[Coin-ipopt] Fixed variables in IPOPT

[Andreas Waechter]

Hi Jens,

> I have another question regarding IPOPT.
> How does it handle fixed variables, i.e. variables with equal upper and
> lower bound.
> We are currently frequently using this feature when performing what-if
> analysis.

In the general purpose code (not using DYNOPT), IPOPT simply takes the
fixed variables out of the problem that it solves, and keeps those
variables (when evaluating functions etc) at their fixed value.

In DYNOPT is it not possible to fix inidividual variables (corresponing to
discretized variables), since it destroys the structure that is assumed to
be identical in each finite element. When we had to fix such variables,
we specified very close (but different) bounds arount the set point.
That's not great, but usually worked.

Best regards,

Andreas