MultiAgentDecisionProcess (MADP) is a toolbox for scientific research in decision-theoretic planning and learning in multiagent systems. It is designed to be rather general, but most effort has been put in planning algorithms for discrete Dec-POMDPs.

Comments, bug reports, patches, etc, are welcome. A mailinglist is available at [email protected].

Authors:

• Frans Oliehoek,
• Matthijs Spaan,
• Bas Terwijn,
• João Messias,
• Philipp Robbel

Please refer to the file AUTHORS for contact information and current affiliations. MADP includes other software, for details see COPYING.

MADP includes all required software, although for OSX argp-standalone needs to be installed.

Optional software:

• Doxygen (doxygen) [for generating documentation]
• Graphviz (graphviz) [for dependency graphs in the generated documentation]
• lp-solve (liblpsolve55-dev) [for POMDP vector pruning]
• GMPlib (libgmp3-dev) [for using arbitrary-length integers as indices]
• libxml2 (libxml2-dev) [for using the XML-based factored model parser]
• Cuda [for policy iteration with GPU policy evaluation]
• cplex [for the DP-LPC solver]

For enabling optional software, see src/Makefile.custom and src/include/configuration.h.

See doc/MADPToolbox-0.XXX.pdf, which also includes detailed installation instructions.

To (re)generate API documentation from source, run:

make htmldoc

Open doc/html/index.html in a webbrowser

To compile, execute the following

sh autogen.sh
./configure
make

Problem descriptions can be loaded without specifying a path if ~/.madp/problems is a symlink to the problems subdir in the MADP tree. Similarly, results are saved in (subdirs of) ~/.madp/results, so it will be convenient to make a symlink to the desired results locations. This can be accomplished as follows:

mkdir ~/.madp
cd ~/.madp
ln -s ~/<PATHTOMADP>/problems
ln -s ~/<PATHTOMADP>/results

Please see the AUTHORS file.