Line Based Robot Localization under Natural Light Conditions

In this paper we present a framework for robot selflocalization in the robocup middle size league. This framework comprises an algorithm for robust selflocalization and a set of benchmarks which can be used offline to test other algorithms and to compare their outcomes with our results. The algorithm is part of our competition team BrainstormersTribots which won the Robocup German Open 2004. This is a multi agent real time environment, therefore our algorithm is prepared to work with 30 frames per second, leaving enough time for other tasks like robot control or path planning. Our approach uses a particle filter method relying on features found in the image. The features are points on field lines. They can be recognized reliably under natural light conditions, so the is no longer a need for a well defined and constant light source. Also color coded landmarks or goals are not required for a stable selflocalization. We present results for different runs on our benchmark suite, which is an outdoor soccer field with the size of 16x10m. This field size is bigger then in current competitions and anticipates the trend of using larger fields in future competitions.