NeuroCars is a personal project that demonstrates
neuroevolution
using
planck.js for physics simulations.
In the simulation, self-driving cars contain brains (neural networks) that control their steering while they try to
navigate a track without touching the walls. Input into the brains:
distances to the nearest track wall in 5 directions. Output: whether to turn right, left, or do nothing. After each
iteration, a fitness score is assigned to each bot proportional to
how long they stayed alive and how far they traveled, higher fitness scores make a bot more likely to reproduce.