Multi-Agent Reinforcement Learning Environment Framework

Train your own reinforcement learning agent to compete against others in multiplayer games. Designed for the UC Irvine reinforcement learning competition.

ColosseumRL contains a number of multiagent free-for-all games. Currently, we have Tron, Blokus, and 3 and 4-player tic-tac-toe. In the future, we will be adding Chinese checkers and other similar games. Tron is a fully-observable multiagent free-for-all turn-based snake variant where players try to survive the longest without crashing into walls or each other. Blokus is a fully-observable multiagent free-for-all turn-based game in which players place pieces on a board to claim space and strategically block opponents from placing their own pieces.

Multiagent free-for-all games are interesting from a research perspective because unlike in two player games, there is no clear optimal strategy. Instead, an gaent's success depends on the other players in the game. For example, agents that learn to implicitly team up with other agents might have an advantage over agents who don't.

Fall 2019 Competition

We are holding a competition to see how well people can perform in the task of Tron! In order to participate, simply train an agent to play the Tron environment on a 25x25 board with 4 players. We will be evaluating all of the agents using a round-robin tournament involving all submitted agents. We will also be releasing our own baseline agents periodically to act as benchmarks for your agents, but they will not be participating in the tournament. All agents should be finished by December 7th 2019.