I have a multiphase Dymos problem I am trying to optimise. It is not working so far and I think it is because of my sad looking Jacobian, as seen below. All driver values in the problem are scaled such that they are ~ within (0.1, 1), but that does result in a few elements in the Jacobian that have really small magnitudes.

I have tried to drastically alter the scalings of variables, but this has not resulted in success and the Jacobian I get looks equally bad.

So, if I undo the outragious scaling that resulted in the second jacobian above, I still have the problem of some really low ones. They are dealing with variables that I mostly have as states so they can be outputs, so I don’t think it’s a bad idea to just call their partial elements small enough to be 0 if they’re causing a problem otherwise. So, I would like to know how I could modify the threshold for how small these partials’ magnitudes have to be in the jacobian for them to just be considered 0.