An OpenAI Gym environment with complex puzzle maps, highly adequate for reinforcement learning approaches, based on Geometry Friends. You can find a more recent version of GeoFriends2-v2.
To ready the environment, we recommend using VirtualEnv. You will need the PyGame and Gym environments.
python3 -m venv venv
source venv/bin/activate
pip install --upgrade pip
pip install pygame gym
python TestEnv.py
We also included some deep reinforcement learning examples, using Asynchronous Advantage Actor-Critic and Asynchronous 1-step and n-step Q-Learning. For those, you will need TensorFlow 1.1+, and SciPy. You can run both algorithms locally with
export PYTHONPATH=$(pwd)
python A3C/A3C-LocalThreads.py
python DQN/DQN-LocalThreads.py
Or distributed with the scripts (for example, 12 processes):
./start-a3c.sh 12
./start-dqn.sh 12
We have published a paper in IJCNN18 with more details and our results, titled "Guided Deep Reinforcement Learning in the GeoFriends2 Environment".