[Ipopt] wrong "infeasibility" message

[Juan Carlos Lopez Alfonso]

Hi Jonathan:

Thank you for your answer, because I need to finish my model in order
to publish a paper and  now I dont know the way to continue. I think
that the best way to help me, is not sending you my code, because is
the implementation of a variational problem that uses the finite
element discretization and even my reduced problem (reduced, because I
have assumed a cubic domain with a regular mesh)  is really big. In my
real problem, I have more that 2 millions of variables, where I have
discretized the head of a patient in order to simulate a radiotherapy
treatment. For this reason, I can not follow the procedure that I have
mentioned in my last email and I need other suggestions or comments to
solve this problem with volumetric constraints.

Could you have ideas about this kind of situations? where start with a
feasible initial point is really complicated. I am sure that the
implementation of my constraint and its gradient is correct (when
change epsilon), so the problem is the Ipopt algorithm and the way to
compute the step for select new feasible solutions in each iteration.

Do you give me ideas?
Best reagards and thank you
Juan Carlos

On Fri, Jun 15, 2012 at 12:12 PM, Juan Carlos Lopez Alfonso
<jc.atlantis at gmail.com> wrote:
> Hi all,
>
> I have written two days ago to this list and I have not received any
> answer until this moment, sorry but I need suggestions and comments. I
> need some help with my problem, because I have included a volumetric
> constraint, which guarantee the existence and uniqueness of an optimal
> solution, but in some cases the optimization algorithm ends with this
> message: "EXIT: Converged to a point of local infeasibility. Problem
> may be infeasible". I have made some test, and when I change the
> initial point in the same problem, in some case the optimal solution
> is reached and in others not. For example, I have computed the optimal
> solution of my problem without this volumetric constraint, and after,
> I have used this optimal solution as the initial point of the same
> problem when I impose the volumetric constraint not so restrictive.
> Therefore, I have used always the optimal solution of the last problem
> to change the volumetric constraint more and more restrictive. With
> this procedure, is possible to obtain the optimal solution of my
> problem with a volumetric constraint more and more restrictive.
>
> I cant use this procedure in my real problem, because is a big one, so
> I can not compute a lot of optimal solutions in order to obtain the
> optimal solution of my problem with my volumetric constraint.
>
> Could any one give me suggestions or comments? Is there any way to
> change the way to compute the step in each iteration in IPOPT? Please,
> any comment will be really useful for my.
>
> Thank you in advance and waiting for comments
> Juan Carlos