The death of Elaine Herzberg in March 2018 was caused by a self-driving vehicle system that could not see "jaywalkers" and failed to classify Herzberg as a pedestrian.
Herzberg, 49, was killed when the Volvo SUV struck her as she ignored the crosswalk signs and walked her bike onto the road. The bike had no side reflectors and the front and rear reflectors faced away from the self-driving Volvo.
Additionally, Herzberg's toxicology tests came back positive for marijuana and methamphetamine.
In a final report from the National Safety Transportation Board (NTSB), conclusions reached by investigators match nightmare scenarios predicted by dozens of safety advocates who claim consumers are being treated as lab rats when self-driving cars are on the roads.
In the NTSB report, the 2017 Volvo XC90 that hit Herzberg had its Volvo collision avoidance system deactivated at the time of the crash.
The SUV was factory-equipped with several driver assistance systems offered by Volvo, but those features were disabled because Uber had equipped the Volvo with its "Advanced Technologies Group (ATG) developmental automated driving system (ADS)."
When the XC90 SUV was operated in manual mode and controlled by a human driver, all the Volvo driver assistance components were activated, but those Volvo components were automatically disabled when the SUV was operated in autonomous mode.
Federal investigators found the Uber software completely failed to accurately detect Herzberg as a pedestrian and couldn't predict which path she was taking while crossing the road.
According to the NTSB report, the vehicle was traveling 44 mph 5.8 seconds before the Uber struck Herzberg.
The NTSB report says the self-driving system never classified Herzberg as a pedestrian, and she was not classified as a jaywalking pedestrian because "the system design did not include consideration for jaywalking pedestrians."
Uber and Volvo say multiple changes have been made to the self-driving test vehicles since the death of Herzberg, and both companies claim the same situation on a road today will be avoided.
Uber told NTSB investigators that new software would detect and correctly classify the pedestrian 4.5 seconds before the time of impact.