[Ipopt] Runtime modification of the model

Horand Gassmann Horand.Gassmann at Dal.Ca
Fri Feb 25 10:59:06 EST 2011

Paul Smith <phhs80 at gmail.com> wrote:

> On Fri, Feb 25, 2011 at 12:11 AM, Paul Smith <phhs80 at gmail.com> wrote:
>> Is it possible to modify the optimization model while the C++ program
>> is running? For instance, is it possible to write a C++ program to
>> solve different optimization models inside a loop and with theses
>> models being not known until runtime? Basically, what I am wanting to
>> do is to change the coefficients of the initial model as new
>> information comes into the system.
> Should I interpret your silence as being impossible what I am wanting
> to do? Can a C++ program solve not more than a single optimization
> model?

Modification is always a tough thing. What do you want to modify? Just  
parameter values, or possibly problem dimensions? Even if it's just  
parameter values: You are using Ipopt, so presumably you have  
nonlinear problems. Do you know if these problems are convex? Can the  
convexity be destroyed by your modifications? How do you run Ipopt? If  
you call it from AMPL, you can set up the loop inside of AMPL, calling  
Ipopt from inside the loop.

If you give more information, perhaps someone can advise you better.


gus gassmann

More information about the Ipopt mailing list