Godot Gym API is an Open-Source framework to enable use of Godot 3.5 as 3D- or 2D-environment to train Reinforcement Learning (RL) algorithms.

Documentation available here.

1. Install the framework in python with the following command:

   pip3 install git+https://github.com/ChernyakKonstantin/godot_gym_api.git#subdirectory=python
   

2. Install Godot addon into your Godot project with godot_gym_api CLI tool:

   godot_gym_api install_addon <PATH_TO_YOUR_GODOT_PROJECT>
   

   and activate installed addons in Project/Project Settings/Plugins tab.

3. Install protobuf into your OS following Protocol Buffers repository.

To run an example launch main scene of Godot project examples/apple_seeker/godot from Godot engine, then launch python training code with python examples/apple_seeker/python/train.py command.

There is persistent TCP communication between a Godot application and Python implemented inside a gym.Env inhereted class. Thus, feel free to use any of your favorite RL-frameworks!

Currently, there is support of Python as the programming language to implement the RL-algorithms. In Python, the environmnet provides gym / gymnasium interface.

If one has a desire to implement the client for any other languages, submit pull request, please.

• Vectorized Environment support;
• Multi-agent Environment support;

Join this Telegram channel to watch for new updates.

You can contact me through E-mail or Telegram.