On Sunday at approximately 10:00 p.m., a self-driving Uber vehicle struck and killed a Tempe pedestrian. Police report that the preliminary investigation shows that the vehicle was traveling at approximately 40 mph, with no significant signs of slowing down when it struck and killed a woman.
The vehicle, a 2017 Volvo SUV, was travelling north bound on Mill Road. The pedestrian, Elaine Herzberg, age 49, was walking a bicycle across the road and was struck by the vehicle when she entered the lane in which the the vehicle was travelling. She was pronounced deceased at the hospital.
The vehicle was manned by Raphaela Vasquez, age 44. The vehicle was in autonomous mode at the time of the collision, but Vasquez was behind the wheel. According to police, Vasquez showed no signs of impairment.
Expert Lionel Robert, University of Michigan associate professor of information, who leads a study that’s investigating how and whether pedestrians trust autonomous vehicles, stated, “Driving is a social activity that involves explicit and implicit communications between humans. Drivers of cars can predict the behavior of pedestrians in part because they have been pedestrians. The same goes for pedestrians who have been drivers.”
“That understanding facilitates a social communication process where both pedestrians and drivers can anticipate the future actions of each other and act accordingly. We are still struggling with how to teach an autonomous vehicle to be social and interpret the behavior and intentions of pedestrians,” concluded Roberts.
Anuj Pradhan, assistant research scientist at the University of Michigan Transportation Research Institute, studies autonomous vehicle human factors. Pradhan, who is involved in research investigating pedestrian trust in autonomous vehicles stated, “The main issue is communicating intent between autonomous vehicles and other road users. This is an important issue that is still being researched,” he said. “The autonomous vehicle needs to infer the intent of a pedestrian from limited information about the pedestrian’s speed, heading, gait, and a host of other features—something that a human driver does quite naturally.”
“Similarly, the pedestrian has to be able to detect the intent of an autonomous vehicle from its speed, heading, slowing-down behaviors, etc,” said Pradhan. “If it were a human driver, this would normally include eye-contact. Current deployments of autonomous vehicles do not communicate their intent externally, other than with the signals we have in all vehicles such as brake lights and turn signals. Other critical factors could be the expectations of other road users—with potentially disastrous results if the expectations of the road user does not match the actual behavior, capabilities, or limitations of the autonomous vehicle.”