Uber self-driving car fatality blamed on faulty software

Uber seems to have determined their own software was, unsurprisingly, the primary cause of last months fatal self-driving car accident.  In that accident a pedestrian crossing a street was struck by an experimental Uber self-driving car, and died of injuries shortly thereafter.  According to sources within Uber cited by “The Information“, a news site covering Silicon Valley, the determination is from Uber’s investigation, but this would be a preliminary decision because the NTSB and NHTSA investigation has not concluded.

The Information’s post is hidden behind a paywall, but TechCrunch has a summary.  That summary reads much like what I wrote last month.

Uber’s self-driving software is making poor decisions and to put it bluntly is not fully baked yet.  So, why is it being used on regular roads?

In TechCrunch’s summary, they suggest two possible scenarios for that particular failure:  A) The “object recognition” system failed to detect the pedestrian pushing their bicycle across the road, or B) higher level reasoning systems decided the detected object was a false positive and should be ignored.

According to sources cited by The Information, the fault was the latter.  That the car did detect the pedestrian pushing her bicycle across the street, but decided it wasn’t worth paying attention to that object.

The decision matrix involved is very complex.  For example it would be inappropriate if autonomous vehicles stopped for every plastic bag blowing across the street.  Somehow the software must decide whether a given object detected by sensors is a risk, or is a nothing.  And it must do this, in real time, using digital video cameras, and with data captured by LIDAR devices.   And it must do this without the benefit of human knowledge and experience.

Another point to make is those two scenarios ignore the blaring reality of the third cause.  The human being sitting behind the steering wheel was clearly not paying attention.  Dashcam video shows that person, who is supposed to be an attentive driver, instead interacting with a device held at waist level, probably a cell phone in a messaging application or a game.  Every so often that human is glancing up at the road until the last time when he reacts in horror as to what’s happened.  That’s the image we see at the top of this posting.

Uber and the relevant government authorities (the NHTSA) will eventually present a report.  At that time we’ll have a clearer story than the leak published by The Information.

According to Reuters, Uber has hired the assistance of a former NTSB Chair.  Uber is quoted as saying “We have initiated a top-to-bottom safety review of our self-driving vehicles program, and we have brought on former NTSB Chair Christopher Hart to advise us on our overall safety culture.  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.”

Those are nice words.  Really nice words.  Well crafted words.  Will it result in any real change?

 

 

About David Herron

David Herron is a writer and software engineer living in Silicon Valley. He primarily writes about electric vehicles, clean energy systems, climate change, peak oil and related issues. When not writing he indulges in software projects and is sometimes employed as a software engineer. David has written for sites like PlugInCars and TorqueNews, and worked for companies like Sun Microsystems and Yahoo.

About David Herron

David Herron is a writer and software engineer living in Silicon Valley. He primarily writes about electric vehicles, clean energy systems, climate change, peak oil and related issues. When not writing he indulges in software projects and is sometimes employed as a software engineer. David has written for sites like PlugInCars and TorqueNews, and worked for companies like Sun Microsystems and Yahoo.

Leave a Reply