Humanoid localisation in a robot soccer competition using a single camera

Abstract

One of the main challenges in a humanoid robot soccer competition is registering robots on the field so that they can estimate their best possible positions. This paper proposes an initial self-localisation algorithm to estimate the distance between a robot and a goal post, which is used as a landmark. By manipulating a single camera, we can apply analytic geometry to determine the real world coordinates of a robot from the transformation image plane. An experiment was conducted from the perspective of a robot in a soccer competition. The robot was able to locate itself with a minimum mean square error rate. This technology has great potential for boosting attacking and defensive performance.