Tracking Dynamic Objects in a RoboCup Environment - The Attempto Tübingen Robot Soccer Team

This paper describes the Attempto Tübingen Robot Soccer Team which played at RoboCup 2003 in Padova. The robot platform, its sensors and actuators, and the software system running on the onboard computer are presented. The main part of the paper concentrates on our current scientific work on modelling and tracking a dynamic environment. Information about dynamic objects moving around in the environment can be useful especially in RoboCup to predict the motion of the ball, to avoid collisions, or to consider objects which can not be detected over a short period of time. In our robot soccer team we implemented an object tracking approach which on the one hand combines the specific advantages of Kalmanand particle filters and on the other hand uses an interacting multiple model filtering approach to model object dynamics as accurately as possible. In addition to the general tracking techniques we present a new real-time approach to detect and track uncoloured objects, such as a standard soccer ball.