Question about Jacobian in example se2_sam/se3_sam.cpp about manif HOT 4 CLOSED

Hello extr15 -- whatever your name is!

Yes you are right. What happens is that this factor is so trivial that it works anyway. But indeed one should be using the derivative of log() which is Jr^{-1}. I will double check it and change it if necessary when I have a minute.

As a side note, your suggested code however could make use of the Jacobian facility of manif, by doing

poses[0].log().rjacinv(Jac);
J.block<DoF, DoF>(row, col) = Jac;

Joan

from manif.

Hello. I had a look at this. In principle, you need Jr as indicated above. However, this Jacobian is very close to the Identity at the point of evaluation, which is the origin, so the math in the code is reasonably fine, and at the convergence point it is exact.

I am evaluating if the example code should be made clearer, and will fix this.

Thanks again

from manif.

Fixed. Waiting for PR acceptance.

from manif.

Thanks for your reply and work!

from manif.