[Coin-ipopt] IPOPT exit codes

[Andreas Waechter]

Hi Zhiwen

I guess there is no detailed documentation on all exit codes, except what
you see in README.IPOPT and the return message (at least I don't remember
having written one :).

Here a bit of information:

Code 17 (EXIT: Convergence to stationary point for infeasibility) means
that Ipopt converged to a point that is infeasible, but where the
constraint violation (measured in the 1-norm) is minimized.  This could
mean that your model is incorrect (in the sense that you pose a constraint
that cannot be satisfied), i.e. it is infeasible.  It could also mean that
the iterates got into an area of the domain where locally the constraints
cannot be satisfied.  In the first case you should revisit your model, and
in the second case a different starting point might help...

Code 16 and 18:  This could mean that the problem is really nasty, and
Ipopt fails.  This can for example happen, if the gradients of the
constraints are linearly dependent, in which case the constraint
qualification does not hold.  Since the mathematical algorithm behind
Ipopt assumes that such a constraint qualification holds, it fails.  In
that case it might help to revisit the problem formulation and see if such
linear dependency can be removed...

Maybe that helps a bit...

Andreas

On Thu, 30 Jun 2005, Zhiwen Chong wrote:

> Hello Andreas and everybody else,
>
> Where might I find documentation on all IPOPT's exit codes and their
> corresponding explanations?
>
> IPOPT is usually very good at returning a one-liner describing the
> nature of the error, but for some
> situations (e.g. exit code 17), it doesn't really tell you what's
> going on.
>
> Thanks a lot for your help!
>
> Zhiwen
> McMaster Advanced Control Consortium
> --
> "If the women don't find you handsome, they should at least find you
> handy." - Red Green
>
>
>
> _______________________________________________
> Coin-ipopt mailing list
> Coin-ipopt at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/coin-ipopt
>