A vision-based method for localization of mobile robots

Abstract

The ability to determine its positions and orientations accurately is very important for a mobile robot when performing tasks such as free-range path tracking and navigation. In this paper, a cost-effective method is proposed for localization of a mobile robot, where a low-resolution charge-coupled device (CCD) camera, a stepping motor, and three artificial landmarks are used. The CCD camera is mounted on the robot and its rotation is controlled by the stepping motor. Then, from the images captured by the CCD camera, image-processing techniques are used to determine the distances between the CCD camera and the three landmarks. On the basis of these distances, it can be proved that the position of the robot is determined uniquely provided that the three artificial landmarks observed are not installed in a line. Moreover, the heading angle of the robot is determined from the rotation angle of the stepping motor. To account for possible installation errors, a non-linear programming method is proposed. Meanwhile, a least-squares method is also used to increase the accuracy of localization. To show the feasibility of the proposed method, experimental results are included for illustration.