Uber Software Classified Pedestrian as a False Positive, Decided Not to Avoid

Software glitch might be responsible for the severity of the Tempe crash 1 photo
On March 20, an Uber self-driving car became the world’s first such system involved in a deadly pedestrian crash. A 49-year-old Tempe resident was killed while jaywalking in the middle of the night in Tempe, Arizona.
Officially, the blame for the incident has been pinned on the pedestrian, with local authorities saying “it would have been difficult to avoid this collision in any mode based on how she came from the shadows right into the roadway.

Even If humans make mistakes, automated systems are supposed not to. And even if the crash was nearly unavoidable, as seen in the video released by Tempe police in the days following the accident, its results might have been mitigated should the system had reacted properly.

Aside from suspending its autonomous vehicle testing, Uber started checking the software that powers the Volvo XC90. And, according to some sources, the software is partially responsible for the woman’s demise.

As per The Information, citing two unnamed sources close to the investigation, an algorithm failure led to the system “deciding” to classify the pedestrian as a false positive.

That means the car’s sensors did pick up the woman in the middle of the road, but just as it does when it comes to pedestrians walking along the side of the road or small debris, the software believed it could continue driving.

At fault for this seem to be the engineers working on the system, who failed to integrate the possibility of a human walking alongside a bike popping up in the middle of the road at night.

As a result of the crash, and possibly of the investigation’s findings, Uber hired a former chair of the National Transportation Safety Board (NTSB) to act as safety advisor.

According to Uber, Christopher Hart has been brought onboard to “advise us on overall safety culture.” Additionally, a complete review of the autonomous program will be conducted.

“Our review is looking at everything from the safety of our system to our training processes for vehicle operators, and we hope to have more to say soon,” says Uber.
If you liked the article, please follow us:  Google News icon Google News Youtube Instagram
About the author: Daniel Patrascu
Daniel Patrascu profile photo

Daniel loves writing (or so he claims), and he uses this skill to offer readers a "behind the scenes" look at the automotive industry. He also enjoys talking about space exploration and robots, because in his view the only way forward for humanity is away from this planet, in metal bodies.
Full profile


Would you like AUTOEVOLUTION to send you notifications?

You will only receive our top stories