31 Mar First time: Object-based traffic monitoring in Graz
For the first time, object-based traffic monitoring is being used in Graz, operated by ALP.Lab. We collect data anonymously e.g. to increase road safety and as training data for AI to develop automated driving systems.
“We mount the sensors of modern cars on street poles in order to see accidents that almost happened,” explains Gerhard Greiner, Managing Director of ALP.Lab.
The classic method of collecting training data for Artificial Intelligence (AI), in order to develop autonomous driving functions, consists of driving test vehicles equipped with sensors many thousands of kilometers on public roads. With the current project, ALP.Lab is taking an innovative, more sustainable path to gather additional data – saving many test kilometers (= CO2, noise, etc.).
Radar-, lidar- and optical sensors, similar to the once that are already installed in modern vehicles, are being used in a permanent infrastructure based installation. We mount them along crossings to provide data 24/7. With the help of suitable algorithms, we calculate the position of each object from this bird’s eye view and can thus represent situations from the perspective of pedestrians, cars or cyclists.
The sensors can thus not only capture a large number of typical and dangerous scenarios, but also link them to a wide variety of conditions – such as traffic volume, weather, time or temperature.
“Such traffic monitoring is a very new approach, ”explains ALP.Lab managing director Dr. Jost Bernasch. “We are pleased that ALP.Labs strategic decision, to break new ground in challenging times, is bearing fruit and we are looking forward to detailed results.”
The main aim is to collect data about critical scenarios, like near misses: In order to increase the active safety of vehicles, i.e. to avoid accidents as best as possible, it is essential not only to have data on real accidents, but also to have highly accurate real traffic data on accidents that almost happend. The collected data will be analyzed and summarized based on object lists.
Right now we run sensor systems on different rural and urban crossings in Austria, e.g. in Graz, Feldbach, Straden etc.