JavaScript - Self Driving Car demonstrates the application of Neural Network, 500 cars were generated and the best car which performed no collision is set as best car and the sensors (rays) are set on the best car, the screenshots are present in Source Code. Basic physics are applied on the car like max speed, backward speed, left and right rotation angle, getting disabled on collision and so on. More traffic and features to be added, click on restart the car at starting position.

On the right hand side the visualizer is built based on math to demonstrate the Neural network and to demonstrate the move decision of the car. The bottom layer is the first input layer, whereas the topmost layer is the output layer; the forward, backward (or break), left and right arrows are animated when the car decides to move in that particular direction.

In training mode 100 cars will be generated and they will go in random directions as soon as you see a best car that has performed well then click on save which will let the Neural Network Know that what it needs to follow during next round. Click reload to begin the training and see the progress. Clickling on reset will discard all the saved progess and learning of the car and it will again begin from the start.

https://paragpalkar.github.io/Self_Driving_car/ Click on the link to view the project in Training mode

FREECODECAMP