Conversation
c2a5b1c to
e0cee6c
Compare
|
LGTM, thanks @alexbeattie42 👍 |
|
@nickbianco Do you think these minimal changes in inertia would break any tests/regressions or prevent reproduction of published work? |
|
@aymanhab I don't think we have any regression tests based directly on these models, and I doubt that these small changes would make much of difference in simulation. Users should be packaging models they used to create their simulations for reproducibility. I'm still confused on why this inertia error is appearing all of a sudden. I can't think of any changes in OpenSim or Simbody that would lead to this. That said, the changes seem reasonable enough. |
|
@nickbianco Using the latest version of the GUI build/artifact from here: I can open the Rajagopal2016 model without an issue on windows, and we don't support/publish a linux GUI. My guess is that these numerical differences are platform dependent, maybe due to lapack/blas differences so i see little harm in making the change but let me know otherwise. |
Resolves this issue
I created this python script which identifies in a
.osimmodel which inertial constraints will fail the triangle inequality and calculates the minimum change in one element required to fix them. I have tested with the models before and after and in my testing the models do not work prior to the repair and do work after.