OpenAi's gym environment for Xmoto game - reinforcement learning research


Keywords
artificial-intelligence, deep-learning, deep-neural-networks, deeplearning, machine-learning, machinelearning, python, reinforcement-learning
Install
pip install gym-xmoto==0.1.2

Documentation

Xmoto AI

OpenAi Gym with Xmoto !

Actions

  • W = ADVANCE
  • A = LIFT FRONT WHEEL
  • S = BRAKE
  • D = LIFT BACK WHEEL
  • SPACE = change direction
  • NA = no action

Observations

  • Screen pixels (200, 150, 3)

Installation

pip install gym_xmoto
sudo apt-get install faketime
sudo apt-get install xmoto

Usage

Example

python example.py

For more details check this

Docker

(Currently doesn't work Seg fault at example in run)

docker build .
docker run -p 5900:5900 <image hash>
vncviewer localhost:5900

Second agent:

docker run -p 5901:5900 <image hash>
vncviewer localhost:5901

...

Roadmap

  • More observations from Xmoto API
  • Better score detection performance
  • Synchronisation algorithm - faketime
  • Distributed training
  • Reward on hit apple (decrease score)