[Coin-ipopt] interpretation of dtol

[Andreas Waechter]

Hi Ned,

> I am looking for a mathematical description of how the
> parameter dtol is interpreted by IPOPT. Any pointers will
> be helpful.

DTOL determines how well you want the optimality conditions to be
satisfied before IPOPT stops (with a "success" message).

If you look into my thesis at Eqn. (2.40), those are the optimality
conditions that IPOPT considers for the stopping criterion (for mu=0).
(You can find my thesis on my home page,
http://dynopt.cheme.cmu.edu/andreasw )

In case you use the default setting for ISCALERR (which is 1), those
equations are scaled; see the first equation on p.34 (starting with "max")
for details.  DTOL the specifies how small that expression on the left
hand side has to be (for mu = 0).

If you specify ISCALERR = 0, then those optimality conditions (2.40) are
taken as they stand (no scaling), and IPOPT stops if the max-norm of those
is less than DTOL.

I hope this answers your question.  If not, please try again :)

Regards,

Andreas