Robust localization using a top-down approach with several LIDAR sensors

Abstract

Robotic devices gain more and more use for performing tasks that are more and more complex. In the case of a mobile robot, one of the pre-requisites for doing so is to have a way of knowing its state and a knowledge about the neighbouring environment. This is generally adressed by a localization strategy. This paper adresses a robust Top-Down real-time localization scheme based on a Bayes Network and using only LIDAR range finders in a urban-like environment. This localization scheme is able to focus on the most informative observable landmark, one at a time thus reducing computational burden. We focus mainly on the improvements made in precision from using one LIDAR towards using 5 LIDARs, thus showing the capability of our approach to cope with several sensors in the same scheme as for one. Experiments are achieved on real data to evaluate the precision of the localization and the computation time.