Failure is not the end; it’s how you learn to drive. In Driven to Evolve, your crashes fuel your progress. Harness the power of evolution to train a neural network of AI drivers, using a genetic algorithm to improve with every generation.

Watch as your cars attempt to navigate the track, learning from their mistakes. Each failure earns you money, which you can use to purchase upgrades like higher population size, increased mutation chance, and more. The more they crash, the better they become.

When all cars crash (or when you choose to restart), a new generation begins: stronger, smarter, and hopefully… less crashy.

Features

• Evolution-driven AI learning with a genetic algorithm
• Interactive upgrades that shape your AI’s growth
• Visual debugging with neural network ray casts
• A feedback loop where failure = progress
• Watch your cars improve over time — or guide their evolution yourself

Controls

Mouse: Navigate menus, purchase upgrades
Space: Pause / Unpause simulation
R: Force next generation
T: Toggle neural network inputs (ray casts) display

Made using Godot game engine and the NumDot library for neural network computations. The source code is available on this GitHub repository.