Closed fiorisi closed 6 years ago
In [5]: numpy.finfo(float).eps*4.0
Out[5]: 8.8817841970012523e-16
We are increasing this tolerance of eight orders of magnitude!
I must confess that I am bit afraid of the possible unforeseen consequences of this change. Do you think it is doable to just turn this in a settable parameters, and set it to 2e-7
only in the iCub generation pipeline?
Thanks @traversaro, I agree with you. Now there is the possibility to set the value of epsilon form the YAML file.
Great! It would be ideal if you could also update the documentation of the YAML parameters in the README.md file, but we can also merge in the current form!
Added also a description of the model base origin parameters.
With the new iCub2.5+ model sometimes the check on the alignment of a frame with the joint axis fails because the error is greater than "numpy.finfo(float).eps * 4.0" but smaller than 2e-7.