Suche
Mein Konto
Robo-cars learn from near misses
ALP.Lab, the Austrian test region for automated driving, is installing camera systems at intersections for the first time that can detect and analyze near-accidents.
Robo-cars learn from near misses
The object-based traffic monitoring systems are currently being installed by ALP.Lab technicians in Graz and on other rural and urban intersections in Austria. “We mount the sensors of modern cars on street poles to detect accidents that almost happened,” explains ALP.Lab managing director Gerhard Greiner. The intersection area in particular is an enormous challenge for the development of autonomous vehicles. Cars, trucks and pedestrians, e-scooters, e-bikes and skateboards, as well as soon the first generation of semi-autonomous vehicles, meet here. The ALP.Lab traffic observation systems use radar, lidar and optical cameras to anonymously record road users and divide them into categories. The data can then also be used by municipalities and traffic planners to increase road safety and develop efficient forms of public transport.
Intelligent driving assistants
The data collected is primarily used as training data for autonomous driving systems. In contrast to human driving students, computer systems have to store and process huge amounts of data in order to derive meaningful actions from them. ALP.Lab can now offer this data to automotive suppliers and vehicle manufacturers as well as scientific research projects. “The traffic observation data is an ideal complement to the real tests of automated driving functions offered by ALP.Lab,” explains Gerhard Greiner, Managing Director at ALP.Lab. In order to avoid accidents as best as possible, data not only from accidents that have occurred but also from near-accidents must be evaluated. These so-called critical traffic scenarios include, for example, pedestrians who have difficulty crossing the street before the end of the green phase or vehicles that suddenly change their turning radius in order to avoid crossing cyclists. The sensors used are in use 24 hours a day, 365 days a year and can therefore not only record a large number of typical and dangerous scenarios, but also correlate them with a wide variety of general conditions - such as traffic volume, weather, time or temperature. ALP.Lab Managing Director Jost Bernasch: “Such traffic monitoring is new international territory and is already arousing great interest among experts.”