Ai Car Project

Watch as a 20 Ai cars learn to drive around a track.

This is a neural network which controls each car, consisting of 20 nodes and 1 hidden layer. Raycasts are sent in 7 directions from the car, which provide as inputs into the network, along with the speed of the car.

Following principles from evolution each generation will be modified and competed.

Only the fittest will survive, and the slowest 10 cars will be eliminated and replicated as a modified version of one of the winning cars.

This was inspired by Sebastian Lague and 3Blue1Brown.

This project is no.3 of my Ai Car project. For the previous two projects failed at lack of neural network understanding and inability to correctly implement gradient descent from 3Blue1Brown. 

This is a simulation which can take up to 4 hours to complete.