Chrome Dino Game

Neural networks evolve to jump cacti and duck under pterodactyls in the classic Chrome offline runner.

Generation0

Alive0 / 0

Best Fitness0

About This Simulation

This simulation demonstrates the NEAT algorithm applied to the Chrome offline dinosaur runner. An entire population of dinosaurs plays the same obstacle course at once, and each neural network decides when its dinosaur should jump or duck. Press the spacebar to start or stop training at any time.

Technical Details

Neural Network Inputs: Dino Y position, vertical velocity, current game speed, and the next obstacle's distance, width, height and Y position
Neural Network Outputs: Jump and duck decisions
Fitness Function: Distance run before crashing into an obstacle
Population: 150 genomes per generation sharing a single horizon of cacti and pterodactyls