Bozcomlekci / laser-hockey

A simple laser-hockey gym environment for RL agents

Team What the Puck?

An implementation of the laserhockey environment agents by:

DQN: Agniv Sharma

SAC: Batuhan Ozcomlekci

DDPG: Dhruv Behl

Install

Create a new environment_<method>.yaml file using anaconda like:

conda env create -f environment_sac.yml

Then execute the code below:

python3 -m pip install git+https://github.com/martius-lab/laser-hockey-env.git

or add the following line to your Pipfile

laserhockey = {editable = true, git = "https://[e-mail]/martius-lab/laser-hockey-env.git"}

Checkpoint files can be found in the drive https://drive.google.com/drive/folders/1zDMBWRs9vSFvKmyMMCFNzA-KhtNL4zVh?usp=drive_link

HockeyEnv

laserhockey.hockey_env.HockeyEnv

A two-player (one per team) hockey environment. For our Reinforcement Learning Lecture @ Uni-Tuebingen. See Hockey-Env.ipynb notebook on how to run the environment.

The environment can be generated directly as an object or via the gym registry:

env = gym.envs.make("Hockey-v0")

There is also a version against the basic opponent (with options)

env = gym.envs.make("Hockey-One-v0", mode=0, weak_opponent=True)

LaserHockeyEnv

A laser-hockey game implementation within openAI gym using the Box2D engine. It is quite a challenging environment See Laser-Hockey-Env.ipynb notebook on how to run the environment.