Self-driving Ubers had many crashes before fatality

Development prototypes from Uber’s self-driving technology arm were involved in 37 crashes in the 18 months leading up to the company’s fatal Arizona crash last year.

The US National Transportation Safety Board (NTSB) confirmed yesterday that all 37 of the crashes involved Uber prototypes operating in their autonomous modes.

Frighteningly, 33 of the 37 autonomous Uber crashes involved other vehicles, though the autonomous Ubers were not always at fault.

The prototype involved in the crash that killed Elaine Herzberg was in its autonomous mode, while its “driver” was watching The Voice on a mobile device. It was the world’s first confirmed autonomous-vehicle fatal crash.

While police argued that the crash was “entirely avoidable”, prosecutors in Arizona said Uber wasn’t criminally liable.

While the NTSB will hold a probable cause hearing on the Arizona crash on November 19, Uber put its autonomous vehicle development program on hold in the aftermath of the fatal crash.

It resumed testing with upgraded software in a different part of the country (Pennsylvania) in December last year.

Uber has: "Adopted critical program improvements to further prioritize safety,” Uber spokeswoman Sarah Abboud said in a statement.

“We deeply value the thoroughness of the NTSB’s investigation into the crash and look forward to reviewing their recommendations."

One finding already announced by the NTSB is that Uber intentionally disabled the Volvo SUV’s built-in autonomous emergency braking system.

The NTSB said previously Uber had disabled an emergency braking system in the modified Volvo test vehicle.

While Volvo has not been involved in the NTSB discussions, Uber insisted to the organisation that its upgraded software would have detected Herzberg 4.5 seconds before the impact and braked for four full seconds before the collision.