A hybrid localization method for a soccer playing robot

Abstract

Self-localization is the process of estimating the robot position exploiting noisy measurements. Since localization is a key issue for soccer playing robots, some probabilistic approaches have been developed over last years to address it. Methods based on Monte Carlo Localization (MCL) show good ability in dealing with kidnap problem, however, most of them are unstable with limited number of samples. On the other hand, Kalman filter extensions are among the best light weight estimators for position tracking. Their drawback is that they are unimodal and can't be used for global and kidnaped problems. Combining the advantages of these two approaches can lead to a valuable method. In this paper we propose a new hybrid localization method that utilizes the MCL and UKF to reach a stable, multimodal, and low weight localization method. The advantages of our method are evaluated in several experiments.