RRA crash or runtime error

[Sylvia Wan]

Hi,

RRA initially crashed without any warning. After reading http://wiki.simtk.org/opensim/rraCmcPage and other forum post I have a few things that I am not sure about.

- I know how to ensure that a joint is locked in the osim file. However, how do I check it in RRA files so that it is consistent with the osim file?

-I chose body COM to be adjusted in CMC tool as my model is a lower half body model. I think it should be ok since pelvis has the largest mass. Do you agree?

-How do I uncheck “use_fast_optimization_target” for RRA. I can’t find this button.

Afterall, I only changed the COM of pelvis in the actuators xml file to be the same as the scaled model.

RRA ran up to 45%, and then a Runtime error by Microsoft Visual C++ Runtime Library came up and said ‘contact the application’s support team for more information’. After that, OpenSim shut down.

As a side, the only time that ID ran with my files, the output (don’t know if it is moment or force as it has no units.) are very high and may indicate the GRF may not be applied properly. I have checked and think that it may be that my force plate 1 was stepped by left foot but I applied the GRF on the calcn_r and vice versa for calcn_l. However, I don’t know if this is the only cause for the large force/moment. The curves look more like a contact force curve but the columns were labelled as moment. I am most confused.

When I tried to run ID again with the correct GRF on the corresponding calcn, ID crashed.
As RRA does not require data output by ID, does it imply that I can proceed to RRA and CMC even without ID results?

Please help on this urgent matter. It will be very much appreciated.

Sylvia

[Ayman Habib]

Hi Sylvia,

Some answers below:
1. RRA uses an existing model so if the joint is locked in the model it should stay locked while running RRA. The only problem that could happen in CMC (not RRA) if you specify a task to track a coordinate that is locked.
2. You can use the pelvis instead of the torso, you just treat that as if there was a torso that was welded to the pelvis (though there's some evidence that this may affect the interpretation of the muscle function). If you go this route you'd need to change the center of mass and inertia accordingly, since the only reason we use the torso is that it's poorly measured/scaled with markers and that it has teh largest mass/inertia.
3. use_fast_optimization_target maps to (the opposite of) the flag to "adjust kinematics" in the GUI.

4. There's an outstanding bug where ID would crash the first time you run it from the GUI but if you save the settings file, then load it and rerun it works. I'd also recommend that you keep your setup and data files for External loads in the same folder.

Hope this helps, and please let us know how it goes,
-Ayman

[Sylvia Wan]

Hi,

Thank you for your help. RRA is still not running.

I have ID output with the GRF applied to the correct calcaneous now. But the net joint moment is still very high, of magnitude up to 10000.

I have seen other user reporting around 6-100 for the pelvis residual actuator in RRA by checking pelvis values from ID. I have no idea why my ID outcome is so high and it resembles shape of force curve not moment curve.

Is it most likely that RRA is not running because GRF is not applied properly? How do I apply it properly then?

Now that I have increased the residuals actuator opt. force to around 10x original, i get a message saying the ID mass matrix is sigular.

Please help.

Sylvia

[Rudolf Klemetti]

Hi Sylvia -

You could try to see if COP data and foot marker data match. In other words, COP should lie 'within' foot markers. If COP is outside of a foot, it may cause problems.

How did you apply GRFs? If you used 'Add external load' option, make sure you have checked checkboxes 'Force is Global' and 'Point is Global'.

Hope this helps!

Rudi

[Sylvia Wan]

Hi Rudi,

Thanks for your comment.

I didn't tick 'point is global' before but i tried with it ticked just now and the error message said

'Exception:
InverseDynamics: ERROR- mass matrix is singular'

any idea?

also, how do you check if COP data are applied to foot markers? do you mean visualising the GRF using Preview Motion Data?

I tried that way, but I can't see any force in the view window for GRF.

Thanks once again.

Sylvia

[Ayman Habib]

Hi Sylvia,

The code checks that the mass matrix is singular because if it is then a force can produce an infinite acceleration. I'd ignore any problems downstream from that and try to figure out what's wrong with the model. ID is the easiest test, until you get that working it's pointless to try to do anything else (SO, RRA, ....).

You should be able to run ID without external forces and get residuals/forces at the pelvis (that resemble GRF). If you can't do that then you go back one more step and inspect your model and so on..

Hope this helps,
-Ayman

[Rudolf Klemetti]

Sylvia -

Just plot your IK/raw marker data for foot markers, e.g. z-coordinates of lateral and medial markers. Then plot COP z-coordinate to the same graph with markers. Depending on your marker set, COP should be between those markers or at least close to them. And of course, do similar plot for x-coordinates with appropriate markers. If COP is far from markers, maybe you should re-run IK or make sure COP calculattions are correct.

Unfortunately/fortunately, I haven't got that kind of error messages, so I don't know what they are all about.

Good luck!

Rudi

[Sylvia Wan]

Hi,

I see. I am just wondering for the walker_tibia and walker_femur associated with tibia and femur in Arnold's model 2010, they have mass = 0. Could that be the problem?

I have just tried ID without external force. Pelvis_list and tilt moment resembles grf curve, but with magnitude of 15000 the highest.

pelvis_ty is offset to 1000 with shape that looks like momeent curve, pelvis_tx has charecteristic shape for the first peak, then it revolutes around 0. pelvis_tz fluctuates around 0.

does that mean my residuals do not resemble GRF?

Many thanks.

Sylvia

[Sylvia Wan]

Hi Rudi,

Thank you for your useful comment.

My COP in x direction goes up to 300 but kinematics of markers range from -1 to 1.5 in x direction througout the motion.

Similarly, z COP is much bigger in magnitute than markers kinematics.

Maybe they are in different units?

Thanks for your help once again.

Sylvia

[Ayman Habib]

Hi Sylvia,

Since this's specific to the Arnold model, I'd suggest you check that with the model author. Having massless bodies is allowed only if they are welded to bodies with non-zero mass properties.

As long as you got some reasonable answers from ID (not NaNs), I'd rule out mass properties issues, however if the forces at the pelvis do not "resemble" your ground reaction forces (grf) it may mean that your grf are in the wrong frame or units. You can use the "Preview Motion Data" functionality to verify that for yourself. Please consult the user guide for more info on this functionality.

Hope this helps,
-Ayman