Calibration Error: calcGoal

[Petko Petkov]

Hi I am a newbie to the OpenSim community and I would appreciate your help.
I am facing consistent trouble with calibrating captured data with both the Rajagopal_2015.osim and upper limb model.
The tutorial example works great but any sort of attempt of input of gathered 1,2,3 upper limb sensor or data from the UC Irvine Opportunity data set yields no skeleton (just gray window in Opensim) and the following error:

Exception in opensense: AssemblySolver::assemble() Failed: SimTK Exception thrown at Assembler.cpp:227:
Error detected by Simbody method AssemblerSystem::objectiveFunc(): calcGoal() method of assembly condition OrientationSensors returned a negative or non-finite value -nan(ind).
(Required condition 'goalValue >= 0' was not met.)

I read in the forum that this could be associated with the lack of markers in the model file which I have confirmed to be true for both the upper limb and Rajagopal_2015.osim models.
1) I have seen this error popup in people using motion capture systems which I am not. What would be the problem and proposed solution for the error?
2)Why does the tutorial work yet it has "no markers"?
3)In what cases are markers a must have other than in the motion capture scenario?

I am using a windows cmd prompt to run this.
Thank you!

[Ayman Habib]

Hello Petko,

Please file a bug (using Help ->file a bug) and include the model, processed data files with quaternions and the xml files you used to reproduce the problem. Answers to your questions below:
1. Likely issue is either that your provided orientations are inconsistent so that can't all be met or are not transformed correctly so that the model can't match the orientations. Use simplest model possible to troubleshoot.
2 & 3. We can do pure imu tracking without markers, the problem formulation is very similar. Eventually we could allow mix-and-match of markers and orientations but as of now we allow one or the other. You see discussion regarding mocap only because it has been there for a while.

Hope this helps,
-Ayman

[Petko Petkov]

Thank you !
I decided to pursue the orientation issue track further. I am using non APDM sensors and initially relied on manual conversion to .sto. Now I converted the original raw data into APDM format and used it through the standard OpenSim pipeline and I obtain the exact same results as previously (calcGoal failure).

The data is coming from 3 sensors attached to an arm at the shoulder,elbow and wrist positions. I am thinking that the position is just fundamentally incompatible with the model selected. And to get to see if the pipeline works we must place the IMUs at the torso similarly to the tutorial example. Is there any limitation or violation of assumption to using these sensors in the position described? Many thanks.

Attached is the model and .sto file used for reference.