Autonomous Driving Model: Lane Detection and Visual Odometry for a Robotic Autonomous Vehicle

Abstract

The goal of this thesis is to demonstrate that an autonomous vehicle can effectively navigate a roadway using only a single low cost RGB camera. More specifically, the project studies how to develop a lane detection and visual odometry system for autonomous driving. Using a scaled down robotic model of a car to test the algorithms, the thesis first models a lane detection algorithm that allows the robotic car to follow roadways using similar methods that a full scale car would use to track lane markings on a paved roadway. Building off of the lane detection algorithm, this project develops a monocular ORB visual odometry system for the robot car which allows it to build a map of the roadway on which it is traveling.