Tempe police have just released a footage of a driverless Uber car killing a jaywalker while crossing the street with her bike. Elaine Herzberg, 49, was taken to the hospital immediately but died from her injuries shortly afterward.
The video shows the woman behind struck by the autonomous car as well as the safety driver’s reaction. The driver seems to have spotted Herzberg a fraction of second before the accident occurred.
One question remains. Why didn’t Uber’s state-of-the-art Lidar sensor system detect the victim in time? The video shows that she obviously didn’t run in front of the car. Uber boasted that Lidar can spot any obstacle on the road even in the dark as it is equipped with infrared sensors for nighttime view.
Steven Shladover, a UC Berkeley researcher who specialized in self-driverless cars, thinks that the car’s sensors should have detected the pedestrian. The car was moving at 40 miles per hour at the moment of the accident.
Driverless Uber Car’s Sensors at Fault
Shladover couldn’t understand how the system failed to detect a pedestrian that was slowing pushing a bike on the road and was not doing any erratic moves. If Herzberg had walked erratically, the car’s AI system would have had a hard time in spotting her.
Also, the pedestrian walked just in front of the car. If she had approached the vehicle from an odd angle, the system could have been confused since it would have been harder to tell if the pedestrian was moving in the car’s path.
In the latest crash, the pedestrian was fully visible and about to obstruct the car’s path. What’s more, she was clearly moving towards the car as she had crossed a lane of road before she was hit in the right lane.
Shladover thinks that the fault lies with the cars sensors, the AI system that interprets the data from those sensors, or both.
Image Source: Flickr