/Self-Driving-Pi

A self driving car project using Reinforcement learning

Primary LanguageJupyter NotebookMIT LicenseMIT

Self Driving Pi

This is the repo for John Papetti III and Mark Zhang's self driving car project for the Wardlaw + Hartridge School 2019 Family Science Night (FSN)

An Intro to Self Driving Car

In this project, a self driving robot is created using a Raspberry Pi 3. The robot can successfully follow a road (black strip on the white surface) using machine learning models. The program process the pixel data from a webcam and control the robot's movement. A 80-to-2 dimension autoencoder was used to reduce the dimension space of the environment. DQN, a reinforcement learning model, was used to learn how to drive. The DQN network was trained in a computer simulation before being implemented on the actual robot.