A Study on Markov Localization for Mobile Robots

Abstract

The Markov localization algorithm is a means of estimating position of a mobile robot using a probability density over the environment of the robot's moving. By means of sensory data and motion model, it can be used to estimate robot's position under global uncertainty. However,some problems are found in our study. For example, the probability density cannot be recovered when it decreases to zero. A robot with only distance sensors cannot find its position in a symmetrical environment by means of Markov localization algorithm alone. In order to solve these problems a modified Markov localization algorithm is presented, and an approach in which a robot is equipped with a compass or gyroscope, has been proposed. An angle Guassian distribution defined in this paper is used to construct a new perceptual model for the robot and the new localization technique based on these ideas is thoroughly presented. A simulation program is used to demonstrate the effectiveness of the new technique for a robot moving in a symmetrical environment.