Abstract:
The localization problem is the detection of the pose of a robot relative to theenvironment using the information about the environment sensed by the robot whenthe starting position is unknown. In short, it is answering the question "Where am I?". Localization is an active field of study where many approaches are introduced intothe literature.Robot soccer is a good platform to develop and test localization techniques sincethe robots have limited and noisy sensorial information as in the real life and theenvironment is also highly dynamic.In this work, a new module that will stand between the perception module andthe other modules that uses its output and a new localization technique are introduced,and they are together proposed as a new localization system.The proposed new module is My Environment, which stores the perceptional dataand provides a filtered and more robust data, and the new localization technique isS-Loc, which is a sample based localization technique where only one sample is usedfor each perception data.This system is implemented in Cerberus̕05, which won the Technical Challengesin the RoboCup 2005 - Sony Four-Legged League. This success, together with the experimental study, has shown that the proposed solution has a high performance forthe application domain.
