How can AI help reduce traffic fatalities? Computer vision expert Torsten Schön (AImotion Bavaria Institute) outlines both challenges and opportunities for the technology in his insightful DLD Future Hub talk.
Currently, human error causes over 90% of fatal accidents in Germany, amounting to nearly 2,800 deaths annually, Schön points out. This number has remained stubbornly high, driving the ambition for “Vision Zero”, where road deaths are a memory of the past.
Autonomous cars are a promising way to achieve this goal, Schön explains – but the vision has turned out to be surprisingly complex. Ten years ago, “we were sure that we are close”, the former Audi executive recalls, but the challenge lies in making self-driving cars fail-safe.
“The problem that has been solved very quickly is that this works in 95 to 99% of the cases”, Schön says. What keeps vexing developers are the rare and complicated scenarios, “and they do happen all the time.”
To overcome these hurdles, Schön’s research focuses on cooperative perception, which relies on sensor data of many cars, not just a single one. “We utilize all of these sensors cooperatively to have a perception that uses all the information available,” he explains.
Watch the video to explore this topic further and learn about Ingolstadt’s unique lab for automotive safety research, CARISSMA.
