https://github.com/vani-or/haxball-ai

Reinforcement Learning based bot for Haxball 1 vs 1 problem
https://github.com/vani-or/haxball-ai

Last synced: 3 months ago
JSON representation

Reinforcement Learning based bot for Haxball 1 vs 1 problem

• Host: GitHub
• URL: https://github.com/vani-or/haxball-ai
• Owner: vani-or
• Created: 2019-02-23T08:14:41.000Z (over 5 years ago)
• Default Branch: master
• Last Pushed: 2023-03-24T23:12:17.000Z (over 1 year ago)
• Last Synced: 2024-04-10T08:36:35.583Z (7 months ago)
• Language: Python
• Homepage:
• Size: 2.23 MB
• Stars: 28
• Watchers: 4
• Forks: 4
• Open Issues: 3
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

• awesome - vani-or/haxball-ai - Reinforcement Learning based bot for Haxball 1 vs 1 problem (Python)

README

        
# Haxball-AI

![Bot-vs-Human](docs/bot_vs_human.gif)

* [Talk on Machine Learning Milan event](https://youtu.be/Ma_MopOXLCg)

* [Full gameplay demostration video](https://youtu.be/fpIe6lNM1NE)

This is an implementation of a bot for [haxball](https://www.haxball.com/) problem using Reinforcement Learning.

*Attention! This repo does not provide any connector to the browser!*

### Requirements

python 3.6+

Cython

Tensorflow (1.13+)

### Installation

Installation of dependencies

    pip install -r requirements.txt

Simulator compilation with Cython

 

    cd simulator && python setup.py build_ext --inplace

    

### Training

[PPO2](https://openai.com/blog/openai-baselines-ppo/) algorithm from [OpenAI-baselines](https://github.com/openai/baselines/) was used for training.