Shortly after the Uber driverless car killed Elaine Herzberg in Tempe, Arizona, the Tempe police announced that it wasn’t Uber’s fault. “It’s very clear it would have been difficult to avoid this collision in any kind of mode based on how she came from the shadows right into the roadway,” Tempe’s police chief told the San Francisco Chronicle. However, the Chronicle noted that “the police have not released the videos.”
Based on the police description of the accident, the Antiplanner’s analysis presumed that Herzberg had just stepped into the street from the left side (where people on the curb would be obscured by vegetation) and the car was in the left lane when it struck her. This would have given any driver almost no time to see and prevent the accident. But the release of the video above reveals that the car was in the right lane when it hit her, which has led to claims that the accident was in fact avoidable and the fault belongs to Uber’s technology. In particular, Waymo’s CEO, John Krafcik, stated, “We have a lot of confidence that our technology would be robust and would be able to handle situations like that one.”
Looking at the video, there is only about one second between the moment the Herzberg’s shoes are reflected in the car’s headlights and the car hits her. So I don’t think a human driver would have fared any better that Uber’s car. And while it appears that the Uber backup driver wasn’t paying any attention, I doubt that he could have done anything even if his eyes had been on the road full time.
But a driverless car should have done better because it isn’t restricted to optical sensors. It should have radar, lidar, and/or other sensors that would have seen Herzberg as she was crossing the left lane of the street. Waymo’s use of such sensors is what gave Krafcik confidence that their system “would have been able to handle” the situation. I don’t know exactly what sensors Uber’s car has, but something between the sensors and the software failed to work as it was supposed to.
The Antiplanner’s analysis concluded that part of the fault belongs to Tempe’s design of putting pedestrian paths that cross streets at unsafe locations. The Drive‘s Alex Roy agrees. “She was jaywalking,” he observes, “from a median where she shouldn’t have been, where the city of Tempe had laid a beautiful paved walkway with a sign saying No Crossing in a city where crosswalks are hundreds of feet apart.”
Shilajit capsules is branded as cialis 5mg sale to erectile dysfunction. Third, one of the biggest roles of sitz bath are cipla generic viagra mainly composed by the following. Although it’s not entirely clear how the bacteria are transmitted, it’s likely they spread from person to person through discount on cialis the oral-fecal route or are ingested in contaminated food or water. This order viagra cheap is especially true for men listening to battered men.
Roy’s answer, however, is to “invest in mass transit . . ., invest in bike lanes, e-bikes, and hybrid and electric vehicles,” and don’t trust driverless cars because they will somehow reduce our freedom of movement. But since 1970 we’ve “invested” well over a trillion dollars in mass transit and all it has done is seen a reduction in transit usage from 50 trips per urban resident per year to 37. Moreover, Tempe has clearly invested in bike lanes, as Herzberg was crossing a street at the place where two pedestrian paths met that street. Spending more money on programs that we already know don’t work is a waste.
The Antiplanner frankly doesn’t understand Roy’s claim that driverless cars are such a threat to our freedom that he is willing to sacrifice safety to preserve that freedom. I can sort of understand it when people say that gun ownership is important for freedom, but just how does having your hands on the wheel of a car preserve your freedom? Many modern cars already drive by wire anyway, which means there is a computer between your hands and the wheels that often has the power to override your driving decisions.
As Roy himself points out, more than twice as many people are killed by cars each year than by guns. If driverless cars, like transit, only went to a small fraction of the destinations that can be reached by human-driven cars, then I could understand why Roy might be concerned. But as far as I know, no one is arguing that people should ever be forbidden from driving their own cars, only that driverless technologies have so many advantages in addition to safety that people will want to use them.
Eventually driverless cars will be able to go everywhere a human-driven car can go. When that happens, I can imagine that state and local highway departments will decide that closing some roads to human-driven cars is important for safety. But, just as there are lots of places where you can go horseback riding today, there will always be places where people can take their life, and a steering wheel, in their hands if they want to get a thrill from driving their own car.
What happened to Elaine Herzberg was a tragedy. In the short run, Tempe should respond by either fencing off the median strip so as to discourage pedestrians from crossing where Herzberg crossed or by installing a pedestrian-actuated crosswalk. In the long run, driverless cars should prevent future such accidents and make motor vehicles safer than guns.
What struck me was why didn’t the headlights see her much earlier. It looks like the headlights only illuminated a few feet before the car.
thanks
JK
The vehicle in question, Volvo’s XC90, is equipped with the company’s Pedestrian Detection in Darkness System that “combines nighttime pedestrian detection and auto-brake technology to prevent collisions with bystanders and cyclists in low-light conditions”. It seems to work in this almost identical situation: https://www.youtube.com/watch?v=xfgWE5nIyG0
The question is whether it was disabled in favor of the add-on technology.
Actually, there is a issue of freedom here. Some people are already proposing that driving be banned.
https://splinternews.com/driving-should-be-illegal-1793851503
This will be a fight in the not-too-distant future – unless, as is quite possible, self-driving cars never become all that safe. It starts the first time a human-driven car and a self-driving car meet in a fatal accident.
This failure was especially bad because Uber uses Velodyne LiDAR sensors (see e.g. https://www.forbes.com/sites/alanohnsman/2018/03/23/lidar-maker-velodyne-baffled-by-self-driving-ubers-failure-to-avoid-pedestrian/#4267adc25cc2). The car simply should have detected the pedestrian with plenty of time to stop safely. Instead, it appears never to have even decelerated.
Thank you for that link, RWS:
A video of the accident from cameras on the Uber test vehicle released by the Tempe Police Department this week shows what looks like a total system failure.
Which agrees with the Antiplanner’s conclusion. Someone screwed up, bad.
Thank you for correcting your earlier post as more information became available.
Clearly no driver or driverless car should drive faster than it can see. Video is notoriously bad at conveying exactly what can be seen under low light conditions, so you can’t take it literally. If it was a true representation, then the car was illegally going too fast and out-driving its headlights.
Human drivers have to dim headlights to keep from blinding oncoming traffic. What limits AI cars? Why aren’t they beaming high luminosity ultra violet and infra red lights?
The Tempe accident was a total failure of the Driverless Car. They will pay a huge amount of damages to keep it out of the courts, and hopefully learn from it. It will be safer eventually, but it has to be 10x safer to get acceptance.
I’m guessing that AVs are currently 10x safer than human driven cars.