Vision-based line tracking and navigation in structured environments

Abstract

This paper describes a vision-based, low-cost line-tracking system suitable for robot or AGV navigation in structured environments. Vehicle navigation takes advantage of the visual information provided by artificial or pre-existing landmarks, specifically lines and signs. This information is efficiently processed using specialized perceptual behaviors, including neural networks and focus of attention techniques, with the help of a multi-threaded real-time control architecture. Attained system performance is compatible with present requirements and practices for typical AGV applications.