Scrutiny Turns Toward Uber After First Pedestrian Death by Autonomous Vehicle

It's not just the autonomous vehicle technology, but also the company that operated the self-driving SUV receiving scrutiny after the death of Elaine Herzberg in Tempe, Az. on March 18.
March 31, 2018, 5am PDT | Irvin Dawid
Share Tweet LinkedIn Email Comments
Foo Conner

The details of the death of 49-year-old Elaine Herzberg, who was walking her bike across North Mill Ave when she was struck by an Uber SUV operating in autonomous mode with a backup driver, are under investigation by the National Transportation Safety Board.

Initially, the Tempe police chief, after viewing the videos, suggested the crash may have been unavoidable, be it from human or self-driving vehicle. After the two videos were made available to the public, scrutiny was directed at autonomous vehicle (AV) technology, with some experts claiming that fully autonomous vehicles were not ready to be operated on public streets. A pedestrian fatality was "inevitable," and the public was being treated like "guinea pigs," some charged.

Now Carolyn Said, who writes on autonomous vehicles for the San Francisco Chronicle, and other journalists have turned their attention to the company, not only its AV technology but also the "company’s procedures and corporate culture," distinguishing Uber from other developers of autonomous vehicles.

"Uber’s haste has already cost it hundreds of millions of dollars in research spending and legal fees and settlements, writes Said on March 23. "Now some ask if it has cost a human life."

“Uber has major explaining to do” about how its system failed so catastrophically, said Michael Ramsey, research director at analysis firm Gartner. Neither Uber’s technology nor its backup driver seemed ready for the open road, experts said.

Herzberg's presence should have been captured by the Volvo's lidar, a laser form of radar, and other sensory equipment.

“This was one of the simplest possible scenarios an autonomous vehicle could be in,” said Sam Abuelsamid, senior analyst at Navigant Research in Detroit. “There was no other traffic, no obstructions, no bad weather and the woman was in the middle of the roadway. There were no extenuating circumstances. She should have been very clearly visible to the sensors on this vehicle.”

In other words, it was exactly the kind of situation in which autonomous cars with their sensors are supposed to outperform humans.

"Marta Thoma Hall, president of Velodyne LiDAR Inc., maker of the special laser radar that helps an autonomous car 'see' its surroundings, said the company doesn’t believe its technology failed," reports Ken Naughton for Bloomberg News [may require free registration].

“We are as baffled as anyone else,” Thoma Hall wrote in an email. “Certainly, our Lidar is capable of clearly imaging Elaine and her bicycle in this situation. However, our Lidar doesn’t make the decision to put on the brakes or get out of her way.”

Said of the Chronicle explains:

There are several ways the [automated driving] system could have gone wrong — but all reflect poorly on Uber. The sensing systems (lidar, radar and cameras) could have failed to detect Herzberg. The classification systems could have thought she was an inanimate object (although the question remains: Why not stop?). The actuation system that controls braking and steering could have not received a message to slow or swerve.

Said delves further into the technology, and also the inattentiveness of Rafaela Vasquez, the 44-year-old backup driver behind the wheel, who was shown in the inward-facing dashcam video looking downwards the entire time before the crash. It has even been speculated that an attentive driver would have seen Ms. Herzberg and evaded the crash, according to forensic crash analysts who reviewed the video.

Uber's self-driving division was struggling even before the Tempe pedestrian crash, according to Daisuke Wakabayashi ofThe New York Times. "Uber’s human drivers had to intervene far more frequently than the drivers of competing autonomous car projects," he writes on March 23.

Waymo, formerly the self-driving car project of Google, said that in tests on roads in California last year, its cars went an average of nearly 5,600 miles before the driver had to take control from the computer to steer out of trouble. As of March, Uber was struggling to meet its target of 13 miles per “intervention” in Arizona [according to documents obtained by The New York Times.]

Unlike California, which requires self-driving car companies to report intervention numbers, "Arizona does not have reporting requirements," writes Wakabayashi, calling it a "regulatory vacuum."

That might be starting to change though, due to Herzberg's death. Gov. Doug Ducey (R) of Arizona ordered a suspension of Uber's autonomous vehicle testing effective March 26, reports Wakabayashi, calling it "a reversal from what has been an open-arms policy by the state, heralding its lack of regulation as an asset to lure autonomous vehicle testing — and tech jobs."

Uber had already suspended self-driving testing in Arizona, San Francisco, Pittsburgh and Toronto following Ms. Herzberg's death.

Full Story:
Published on Friday, March 23, 2018 in San Francisco Chronicle
Share Tweet LinkedIn Email