"It's a major breakthrough," said Davide Scaramuzza. The computer science researcher at the University of Zurich (UZH) developed the new system together with his colleague Daniel Gehrig. 

Today, driver assistance systems can already warn drivers of obstacles or initiate emergency braking. However, these systems are not yet fast and reliable enough. "They still need to be significantly improved if they are to be used in autonomous vehicles where there is no human behind the wheel," says Scaramuzza.

According to the researcher, such systems work with standard cameras that take 30 photos per second. In other words, an image every 33 milliseconds. In order to reliably detect a pedestrian or a car and initiate a braking manoeuvre, these cameras require at least three images, which brings the total time until the braking manoeuvre begins to 100 milliseconds. "Our system makes it possible to reduce this time to less than one millisecond," says Scaramuzza.

How the camera works

One way to improve this would be to use a high-speed camera that shoots significantly more images per second. However, these require enormous computing power.

The researchers therefore relied on so-called event cameras. Just like the human eye, these look for things that change quickly and only take a photo when something suddenly moves. However, they do not reliably register slow movements.

This is where Scaramuzza and Gehrig came in: They combined a camera that captures 20 images every second with an event camera and artificial intelligence (AI). The AI was trained to recognise certain obstacles, such as other cars or pedestrians, in the images from the first camera. Another AI system analyses the data from the event camera. The data from the event camera is used to predict what the conventional camera will see.

However, it could be many years before the new camera is actually used in road traffic.