RL-Gridworld is a simple project to implement grid world environment, like: maze. You can also implement your agent under /agent.
PR and fork are welcome :)
In this project, I use CPU-version pytorch to implement my agent.
you can run the following method to create virtual environment for this project.
virtualenv -p python3.5 venv
source venv/bin/activate
pip install -r requirements.txt
- agent package: We can implement the AI agent here.
- env package: Defined environment base class, grid etc.
- maze package: I put a example maze here, which is inherit from env.env.GridWorld
- models directory: store pre-trained model here.
- a2c_client.py: A script to train A2C Agent.
- deep_q_client.py: A script to train Deep Q Agent.
- play.py: A example to demonstrate the interaction between SimpleMaze and AI agent.
Just run the play.py
(venv)$ python play.py
The MIT License