You are currently browsing the category archive for the ‘Software’ category.

Cabal cannot install the GLFW package version 0.4.2 for GHC 7.0.3, at least on Mac OS X 10.6. 64-bit. This seems to be a well-know bug but I couldn’t find a solution online. To fix the problem, you can install the GLFW library directly, then do

sudo cabal install glfw -fdynamic

The “dynamic” flag avoids that cabal tries to recompile its version of GLFW.


Bonmin is a solver for mixed-integer nonlinear programs. It is a global solver if the continuous relaxation is a convex program, and my hope is that it will perform better than the naive branch-and-bound I recently had to implement for a project (written in Python with cvxopt for quick development reasons, its performance was about the same as Yalmip bnb).

I struggled quite a bit yesterday to compile the library on my Mac running Snow Leopard though. Here is the final configuration I used, and so far that seems to be working (at least make test succeeded):

./configure FFLAGS="-arch x86_64" ADD_CXXFLAGS="-mmacosx-version-min=10.4" \
ADD_CFLAGS="-mmacosx-version-min=10.4" ADD_FFLAGS="-mmacosx-version-min=10.4" \
LDFLAGS="-flat_namespace" --with-blas='-framework vecLib' \
--with-lapack='-framework vecLib'

While I’m at it, if you want to compile Ipopt, the coin-or interior point optimizer on which Bonmin relies, then the following simpler configuration should do it:

./configure FFLAGS="-arch x86_64" --with-blas='-framework vecLib' \
--with-lapack='-framework vecLib'

The dominant software for control system design currently is clearly MATLAB. It has a nice Control System Toolbox, a Model Predictive Control Toolbox, a Robust Control Toolbox, and various other related toolboxes, such as Optimization and Signal Processing. Simulink is very useful for system design, and can be coupled to Stateflow to analyse hybrid systems.

Nevertheless, the main drawback of Matlab is that is it not free, and there has has been a desire to develop more widely accessible alternatives. The first that come to mind are Scilab and Octave. However, despite their qualities these packages have suffered from an insufficient number of users for their usage to become more widespread. For example, Scilab was created in 2003 by INRIA (the French national institute for research in computer science and control) but the overwhelming majority of the Scilab consortium still consists of French research institutes and companies. I believe that one important reason why the situation is unlikely to change is that MATLAB is universally used in the classroom to teach control systems design, in part because it is expected that students entering the industry know this language. Still, this is not a very good explanation: Scilab and Octave are purposely using a syntax very similar to MATLAB.

Anyway, another issue with MATLAB is that it is yet another language to learn, and new features in the language tend to be introduced slowly. For example, object-oriented design has been added only recently. Instead, there is a strong trend currently in using Python for scientific computing, and several scientific communities are enthusiastically developing in Python. Some examples of impressive packages include Numpy and Scipy, Matplotlib for visualization, and Sage, which is a wrapper for a large number of mathematical libraries. One can find an increasing number of libraries for signal processing, machine learning, computer vision, optimization, and so on. With regard to control systems libraries however, there seems to be much less activity. Here are some related links that I know of:

Please let me know of any other Python library related to control systems.