bstriner / gym-traffic

OpenAI Gym Environment for Traffic Control
77 stars 23 forks source link

gym-traffic

OpenAI Gym Environment for Traffic Control

Installation

git clone https://github.com/bstriner/gym-traffic.git
cd gym-traffic
python setup.py install

Environments

Simple Environment

Traffic-Simple-cli-v0 and Traffic-Simple-gui-v0 model a simple intersection with North-South, South-North, East-West, and West-East traffic.

CLI runs sumo and GUI runs sumo-gui. GUI is slower but required if you want to render video.

Agent has 4 available actions, corresponding to traffic light phases:

Simple Environment

Agents

A simple DQN agent is provided, written in Keras.

Questions?

Feel free to create issues, pull requests, or email me.