[Ipopt] Question about ipopt modification

[Andreas Waechter]

Hi Yudong,

The base class for the code that solves the KKT system (or augmented 
system) is AugSystemSolver.  The implementation of this in the Ipopt 
code is the StdAugSystemSolver that handles the matrix data format that 
we use in Ipopt.

If you really want to dig into the code, this would probably be the 
place to start (but don't expect this to be an easy task...).

I do not know of a Matlab implementation of the Ipopt code (there is a 
Matlab interface, though).

Regards,

Andreas Waechter

Associate Professor
Department of Industrial Engineering and Management Sciences
McCormick School of Engineering
Northwestern University
Evanston, IL 60208
USA


On 01/11/2012 06:13 PM, Ma Yudong wrote:
> Hi all
> I am currently trying to modify the ipopt codes to better explore the 
> sparse structure of hessian.
> For some specific optimization problems, the newton steps of the KKT 
> conditions (or cholesky factorization of the hessian)
> can be evaluated in a more efficient way. At the meantime, I want to 
> take advantage of all the heuristics such as filter step lengh control
> implemented by Ipopt for robustness and convergence property.
>
> It would be great if anyone can shed me some light on where to start 
> to replace the section of codes that calculates the KKT Newton steps.
> Another possibility is can someone share some version of simple ipopt 
> codes that are written in matlab that implements the filter-based
> step length control so that I can easily modify the codes.
>
> Any suggestions are highly appreciated.
> Thanks
> Yudong Ma
>
> -- 
> Thank you very much
> Yours
> Yudong Ma
> Ph.D. candidate
> Department of Mechanical Engineering
> 2169 Etcheverry Hall
> University of California
> Berkeley, CA 94720-1740
> email: myd07 at berkeley.edu <mailto:myd07 at berkeley.edu>
> Homepage: http://www.mpc.berkeley.edu/people/yudong-ma
>
>
> _______________________________________________
> Ipopt mailing list
> Ipopt at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/ipopt
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/ipopt/attachments/20120118/80ecc134/attachment.html>