Majority Rule Based Sensor Fusion System with Correlation Evaluation for Mobile Robot Localization

Abstract

Robustness of localization is a crucial issue for a mobile robot moving in various kind of environment. We have researched a robust localization system with various characteristic sensors. In past research, we proposed a selective sensor fusion system with majority rule; representing each sensor's robot position as a probability distribution, evaluating similarities between probability distributions, and selecting probability distributions based on majority rule. Although the effectiveness of the selective sensor fusion system was experimentally confirmed, proper selection could not be done in some case because similarity evaluation had a problem. This paper proposes up-to-date system to conduct more proper sensor selection. Simulational and experimental results show that the proposed system can select proper sensors in various environment.