ways, etc. through all the options. The CMC failed every time, though consistently at a different point for each changed setting.

When it reaches that critical point in the simulation for whatever settings were used, memory usage (java.exe process, as tracked by Task Manager) goes up rapidly from about 350MB to 1.2GB, where OpenSIM closes without an error message. Nothing shows up in the Windows Event Log either.

I also tried disabling anti-virus software and the Windows firewall as well to no effect.

Just in case, here are the relevant system specs as well:

OS: Win XP x64
CPU: two quad-core Xeons
RAM: 8GB buffered
Video: NVidia Quadro 4500

This happens on a 32 bit lower spec machine consistently as well.

[Patrick Riley]

I am having similar issues. I have scaled a model, run the inverse kinematics and inverse dynamics tools without much problem. However, when I try to use the RRA optimization tool I run into problems. I too am using a longer data set, but haven't run into memory issues as I stop the optimization after relatively few frames. The ankle angle, and to a lesser extent the knee angle, very quickly diverge from the inverse kinematics. I was wondering about including marker tracking task in the RRA_Tasks. It now uses joint coordinates exclusively. Is that possible/appropriate?

[Daniel Leib]

Hi Patrick,

I've had this happen a couple times, but it was when the forces applied to the model were inappropriate for the kinematics. Have you checked to make sure the forces and motion data are synchronized/correct?

Have a good one,

Dan

[Patrick Brewer]

Hi All:

I see this is kind of an old thread. Has anyone been able to figure anything out on this issue? I too am having problems with OpenSim shutting down while trying to run RRA, and I can't seem to figure out why. OpenSim does create an error log when it closes, but I am unsure how to interpret it. The log is quite long though, so I'll wait to post it if someone would like to see it.

Thanks for any help you can offer.
Patrick

[Sam Hamner]

Hi All,

Usually when RRA or CMC fail it is because the model does not have enough "strength" to track the desired motion.

For RRA, you can try increasing the optimal force for all of the actuators, especially FX, FY, FZ, MX, MY, MZ. While this may increase your residuals, it will allow you to get a simulation which you can then begin to interpret and refine.

For CMC, you can try increasing the "reserve actuator" optimal forces in the actuators file. The reserve actuators are "ideal" torque actuators at each degree of freedom (imagine if you could have electric motors at each of your joints to help you move faster). Again, this may increase the amount of reserve actuators used in the simulation, but it will allow you to get a simulation which you can then begin to interpret and refine.

Let me know if this helps.

Cheers,
Sam

[Ayman Habib]

Patrick,

Please file a bug (using the Application's Help/Report-a-bug feature) and include the log file (likely hs_*.log) when you report it. This will help us figure out if that's an optimizer exception that was not handled properly or if it's something else.

Thanks,
-Ayman

[Patrick Brewer]

Thanks for the quick replies. Increasing the optimal forces seems to have done the trick; however, now I need to run a second RRA to try and minimize the residuals. Can you give me any pointers as to how to set the ranges for the residuals? What I've done is set the optimal forces equal to the average residuals from the first RRA pass and then set my default mins and maxs in the contraints file such that the ranges span the actuator force ranges as determined from the RRA1 actuator force output. Does this seem like a reasonable approach?

Thanks again.
Pat