Laser proximity scanner correlation based method for cooperative localization and map building

Abstract

This article is concerned with an autonomous robot self-localization, that is the first step of robot navigation. Only if the robot knows where it is relative to a suitable coordinate frame, can it plan its trajectory, and plan the whole mission. The next important part of the autonomous mobile robotic system is a map of the environment. Many self-localization modules suppose the map of the neighbourhood is known. In this paper a method for cooperative robot map building and robot pose-estimation is presented. One SICK Laser proximity scanner is used as the only input sensor for the method. The method is based on angle, x and y histograms and cross-correlation function. As a localization quality estimation, robot evidence grids and modified evidence grids methods are used. The output from the method is a two-dimensional map of the robot's environment and the actual position of the robot in this map. The method is programmed in GNU C programming language and tested on UTAR (Universal Teleprezence and Autonomous Robot) system. Results of practical experiments in a real environment are presented.