Egsolver solves finite, two-player energy games [0,1,2]. It computes the winning region and a winning strategy for the energy player using a simple progress measure refinement procedure (see e.g. [2]).

It is written in python and loosely motivated by the pgsolver suite for parity games.

Get the latest version:

git clone git@github.com:pazz/egsolver.git

In case you have pip installed, the following will do.

pip install ./egsolver

Otherwise, manually install the dependencies:

1. numpy
2. networkx

On debian/ubuntu:

sudo aptitude install python-networkx python-numpy

Now install egsolver using the setup script:

./setup.py install --user

This should install the executable as $HOME/.local/bin/egsolver.

There are currently three subcommands: convert, generate, and solve.

>egsolver -h

usage: egsolver [-h] [-v] [--version] [-l LOGFILE] {convert,generate,solve} ...

energy game solver

optional arguments:
  -h, --help            show this help message and exit
  -v, --verbose         increase verbosity
  --version             show program's version number and exit
  -l LOGFILE, --logfile LOGFILE
                        where to log to; defaults to '-' (stdout)

commands:
  {convert,generate,solve}
    convert             convert game description to another format
    generate            generate a random game
    solve               solve a game

Also check out egsolver solve -h etc..

To generate a random game (with 5 states, edge- and owner probability 1/2 and maximal absolute effect 10):

egsolver generate 5 0.5 0.5 10

To generate and solve a random game:

egsolver generate 5 0.5 0.5 10 | egsolver solve

To further format the result in dot-format and display with xdot:

egsolver generate 5 0.5 0.5 10 | egsolver solve -f dot | xdot -

egsolver uses a custom JSON format, designed with portability and later extensions to other types of games in mind. To see an example, just generate a random game graph.