Slashdot Mirror
Uber's Self-Driving Car Saw Pedestrian 6 Seconds Before Fatal Strike, Says Report (tucson.com)
An anonymous reader quotes a report from Arizona Daily Star: The autonomous Uber SUV that struck and killed an Arizona pedestrian in March spotted the woman about six seconds before hitting her, but did not stop because the system used to automatically apply brakes in potentially dangerous situations had been disabled, according to federal investigators. In a preliminary report on the crash, the National Transportation Safety Board said Thursday that emergency braking is not enabled while Uber's cars are under computer control, "to reduce the potential for erratic vehicle behavior." Instead, Uber relies on a human backup driver to intervene. The system, however, is not designed to alert the driver.
The report comes a day after Uber announced it will be ending it's self-driving vehicle testing in Arizona. The full NTSB report is available here.
It's time for regulatory control - like how the FAA regulates the skies, the autonomous car industry needs to have sensible regulations. Right now it just seems like a bunch of cowboys in the wild west are trying to one-up each other.
Get this: The car "saw" the person 6 seconds before striking the person. The emergency system was disabled. The emergency system was not set up to alert the driver. So many things wrong here that would have been avoided had there been sensible regulations written by sensible engineers.
The woman is a hero for all intents and purposes. Like any newly deployed technology, a few people will be killed while we work out all of the kinks. We've lost early adopters before. Everything from astronauts to roller coaster riders to test pilots. It's unfortunate but it comes with the territory and it's the price of progress. In the end we'll have much safer cars and the world will be in debt to people like this victim. May she rest in peace and may the Lord provide solace to her family.
What did you expect from a company like Uber, exactly? Their entire business model was built around the concept of "laws don't apply to us, because, internet." This is the same company that beyond running an unregistered taxi service (cheaply disguised as "ride sharing", as if people just happened to be going in the same direction), was tracking critical journalists with "God Mode" and hiring private investigators against them to find things to blackmail them with, ran an active campaign to track law enforcement officers in order to evade them, ran a campaign against competitors like Lyft involving the massive use of fake ride requests, and literally dozens and dozens more types of general scumbaggery.
Anyone shocked that this same company shut off all pedestrian safety controls in order to get their "self driving cars" on the market sooner? Bueller?
Give a boy a gun and you arm him for a day. Teach him how to make a gun, and the whole metaphor breaks down.
Apparently the way they had it was that the computer would drive and the driver would stop it from driving, if needed. That doesn't seem like an obviously ridiculous arrangement, even if having the computer ping the driver would have been better.
I suspect that the alerting the driver wasn't gonna be any better. This is all speculation, but if Uber turned off the AEB (automatic emergency braking) system in self driving mode because it would have actuated the brakes too often making the driving erratic, simply notifing the driver about the same potential collisions would eventually result in alert fatigue in the driver.
Then the driver would start ignoring the alerts for basically the same reason that they turned off the AEB system originally.
If you've ever taught someone to drive who was overly cautious and braked all the time you know how that goes. The flaw is that simply seeing the pedestrian isn't enough to make the braking decision and the driver has to learn to anticipate the actions in the environment to drive successfully.
The driving instruction not only has to anticipate the environment, but also the driver to know when to intervene (grab steering wheel or apply the brake). Most self driving systems don't give these driving observers (not instructors) enough information and training to anticipate what the system is going to do to make them effective at intervening.
This is the problem of allowing people to walk before they can crawl... Sometimes you can't bypass stages of learning. If the system's AEB system activates too much, you kind of have to let it do that until it can learn not to do that.
According to the FA she wasn't texting, she was minding the screen in the car, per Uber's instructions, to take notes on interesting events for later research. Uber used to have two people, one to watch the road and one to take the notes and they did away with the second one to cut costs.
Hi! I make Firefox Plug-ins. Check 'em out @ https://addons.mozilla.org/en-US/firefox/addon/youtube-mp3-podcaster/
How would that have helped? The car detected her and didn't brake,
As he pointed out, they're talking about two different systems. It's hard to parse out, but I think what they're trying to say is that the built in factory system detected her but couldn't break because it was disabled, but Uber's own system didn't detect her.
If that's the case then yeah, maybe reflectors and such would have helped the Uber system detect her and stop. On the other hand if the factory system detected her and theirs didn't then it also means that their system is pretty shit.