Press release: When cars go to driving school

Nr. 182/2009 - 29.09.2009

Scientists develop a learning driver assistance system

“Foresight”, as one learns in driving school, is the key to driving safely. But this is particularly difficult when it is dark and visibility is poor. Within the framework of the EU-funded project “DRIVSCO”, scientists have developed a driver assistance system that can solve this problem. The system learns from the driver during the daytime and applies this knowledge at night, when its infrared system can see farther than the human eye. The EU project was coordinated by Florentin Woergoetter, a scientist at the Bernstein Center for Computational Neuroscience and the University Goettingen in Germany. Another eight partners from six different European countries were involved in the research project. The successful conclusion of the project manifests itself in a first prototype of the driver assistance system, which has been implemented in a test vehicle by the company Hella Hueck KG & Co (Lippstadt, Germany).

DRIVSCO is the first driver assistance system that learns from the driver. Using lane markers the system recognizes the course of the road. It saves images and road data and compares them to the driver’s reactions: How strongly does the driver brake if a turn of a certain angle lies ahead? How does he/she steer the car? This is how the system gets to know the individual driving style of a driver. At night, the system uses infrared headlights in order to detect the course of the road – it sees more than the driver and now knows from experience how the driver should react in certain situations. If the driver’s behavior deviates too strongly from the usual behavior, for example when he/she does not recognize a turn at night, the driver is alerted by the system. Besides infrared headlights, the driver assistance system contains a stereo camera that allows detection of other vehicles, calculating their distance away.

“The scientific challenge in the development of the system was the matching between image data and driver reactions”, explains Woergoetter. The driver may react quite differently to images that look very similar. The system has to learn which image aspects are crucial and which reactions they entail – it thereby also considers the driving style of different drivers. “Systems already exist that recognize if the vehicle changes its distance to the road boundary. But our system applies foresight and plans the driving behavior also for the further course of the road”, says Woergoetter.

DRIVSCO was funded by the EU with 2.8 million Euros over a time frame of 3.5 years. Further project partners are the Universities of Leuven (Belgium), Genoa (Italy), Granada (Spain), Muenster (Germany), Kaunas (Lithuania), and Odense (Denmark) as well as Hella Hueck Kg &Co (Germany).
Images are at: http://www.nld.ds.mpg.de/~tomas/drivsco/pictures/

Contact:
Prof. Dr. Florentin Wörgötter
Georg-August-Universität Göttingen
Bernstein Center for Computational Neuroscience
Department for Computational Neuroscience, III Physikalisches Institut - Biophysik
Tel: 0551 3910760, Email: worgott@bccn-goettingen.de