Uber’s self-driving computer system attempted to alert the driver of the vehicle to a pending collision seconds before the car struck and killed a 49-year old woman in Arizona in April, the National Transportation Safety Board said on Thursday.
In a preliminary report on the crash, the NTSB said the driver had the autonomous driving technology on for 19 minutes before the crash. The vehicle, a Volvo, was equipped with 10 cameras that Uber uses for its self-driving feature, as well as a “collision avoidance function” and other driver assistance features from Volvo.
The car initially classified the pedestrian, who was walking a bicycle across the street, as “an unknown object, as a vehicle, and then as a bicycle with varying expectations of future travel path,” the agency said. The self-driving system first registered the pedestrian 6 seconds before the impact, then alerted the driver 1.3 seconds before the crash that an emergency braking maneuver was necessary. The computer itself cannot initiate the maneuver, a feature that Uber says is intended to “reduce the potential for erratic vehicle behavior.”
The NTSB said the driver attempted to intervene less than a second before it struck the pedestrian, but that the operator did not begin braking until a second after the collision. The driver told the agency that while two phones were in the vehicle at the time of the crash, she was using neither and that she had been monitoring the self-driving system.
The NTSB said “all aspects of the crash remain under investigation” and intends to provide recommendations to prevent similar situations. Uber said it is conducting its own safety review of the self-driving vehicles program and that the company expects to share further information in the coming weeks.
Uber announced on Wednesday that it would shutter its self-driving tests in Arizona and focus on other markets, including Pittsburgh.
