Self Driving Uber Kills Pedestrian

Chalo said:

That kind of democracy only works when you don't have 95% of the people born into and indoctrinated as road pirates.

Click to expand...

I can't remember the day I was indoctrinated. I must have missed it.

billvon said:

Chalo said:

That kind of democracy only works when you don't have 95% of the people born into and indoctrinated as road pirates.

Click to expand...

I can't remember the day I was indoctrinated. I must have missed it.

Click to expand...

Yep.

Did anyone watch my post on the Google car video?
https://youtu.be/tiwVMrTLUWg?t=12m33s
Here the google car sees a cyclist being an issue from a massive distance away, while the cyclist at that point is about 11 car lanes worth of distance away, thats about 50meters.
The short story here is that there is no way in hell this accident would have happened if it was a google car.

I think the reason why Google decided to show off this coincidentally quite a similar incident as its last example of issues it can deal with is to show how smart the car is. This is from the LIDAR that sits at the very top of the vehicle for maximal height and 360degree surrounding laser-based vision that it doing all the visual looking around and deciding what to do. What you see in that Uber human eye perspective video dashcam doesn't mean shit, the Uber self-driving car would not be looking at that at all, its real eyes are LIDAR based.

Also as this police vehicle video cam shows, this Uber car just powers straight through the red light, Uber self-driving cars are a joke compared to the Google Cars.
https://youtu.be/pzzQ42D9Srw

[youtube]pzzQ42D9Srw[/youtube]


https://www.reuters.com/article/us-uber-selfdriving-sensors-insight/ubers-use-of-fewer-safety-sensors-prompts-questions-after-arizona-crash-idUSKBN1H337Q

I wonder what she was thinking in the last moments and if she went instantly or did she survive the initial impact but tapped out on the landing.

Hows the insurance work on that, if it was auto driving, who's to blame. Interesting legal argument, no doubt owners insurance. I hope the poor lady had family so her estate can get a few million bucks.

Uber has already settled with the family of the cyclist.

https://www.reuters.com/article/us-...ly-of-autonomous-vehicle-victim-idUSKBN1H5092

Addy said:

Uber has already settled with the family of the cyclist.

https://www.reuters.com/article/us-...ly-of-autonomous-vehicle-victim-idUSKBN1H5092

Click to expand...

Good to here, for a 60billion dollar market cap company, they should be easily able to afford compensation, especially if they're losing billions a year.
The thing that shocked me is leaked data showed self-driving Uber cars in Arizona were "struggling" to go 13 miles between interventions by a safety driver. That is just so crap.
https://www.nytimes.com/2018/03/23/technology/uber-self-driving-cars-arizona.html
For Googles cars it was 5,600 miles before the driver had to take control from the computer to steer out of trouble.
So 5600 / 13 = 430.
So Googles self driving cars are 430 times more autonomous/safer.
Do people still believe considering the compensation and the 430 times less reliable that this was a "legitimate kill" via the Uber car?

Here is the spot in Google Street view, ironically she has basically made it to the bike lane lines start when she got killed.
https://goo.gl/maps/xXy47Ud48Fw

Also a few meters back about where LIDAR should most certainly should have seen her (LIDAR sees in the dark) there is a bike sign on the road that says "Yield to Bikes"
https://goo.gl/maps/oCLVtZ3i6YB2


The creepy thing is you can actually see this "Yield to Bikes" sign in the first few frames of this gif video, sign is on the right.


Also Ubers newer cars have far less sensors, probably to save money. The car that killed only had one single LIDAR, but their previous test car had 7 LIDARS.
https://www.reuters.com/article/us-uber-selfdriving-sensors-insight/ubers-use-of-fewer-safety-sensors-prompts-questions-after-arizona-crash-idUSKBN1H337Q
https://fingfx.thomsonreuters.com/gfx/rngs/UBER-SELFDRIVING-SENSORS/010061BR2TH/UBER-SELFDRIVING-SENSORS.jpg

For me this thread as been a real eye opener, everyones "trust the self-driving car/computer and the private company its pedestrians own fault and this is a legitimate kill."
I really was afraid when this happened people on here would feel this was a legit kill. Even though this forum has hardened me against people who never agree with me I am still overall shocked.

Alan B said:

This a case where the lights might have been on high beam had the human been driving. That might have made a big difference.

Click to expand...

I used to live in Tempe and don't remember any place where I'd have used high beams.

It is common for people to overdrive their headlights, travelling at speeds faster than they can react and stop in the distance illuminated by the headlights (especially to smaller hazards). This is fine where other lighting provides adequate visibility, but not otherwise. Watching cars pass me on the highway at night when I was ebiking showed that a significant fraction of them didn't have effective lights at all, even at 50+ MPH on a dark unlit highway. They can see well enough to stay in the lane, but they are really trusting the lane is clear, not really being safe. My ebike lights were more effective than many cars', and it was still challenging to see debris and animals in the bike lane early enough even at 28.

