Internalsolve of upper_model causes unboundedness issues.

[1ozturkbe]

This is at least true in simple_wing(). This used to not be the case. For now, I am turning down the gamma until we can figure out the issues. (Granted, when the code was written, boundedness checks were not in force.)

[1ozturkbe]

Also, sometimes it triggers unbounded, sometimes it does not...Also, it has nothing to do with gamma either...

[1ozturkbe]

I'm currently hacking it with Bounded to see if it makes any difference whatsoever, or the warnings are fake.

[1ozturkbe]

I don't know if this is related, but the Best Pairs formulation with elliptical uncertainty gives a flat probability of failure curve...

[1ozturkbe]

Alright, I think I figured out the issue. The try/except here:

            try:
                sol_upper = RobustModel.internalsolve(model_upper, verbosity=0)
            except RuntimeWarning:
                upper_model_infeasible = 1

no longer works, because of the boundedness warnings. I believe the robust approximations are still fine, because the solution diffs are the same compared to a few months ago.

[1ozturkbe]

Oh, this is not great. Getting infeasible for sol_lower = RobustModel.internalsolve(model_lower, verbosity=0) Will have to dig up what exactly is going on.

[1ozturkbe]

Also, part of the issue is that cvxopt is failing for certain robustified SPs... This is what is holding up the latest PR, #45