[Ipopt] Restoration phase is called at point that is almost feasible, with constraint violation 3.108624e-14. Abort.

Paul van Hoven paul.van.hoven at googlemail.com
Mon Nov 14 07:40:41 EST 2011


I've implemented an algorithm which is a slightly modified version of
a already existing algorithm. But it is possible for me to run my own
implementation as if it was the original one. Therefore results of my
algorithm and another widly used standard version of it must be the
same. The version I've implemented uses ipopt for optimizing the
objective. Now, to check wether my implementation is correct or not I
ran some tests and compared it with the standard impementation.
Therefore I created some toy problems and it turns out that my
implemenation works fine and coincides with the standard version.

Now, my real problem can be calculated in the standard version as well
and it turns out to work fine, no problems with that. But running it
with ipopt it supprisingly fails with the error message:

Restoration phase is called at point that is almost feasible, with
constraint violation 3.108624e-14. Abort.

And the result of ipopt is:
EXIT: Restoration Failed!
Restoration phase failed, algorithm doesn't know how to proceed.

Actually since I tested my algorithm + ipopt with toy problems and it
worked correctly, I'm pretty shure that my analytical implementation
(such as objective,gradient,constraints,etc) are correct.

Has anybody an idea why ipopt aborts the optimization process?


More information about the Ipopt mailing list