by Enrico Bertolazzi
This is my rewrite of Peter Carbonetto MATLAB interface for IPOPT a software package for large-scale nonlinear optimization.
Source code and documentation for IPOPT can be downloaded from:
On OSX IPOPT can be installed using Homebrew (
Why do a job already done?
The original MATLAB interface (
https://projects.coin-or.org/Ipopt/wiki/MatlabInterface) seems not maintained for OSX and do not
work with recent MATLAB distributions.
I have reorganized and simplified the original interface
(only internally) and eliminated the bug (due to MATLAB change
in the managing of sparse pattern) in the interfacing
of sparse pattern between MATLAB and IPOPT.
Moreover I tried to improve error catching.
How to compile
If you have an OSX or a LINUX 64 bit OS try to use the corresponding precompiled mex file, otherwise try to compile the mex interface following the instructions. Compilation is not an easy task, I try to be as clear as possibile, in any case things may change from different version of OS.
The interface is written in C++. To compile the mex file
- change the working directory to the
- run the script
To compile the mex interface you need a static IPOPT library
installed in your system. To install the library read
On linux system make sure that atlas versioni of blas/lapack libraries are available. If it is missing use the command
sudo apt-get install libatlas-base-dev
to install it.
For the usage of the mex file use
addpath to add
in the search path or move the contents of
lib to a directory
in the search path.
In the directory
example you find the original examples
of Peter Carbonetto which shows the usage of the interface.
Enrico Bertolazzi Department of Industrial Engineering University of Trento email@example.com