Probabilistic localization methods of a mobile robot using ultrasonic perception system

Abstract

Effective localization is a fundamental prerequisite for achieving autonomous mobile robot. In this paper, we propose three probabilistic approaches to solve the global localization problem and the kidnapped robot problem. The first approach named the hybrid Grid-MCL algorithm merges Monte Carlo Localization (MCL) and grid localization. It can solve the global localization problem with very low on-line computational costs. The second approach, sampling in Similar Energy Regions (SER), is used to conquer the kidnapped robot problem. The third approach is a combination of previous two approaches with adaptive samples, which solves the global localization problem and the kidnapped robot problem together. The validity of our approaches is verified through extensive simulations employing ultrasonic perception system.