SAR data based Landmark Navigation for highly precise Vehicle Localization

Abstract

This approach deals with a probabilistic state estimation method with the aid of landmarks in the context of autonomous driving and driver assistance systems. It is mandatory to get a precise ego-position in order to operate vehicles throughout complex urban environments where global positioning methods like GPS provide only low position accuracies and low reliability. On-board sensors like LiDAR are able to detect a wide range of objects in the surrounding of a vehicle. Static urban objects like streetlamps or traffic light posts with absolute coordinates could be a solution to improve stability on the localization of a vehicle. Modern remote sensing approaches like geodetic radar are able to extract certain features like pole-shaped objects in an extensive and fast way. This work demonstrates the possibility of locating oneself within a landmark map based on a low-cost LiDAR sensor. Monte-Carlo localization enhances GPS positioning accuracy and was successfully applied on a test site.