Autonomous mobile robot localization based on monocular vision in RoboCup

Self-localization is an essential task for a robot operating in the RoboCup middle-size league competition. A monocular vision method is proposed to re-position the mobile robot without prior pose estimate. The static environment is described as the global geometrical map consisting of field lines, which are projected into the image. By integrating the line direction derived from the edge detector with the simplified randomized Hough transform, the proposed approach extracts the line segments to build the local map. Then via the map matching process, the mobile robot estimates its absolute position. Experimental results demonstrates the practicability. Performance of the approach suggested is also given.