Woman dragged along

Cruise: tests with robo-taxis halted after accident

In response to an accident involving a pedestrian in San Francisco, the robo-taxi company Cruise has now suspended all its journeys on public roads. The company, which belongs to car giant General Motors, has announced that this is to regain trust while investigations are carried out. Previously, only the operation of cars without people at the wheel had been stopped.

In the accident at the beginning of October, a Cruise driverless car dragged a woman several meters. The pedestrian was hit by another vehicle with a person at the wheel and thrown in front of the self-driving car. According to the accident report, the robotaxi braked immediately - but the woman was still dragged under the vehicle.

Woman dragged six meters
The Cruise cars are programmed to automatically drive to the side of the road in some cases after collisions so as not to obstruct traffic. In this case, the software also decided to do this - even though the woman was still under the car. She was dragged around six meters and the car reached a speed of around eleven kilometers per hour, according to a report by the Californian traffic authority. Cruise has since changed the software for its approximately 950 vehicles to prevent the situation from happening again. Around 70 cars were also affected by the halt to test drives with people at the wheel, it was now reported.

Test area for autonomous vehicles
San Francisco became a unique test case for self-driving cabs last year. In addition to Cruise, Google's sister company Waymo also received permission from a Californian supervisory authority in the summer to expand driverless transportation services throughout the city. The city council and many residents were against it. They argued, among other things, that the vehicles often blocked traffic.

Waymo is allowed to continue offering its driverless robotaxi service throughout San Francisco. Overall, autonomous vehicles are taking much longer to become part of everyday life than predicted just a few years ago. While the technology has long worked under simple conditions, some insiders are now questioning whether the software can be trained for all unexpected situations.