colosseumrl 1.0.0

UC Irvine multi-agent reinforcement learning framework

Codebase for the UCI multi agent reinforcement learning framework

A framework for developing large multiplayer reinforcement learning agents that can play free-for-all games. Currently available games are:

• Tron: A simple snake-like game where the goal is not to crash into the walls.
• Blokus: A board game of controlling territory with various shaped blocks.
• Tic Tac Toe: Generalization of tic tac toe to 2, 3, and 4 players.

Requirements

This library requires at least Python 3.5 in order to run correctly. Python 2.7 is not currently supported.

Basic requirements are listed in the requirements.txt and can be installed from PyPi with pip install -r requirements.txt.

Install

Simply clone the repo and run pip install -e . in the root directory to install a development copy of the library. Full pip install is not supported yet.

Important scripts

python3 -m colosseumrl.matchmaking.MatchmakingServer launches the main matchmaking server for allowing any number of agents to play against each other in a dynamic way. Run python -m rlcompetition.matchmaking.MatchmakingServer -h for more information.

./colosseumrl/examples contains a list of example scripts that will connect to a matchmaking server and launch an example agent.