A fast and low-cost vision-based line tracking measurement system for robotic vehicles

Abstract

Localization and tracking systems are nowadays a quite common solution for automated guided vehicles (AGV) in industrial environments. The bunch of technological solutions developed in this sector is now being revitalized by their applications in service robots, e.g. for safe navigation in crowded public spaces or in autonomous cars. Related to this field are robotic competitions and races. In this particular scenario, the robot often has to track a line painted on the ground. Line tracking techniques typically rely on light dependent resistors (LDR), photo-diodes or photo-transistors detecting the light generated by normal or infrared Light Emitting Diodes (LEDs), arrays of electric inductance sensors or vision systems. Crucial issues for line tracking are: accuracy and reliability in estimating the direction of the moving vehicle with respect to the line and processing speed. Such problems are particularly critical when high-speed mobile robots are considered. To address this issue, in this paper a vision-based technique of moderate computational complexity is described. The proposed solution has been implemented on a low-cost embedded platform and relies on a high frame rate light contrast sensor, a tailored RANSAC-based algorithm and a Kalman filter. The reported experimental results prove that the proposed solution is able to track the direction of the vehicle in real-time even when the field of view of the camera is limited and the vehicle moves at high speed.