It would seem that each company investigating self driving cars should be evaluated for quality of work and only those passing fairly high standards be allowed to continue. Uber's motivations, approach and track record to this development seem less than optimal.

Alan B said:

They can see well enough to stay in the lane, but they are really trusting the lane is clear, not really being safe.

Click to expand...

Unfortunately, in my experience, the part i bolded is a significant portion of the problems with many (perhaps most) road or path users of any kind, be they car/truck drivers or cyclists or even pedestrians.

The drivers of really large vehicles like buses and delivery trucks of various kinds seem to do a better job of watching even farther ahead (probably because due to mass their reaction time is forcibly slowed and they're aware of it), but drivers of stuff that's SUV sized and smaller (all the way down to foot traffic) typically seem to just assume everything is always going to be ok until it just isn't.


I can't count the number of times I've been waiting to turn (left or right) or cross some intersection (usually but not always at a red light), and someone approaching behind me doesn't think they should have to wait, but can't see the conditions on the cross-road, and just go around me. Sometimes they collide with the unseen oncoming traffic, but much more often they either manage to slam on their brakes in time or they accelerate really hard and just gun it into or across the traffic stream as the cross traffic manages to avoid them one way or another (sometimes colliding with each other rather than the cause of the whole thing).

Sometimes the reason I was waiting was crosswalk traffic (or jaywalkers further down), that someone behind me can't see due to angles and/or obstructions. When the impatient person behind me goes around in those cases, it's always been a really really close call as they screech to a halt or swerve hard to avoid killing the crossers (who can't move fast enough to avoid the driver), and sometimes the driver wrecks their vehicle in the process.

The same thing happens when it's another car waiting there to turn rather than me, but since I don't fill the lane even with SB Cruiser, it's much more likely a driver will go around me than it is for them to go around another car/truck.


If everyone assumed there would be problems everywhere along the way and was looking for them all the time, rather than assuming the entire road or path is all theirs forever and always with nothing else on it anywhere, roads and paths would be significantly safer.

But AFAICT, that's not how people generally think, and whenever I've discussed it with people, even friends or coworkers, some of them seem genuinely surprised (even angry) that they should think any other way than "why should I care", and aren't going to change, and most of them never even thought about it at all, and many of *those* don't really see any problem with how things are, since they've never killed anybody yet. Some agree with me, and some even change how they drive/walk/etc to be more aware of their surroundings. Some of *those* even stay that way more than a week. :/

Don't know if ppl have been carefully reading my posts but I will make this more obvious and clear.
The car that hit the cyclist continued to drive for another 40-50 meters AFTER hitting her as the video shows the incident happening right at the water drains/"yield to bikes" sign and the car/bike sitting about 40-50 meters down in front of the blue sign.

So not only did the car not slow down it continued on for about 40-50 meters after the crash.

Again, the video shows sitting right at the "Yield to bikes" sign next to the drains https://goo.gl/maps/oCLVtZ3i6YB2

And this is where all the crash scene screen shots/news video show where the damaged bike and damaged car sit
https://goo.gl/maps/xXy47Ud48Fw

All in one google street view, Yield Sign, Drains,blue sign. https://goo.gl/maps/t36neVhYanF2


^ Click to enlarge

^ Click to enlarge

Couple of interesting short statements from other self-driving / driver-assistance system manufacturers on this collision:

In a related development, the chief of Intel's rival self-driving programme suggested its Mobileye tech would have prevented the accident.

"We ran our software on a video feed coming from a TV monitor running the police video of the incident," blogged Amnon Shashua.

"Despite the suboptimal conditions, where much of the... data that would be present in the actual scene was likely lost, clear detection was achieved approximately one second before impact."

The Bloomberg news agency has reported that the Volvo cars used by Uber feature Mobileye chips and sensors, but their normal driver assistance system was disabled, according to the auto-parts' supplier.

The head of Google's autonomous car division Waymo has also said its tech would have been "able to handle" such situations.

Velodyne - the firm that designed the collision-avoidance sensors that Uber employs - previously told the BBC that it was "baffled" by the accident because its equipment was capable of seeing in the dark.

Click to expand...

That's quite something if the Volvo's standard, basic collision-avoidance system would have prevented or lessened the crash!

Hehe... Fun phrase... "suboptimal conditions"... to describe watt conditions are like over 99% of the time. :wink:

A new one, this a company called Waymo. http://www.azfamily.com/clip/14324579/video-another-self-driving-vehicle-involved-in-crash

http://www.azfamily.com/clip/14325821/video-waymo-releases-video-moments-before-self-driving-car-accident

https://www.bloomberg.com/news/articles/2018-09-17/self-driving-cars-still-can-t-handle-bad-weather

It's someday going to get closer to understanding what we see automatically. Someday.

Maybe we just need self-driving pedestrians that are networked with the cars.

You know that kind of makes sense. Not connecting the pedestrian themselves to the car persay, but maybe the car could "see" pedestrians by detecting the radio signals transmitted by their cell phone before being able to actually see them with it's Lidar, etc. Or when they're hiding behind an obstruction that would prevent the other systems from being accurate.