Uber may have switched off safety system before fatal crash, Bloomberg reports

Ride-hailing app Uber may have turned off a standard safety system in a Volvo SUV before the self-driving vehicle fatally struck a woman in Arizona last week, according to a report.

A spokesperson for Michigan based auto-parts maker Aptiv Plc, which supplies Volvo with the vehicle's radar and camera, told Bloomberg that Uber disabled the SUV's standard collision-avoidance technology.

"We don't want people to be confused or think it was a failure of the technology that we supply for Volvo, because that's not the case," said Zach Peterson of Aptiv, according to the news agency.

A dashcam video of the crash released by Arizona police showed that the self-driving Volvo that Uber was testing did not slow as Elaine Herzberg crossed the street while pushing her bicycle at night. 

The car was in self-driving mode with a human backup driver behind the wheel when it hit the 49-year-old woman.

She later died in hospital and was the first death involving a fully autonomous test vehicle in the U.S.. After the death, Uber halted testing of all its self-driving cars in all cities.

Uber did not respond when asked by CBC News if the company had disabled the car's standard collision-avoidance technology before the crash.

Regarding the next steps it plans to take, Uber said the company's autonomous cars remain grounded.

"We're heartbroken by what happened last week," the spokesperson said in a statement. "As we develop self-driving technology, safety is our primary concern every step of the way."

CBC News contacted the National Highway Traffic Safety Administration (NHTSA), which regulates vehicle performance standards in the U.S., to ask if Uber was legally required to leave the car's standard collision safety technology enabled during testing.

The agency did not answer about specific legal requirements, saying the case was an open and active investigation.

"Consistent with NHTSA's vigilant oversight and authority over the safety of all motor vehicles and equipment, including automated technologies, the agency has dispatched its Special Crash Investigation team to Tempe, Arizona," it said in a statement.  

"NHTSA is also in contact with Uber, Volvo, Federal, State and local authorities regarding the incident. The agency will review the information and proceed as warranted."

Banned in Arizona

The report comes as Arizona Gov. Doug Ducey suspended the tech giant's autonomous vehicle testing privileges in the state on Monday.

In a letter to Uber's chief executive Dara Khosrowshahi, Ducey said the video footage of the crash heightened concerns over whether Uber had the ability to safely test the technology in Arizona.

He said in a series of tweets that he was "personally very disturbed" by the video and it raised questions that need to be answered with the help of federal investigators.

"I want to see the facts," he said. "Arizona has allowed this technology to test and flourish. We will continue to encourage innovation. But public safety comes first. It's always been our focus and it will remain our focus."

Uber told CBC News in a statement that it would "continue to help investigators in any way we can, and we'll keep a dialogue open with the Governor's office to address any concerns they have."

'Experience' in detection

Meanwhile, some experts who have viewed the video of the crash say the Volvo's sensors should have seen the woman crossing the road and stopped before the crash.

In a blog post, Intel Mobileye's CEO Amnon Shashua wrote that it used the video of the crash to test its own software and was able to detect the woman one second before impact despite the poor quality of the video.

Intel Mobileye is a supplier to Aptiv, but it's not clear if the Mobileye technology was in the Uber car that fatally struck the woman.

"To demonstrate the power and sophistication of today's ADAS technology, we ran our software on a video feed coming from a TV monitor running the police video of the incident," he said. "Despite the suboptimal conditions, where much of the high dynamic range data that would be present in the actual scene was likely lost, clear detection was achieved approximately one second before impact."

"The software being used for this experiment is the same as included in today's ADAS-equipped vehicles, which have been proven over billions of miles in the hands of consumers," he added.