Sihai network

Uber crash caused by a software problem

Original title: new auto driving software or culprit found in Uber's car accident

As part of an initial internal investigation, Uber has concluded that its autopilot software detected pedestrian victims in a recent fatal car accident in Arizona, but did not respond immediately, the information said.

At the time of the accident, the Uber autopilot detects a 47 year old female pedestrian Elaine · (Elaine Herzberg), who was pushing her bicycle across the street at the time. ·. It is said that part of the reason Uber software didn't respond is that the technology has been adjusted to ignore objects deemed to be 'false positives', such as plastic bags on the road.

After the accident, Uber and the National Transportation Safety Commission of the United States launched an investigation to find out if there was a fault in Uber's automatic driving software. At present, these investigations are still in progress. However, people who were informed of the preliminary findings told the information that Uber's automatic driving software might be the cause of the accident.

Autonomous driving companies can adjust their technology so that they are not too cautious about obstacles on the road. Usually, when computer vision software detects an object, it becomes too cautious and then decelerates or stops. According to the information, Uber decided to adjust its system so that it would not be too cautious about potential 'false positives', however, this made the software fail to respond immediately to Herzberg, even though the system detected that she was on the road.

At present, Uber has stopped all automatic driving tests on public roads and hired the former chairman of the National Traffic Safety Council as the company's safety consultant.

Uber said in a statement on Monday: 'We conducted a top-down safety assessment of our driverless car project and hired Christopher Hart, former chairman of the National Traffic Safety Council, as our safety consultant. Our assessment involves the safety of the system itself and the training process of vehicle behavior. We'll release more details later. '