A low cost vision-guided car for autonomous racing car competitions

Abstract

This paper presents a brief “case study” report on the design and performance of a low-cost selfdriving, self-learning vision-guided robot car (VIC) that can drive in a left or right side lane and even identify and overtake obstacles or vehicles in front of it. In the future, it is hoped that more of these types of low-cost robots will be built and used for fully autonomous robotic racing car competitions. Topics that are covered herein include a brief description of CMU’s NAVlab self-driving car, which inspired this work, low-cost mechanical and electronic hardware, image analysis, object detection, ANN (Artificial Neural Network) training, the control software, test results and future work to improve robot performance and capabilities