gym-puddle

Puddle world environment for OpenAI Gym with continuous state space and discrete action space.

Installation

Run this line in the project directory.

pip3 install -e .

Usage

import gym
import gym_puddle # Don't forget this extra line!

env = gym.make('PuddleWorld-v0')

Notes

Rendering is available. The program runs much faster without it, but it's good for debugging.

Run this line after each time step to see the puddles and agent location.

env.render()

References

Off-Policy Actor-Critic. Thomas Degris, Martha White, Richard S. Sutton. In Proceedings of the Twenty-Ninth International Conference on Machine Learning (ICML), 2012.

http://rlpark.github.io/ (Java implementation)

https://github.com/samindaa/RLLib (C++ implementation)