Reinforcement-Learning-Curve-Fever
This repo contains a reinforcement learning implementation of the game Curve fever and is based on the work of A. Lilja and E. Bohnsack
It makes use of PLE and gym-ple to create a custom gym env for our Curve Fever.
OpenAI PLE (PyGame Learning Environment) is a learning environment, mimicking the Arcade Learning Environment interface, allowing a quick start to Reinforcement Learning in Python. The goal of PLE is allow practitioners to focus design of models and experiments instead of environment design. Finally, gym-ple allows to use PLE as a gym environment.
However, I had to modify some files of PLE and gym-ple to make the implementation work (e.g. fix bugs caused by pygamewrapper.py, alter registration files to include curve fever, etc.), therefore I included both modified repos in here.
To install PLE and gym-ple, cd into their respective folders once you have downloaded them and then use
pip install -e .
In addition, you will also need to add ffmpeg to your path variable for rendering, you can dowload for instance from here.
If you don't know how to add it to your path variable, you can follow this guide.
The curve fever environment itself is located in PyGame-Learning-Environment-master\ple\games.
Convserion to a gym env is done via the file gym-ple-master\gym_ple\ple_env.py
The main file is where all the magic happens. Here you specify the env that you want to use (e.g. Curve Fever, Flappybird, etc.), and specify the agent that you want to train, etc.. Check out link for a tutorial on using stable baseline agents (e.g. DQN).
TODO:
- implement other agents, see resources.
- take full frame as input via a CNN instead of beam search
- training against non-random opponents (e.g. version by Lilja & Bohnsack)
Some resources
Making a Custom gym env:
https://github.com/openai/gym/blob/master/docs/creating-environments.md
PLE:
https://pygame-learning-environment.readthedocs.io/en/latest/user/home.html
https://github.com/ntasfi/PyGame-Learning-Environment
ple-gym:
https://github.com/lusob/gym-ple
Agents:
https://github.com/openai/gym/blob/master/docs/agents.md
https://stable-baselines.readthedocs.io/en/master/
https://colab.research.google.com/github/araffin/rl-tutorial-jnrr19/blob/master/1_getting_started.ipynb