View Code? Open in Web Editor NEW

Bomberland: a multi-agent AI competition based on Bomberman. This repository contains both starter / hello world kits + the engine source code

License: MIT License

Dockerfile 0.26% Python 2.05% TypeScript 33.23% Go 1.40% Makefile 0.02% C++ 61.80% Rust 1.20% JavaScript 0.04%

bomberland's Introduction

Bomberland engine + starter kits

About

Bomberland is a multi-agent AI competition inspired by the classic console game Bomberman.

Teams build intelligent agents using strategies from tree search to deep reinforcement learning. The goal is to compete in a 2D grid world collecting power-ups and placing explosives to take your opponent down.

This repo contains starter kits for working with the game engine + the engine source!

Contributing

Contributions are always welcome, see our contribution guidelines here

Usage

Basic usage

See: Documentation

Clone or download this repo (including both base-compose.yml and docker-compose.yml files).
To connect agents and run a game instance, run from the root directory:

docker-compose up --abort-on-container-exit --force-recreate

While the engine is running, access the client by going to http://localhost:3000/ in your browser (may be different depending on your settings).
From the client, you can connect as a spectator or agent (to play as a human player)

Open AI gym wrapper (experimental)

docker-compose -f open-ai-gym-wrapper-compose.yml up --force-recreate --abort-on-container-exit

Starter kits

Kit	Link	Description	Up-to-date?	Contributed by
Python3	Link	Basic Python3 starter	✅	Coder One
Python3-fwd	Link	Includes example for using forward model simulator	✅	Coder One
Python3-gym-wrapper	Link	Open AI Gym wrapper	✅	Coder One
TypeScript	Link	Basic TypeScript starter	✅	Coder One
TypeScript-fwd	Link	Includes example for using forward model simulator	❌	Coder One
Go	Link	Basic Go starter	❌	dtitov
C++	Link	Basic C++ starter	✅	jfbogusz
Rust	Link	Basic Rust starter	❌	K-JBoon