Vision-based driverless car in the condition of limited computing resource: Perspectives from a student competition

The "Digital race - Driverless car" was a student competition where each team was given a model car with fixed specification in both mechanical and electronics. The mission was to program the car to run automatically in the condition of road having obstacles, missing lane lines and bridges with completely out lane lines in the field of view. This paper aims to address the issues facing in the competition as well as a sustainable solution to make car run autonomously and rapidly in the condition of real time processing and limited computing resource. The vision-based solution consists of modules to infer the center of lane from images in normal flat road as well as road having obstacles and bridge. It also features a fuzzy controller and a PID controller to control the car accordingly. The proposed solution has been proven in practice by reaching the first position of the competition. © 2017 IEEE.