An implementation of rtNEAT by Kenneth O. Stanley for neuroevolution in a real-time, persistent environment within the Godot game engine. Building on concepts from adaptations like "SharpNEAT" and "NEAT for Godot", this project shifts to a pooled evaluation format. This approach keeps genomes continuously active in the environment, with real-time swapping of underperforming genomes for ongoing optimization.
This project is written in GDScript for Godot 4.2.1
This project is tied to the Bachelors Thesis "Real-time genetic algorithm library for the Godot game engine" (TalTech, 2024)
See the wiki for general information: wiki
- About the engine see the official site Godot or the wiki overview
- How to use this for a project: Usage
- What are the demos: Demos
- Saving and loading genomes
- Saving and loading scenes
- Binary with saving and loading
Special thanks to Paul Straberger for the first NEAT for Godot implementation. This project reuses the config loading and low-level datastructures from NEAT for Godot