Monocular visual self-localization for humanoid soccer robots

Abstract

In this paper a visual self-localization method for a humanoid robot is presented. This one is based on monocular information. The goal of this method is to obtain the position (x; y) and orientation θ of the humanoid robot inside the field of play. The methods proposed include some digital image processing algorithms and geometric interpretation to perform a 3D monocular reconstruction, that allows to measure the relative position between the robot and some known objects (goals and beacons), and to use them to obtain the robot's absolute position, by means of a triangulation method. Finally, the results obtained in Webots and in the real platform are presented.