daleang / godotgymai

Communicating with godot engine through shared memory

GodotGymAI

Make your Godot project into OpenAI Gym environment to train RL models with PyTorch. This project only runs on Linux systems.

Requirements

1. Godot Engine version >= 3.2 compiled from source
2. Boost interprocess (apt install libboost-container-dev libboost-system-dev libboost-locale-dev)
3. Pytorch version >= 1.5
4. Python setuptools

Installation

First, in setup.py change the variable GODOT_PATH to the root directory of godot engine source. Then run:

python setup.py

This script does several things:

1. Copies GodotSharedMemory module and compiles standard godot editor (x11 platform).
2. Compiles x11 export template.
3. Installs python module GodotEnv that is used to communicate with the engine.

Examples

There are several example environments in the directory Environments:

InvPendulum shows the example of environment that relies heavily on the physics engine

LunarLander more complex example that shows how to randomize environment upon reset

Tutorial and API

Tutorial

API

TODO

1. Models deployment: add module to load your traced model into godot engine
2. Learning from pixels: passing godot viewport rendering as a torch tensor
3. Complete tutorial: tutorial about complete workflow, from empty godot project to training to deployment