A Modified Sampling Method for Localization Accuracy Improvement of Monte Carlo Localization

Abstract

Unmanned vehicles are devices that can move around and perform tasks without an operator onboard. Such features are essential in many applications. Localization is a very important task in any autonomous mobile robot; in order to reliably navigate, the robot must keep accurate track of where it is. In the past few years Monte Carlo Localization (MCL) has been one of the most successful and popular approaches to solve the localization problem. MCL is a Bayesian algorithm based on particle filters. This paper is an attempt to increase the accuracy of localizing a mobile robot by modifying the way of generating samples from the proposal distribution of the MCL algorithm. Results show improvements in localization accuracy as compared to the basic MCL algorithm.