Mobile robot self-localization using PDAB

Abstract

The aim of the paper is to make a contribution to the mobile robot self-localization problem, when the initial position is unknown. It is assumed that a simple map of the environment, consisting of a list of 2D-edge coordinates is available. These are regarded as natural or artificial landmarks. It is further assumed that the position of edges can be detected by a sensor in the local coordinate frame of the robot. From every pair of feature measures and map-landmarks, hypotheses of the robot-pose are constructed. These are regarded as measurement inputs into the probabilistic data association using Bayesian formulation (PDAB) approach. The presented first results demonstrates that the suggested method in combination with low level features is a promising approach for self-localization in unstructured and sparse modeled environments.