𐰆𐰺𐰴𐰣

Neuroevolution Flappy Bird

Evolution of Artificial Intelligence

Neuroevolution Flappy Bird merges the classic Flappy Bird gameplay with neuroevolution. Each bird is controlled by a compact neural network; the most successful individuals are selected each generation to reproduce, combining and mutating their networks to form the next generation. The project demonstrates how AI can autonomously learn better flight strategies over time without human input.

Core Features

• Parallel learning: Train with 50+ agents simultaneously (configurable population sizes).
• Real-time evolution: Observe population performance changes live, generation by generation.
• Speed control: Accelerate or slow down simulation (frame skipping) to either fast-forward training or inspect behaviors.
• Detailed metrics: Generation number, average score, best score, alive count, mutation rate and more.
• Visualizations: Neural map (network graph), generation charts, live gameplay overlays.
• Model management: Save/load the best neural network to/from localStorage or as downloadable files.
• Parameter control: Adjust mutation rate, population size, gravity, pipe gap, and pipe speed via UI.