Abstract
Maps are still used for two main purposes, localization and navigation. Localization helps to situate ourselves by comparing the landmarks seen around us with the landmarks given on the map. Navigation guides us from one location to another in an unfamiliar environment. The map format presented in this paper will cover both aspects. Because this map is primarily read by autonomous vehicles it needs to be digital. A graphical representation is not needed or the vehicle, however, to find errors quickly and to edit the map by humans a graphical representation together with a graphical interface for editing will be introduced.
Virtual maps of artificial road networks were created for Berlin test manoeuvres or scenarios in early stages of development. Though these maps were very small, they still allowed us to test multiple crossings, right of way behaviour, traffic lights and blocked roads. In preparation to drive on the German Urban area these maps included long straights, long curves and several adjacent lanes for driving up to 70 km/h, doing lane changes and overtaking slower vehicles.
Some requirements resulted not only from the technical needs but also from the perspective of software development. As any other map the digital map should represent the real world only as detailed as needed. If it is too detailed, the creation of these maps would consume more time than they should. Over 300 km of Berlin’s road network were mapped and autonomously driven. The experiments include runtime and file size comparisons are presented and discussed.
KEYWORDS: 3D Map, autonomous vehicle, localisation , navigation. Urban area