Daimler, KIT send autonomous vehicle on historic course

September 11, 2013 // By Christoph Hammerschmidt
125 years ago, Bertha Benz made the first cross-country excursion in history with an automobile - from Mannheim to Pforzheim in Germany, a distance of about 100 km. Now a current model Mercedes-Benz 500 drove the same route - autonomously, without human driver interaction. At the IAA automotive fair, Daimler introduced the vehicle to the public.

The equipment that enabled the luxury sedan to find its way automatically and to participate in today's road traffic without accidents has been jointly developed by Daimler and the Karlsruhe Institute of Technology (KIT). The sensors used to automate the vehicle - short- and midrange radar systems and stereo cameras - are utilized in a very similar form already in today's series models of Daimler's E and S class. The algorithms and computers that make the vehicle see and identify its environment has been developed by a team around professor Christoph Stiller from KIT. The development focused on ways to calculate how the car has to act in the moving traffic (behaviour generation) and then to identify the best trajectory. "Based on the sensor data, the system determines the steering angle, acceleration and deceleration", says Stiller. The algorithm steers the car through roundabouts and intersections, merges into the traffic, identifies other traffic participants and traffic lights, and drives it around obstacles. "With this research project we prove that highly automated driving is possible with simple video sensors."

Fig. 1: Sensors for the Daimler S 500 INTELLLIGENT. For full resolution click here .

An important topic for the researchers were the digital maps required for automated driving. These digital maps are more detailed than the maps in common navigation systems. They contain data the vehicle cannot acquire during its drive. Examples are the exact location of driving corridors in intersections, the position of traffic lights or right-of-way regulations. To make use of these data the vehicle must be able to determine its position much more exactly than the position provided by standard navigation systems. While normal GPS systems offer an accuracy of a few meters, the autonomous driving algorithm require an accuracy in the centimetre range. The KIT scientists resolved the challenge by means of a video-based localization approach: A specifically equipped vehicle conducts a charting ride; the visual data acquired