GodotAIGym
GodotAIGym allows you to turn your Godot project into an OpenAI Gym environment for training reinforcement learning models with PyTorch. It supports Godot 3.2 and currently runs only on Linux systems.
https://github.com/lupoglaz/GodotAIGymPreview Image
GodotAIGym: Reinforcement Learning with Godot and PyTorch
GodotAIGym provides a bridge between the Godot game engine and the OpenAI Gym environment, enabling developers to train reinforcement learning (RL) models using PyTorch. This integration allows for the creation of intelligent agents within Godot games through RL techniques.
Key Features
- OpenAI Gym Integration: Turns your Godot project into a compatible environment for training RL models.
- PyTorch Support: Facilitates the use of PyTorch for developing and training RL agents.
- Linux Compatibility: Currently designed for Linux systems.
- GodotSharedMemory Module: Uses shared memory for efficient communication between Godot and Python.
- Example Tutorial: Includes an InvPendulum tutorial demonstrating environment creation, training, and deployment.
Requirements
- Godot Engine v3.2 (compiled from source)
- Boost libraries (interprocess and time)
- PyTorch v1.10
- Python setuptools
Installation
Installation involves setting the GODOT_PATH variable in setup.py to your Godot Engine source directory and running python setup.py. The script handles downloading necessary dependencies like LibTorch, compiling the GodotSharedMemory module, compiling the X11 export template, and installing the GodotEnv Python module.