Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Any details on headline - Arizona pedestrian is killed by Uber self-driving car

This site may earn commission on affiliate links.
Krafcik Says Waymo Tech `Robust Enough' to Avoid Uber Crash

Waymo Chief Executive Officer John Krafcik said his company’s self-driving software would likely have avoided the pedestrian death in which an Uber autonomous car was involved.

“We have a lot of confidence that our technology would be robust and would be able to handle situations like that one,” Krafcik said on Saturday in his first public comments since a pedestrian was killed by an Uber self-driving vehicle last Sunday night.
 
  • Informative
Reactions: immunogold
Oooh snap, Mobileye also called out Uber for messing with the video: "...video from a dash-cam that was subjected to some unknown downsampling."

Both Waymo and Mobileye recently issued statements that are essentially throwing Uber to the lions.

The problem with this whole thing is we don't know if it was a detection issue or a failure elsewhere in the system. Like a SW bug in untested code.

I imagine it's beyond frustrating for Waymo/Intel/Cruise to deal with the likes of Uber. They absolutely have to separate themselves from Uber, and I really don't think it will take much convincing.

This fatality wasn't the result of some edge case, or some fundamentally flawed aspect of the way self-driving cars are being tested. This is simply a case of gross incompetence.
 
  • Like
Reactions: zmarty and harry
Accelerated life-cycle testing for all road vehicles is well established

suspension-durability-280.jpg


These facilities will need to develop sections of track which test autonomous systems, ideally in a range of weather / light conditions.

There will be the usual problem of manufacturers designing to the test, much as they do now with emissions, but it would help to set minimum standards.
 
Accelerated life-cycle testing for all road vehicles is well established

placeholder_image.svg


These facilities will need to develop sections of track which test autonomous systems, ideally in a range of weather / light conditions.

There will be the usual problem of manufacturers designing to the test, much as they do now with emissions, but it would help to set minimum standards.

IIHS or NHTSA could create a self driving safety test suite similar to the one for web browser compliance. Multiple test cases, non simulator environment, only a summary provided to manufacturer to reduce gaming/ over fitting.

Later....
That expands to an American Gladitors type TV show where autonomous cars compete in an obstacle course and it all becomes self funding....
 
According to:

Uber’s use of fewer safety sensors prompts questions after Arizona...

Current Uber Volvo vehicles only have 1 lidar for each car while GM has 5 and Waymo has 6.

Prior Uber Ford Fusion had:

7 lidars
7 radars
20 cameras

But current Uber Volvo's sensors count are down for lidars and cameras:

1 lidar (decreased from 7)
10 radars (increased from 7)
7 cameras (decreased from 20)

The problem with current Uber Volvo is: the car is too high, much higher than the previous Uber Ford Fusion so a single roof lidar cannot cover low ground due to vertical beaming limitation.

When switching from Ford Fusion to Volvo the logical way should be to increase the numbers of lidar to cover the low ground as well and not a reduction.
 
According to:

Uber’s use of fewer safety sensors prompts questions after Arizona...

Current Uber Volvo vehicles only have 1 lidar for each car while GM has 5 and Waymo has 6.

Prior Uber Ford Fusion had:

7 lidars
7 radars
20 cameras

But current Uber Volvo's sensors count are down for lidars and cameras:

1 lidar (decreased from 7)
10 radars (increased from 7)
7 cameras (decreased from 20)

The problem with current Uber Volvo is: the car is too high, much higher than the previous Uber Ford Fusion so a single roof lidar cannot cover low ground due to vertical beaming limitation.

When switching from Ford Fusion to Volvo the logical way should be to increase the numbers of lidar to cover the low ground as well and not a reduction.

"Velodyne acknowledged that with the rooftop lidar there is a roughly three meter blind spot around a vehicle, saying that more sensors are necessary."

So that wouldn't be a direct factor in killing Elaine Herzberg. She was much further out when the identification and braking needed to start happening and should have been fully within the LIDAR's sensing torus, right?

EDIT: Further looking at that sensor list they might have switched to a strategy of using radar in cover in close, maybe with some camera augmentation? That's in line with the later portion of the article. Remember Tesla, although still operating at Level 2, isn't using LIDAR at all. They see it as a crutch that if you design your system to it will lead to a dead end (AKA "local minimum").

EDIT2: If the extra height from the Volvo is indeed creating problems sensing the bottom of pedestrians at that distance on level ground? I'm not sure how it would have been reliable even on the Focus during elevation changes, where the sensor's angle to distant ground would change. That'd make the LIDAR little more than ornamental at best, a dangerous "feel good" at worst.
 
Last edited:
  • Like
Reactions: MP3Mike and brkaus
I've watched a video of this incident (didn't find it in this thread, maybe I'm just blind) and it seemed to me the driver wasn't paying too much attention. I understand he was relying on the technology however it was very dark outside and the person came out of nowhere. Based on what I'v seen in the video, I'd have an issue seeing that person myself however the car should see them I think.
 
  • Disagree
Reactions: ℬête Noire
I've watched a video of this incident (didn't find it in this thread, maybe I'm just blind) and it seemed to me the driver wasn't paying too much attention. I understand he was relying on the technology however it was very dark outside and the person came out of nowhere. Based on what I'v seen in the video, I'd have an issue seeing that person myself however the car should see them I think.

The other video you haven't found that is also linked in this thread is the video taken by someone else using their phone driving the same chunk of road at night showing that that video is quite deceptive in just how dark that area is. It's a typical urban road, lit up like a Christmas tree. Something is VERY wack with the gamma on that Uber video that the police released.

I don't see off-hand where that is but here's a still shot of the area. Any details on headline - Arizona pedestrian is killed by Uber self-driving car



P.S. The driver isn't a "he".
 
I've watched a video of this incident (didn't find it in this thread, maybe I'm just blind) and it seemed to me the driver wasn't paying too much attention. I understand he was relying on the technology however it was very dark outside and the person came out of nowhere. Based on what I'v seen in the video, I'd have an issue seeing that person myself however the car should see them I think.

In adddition to Bete Noire's comment you must have also missed in the thread the video of the investigators/police using the same Uber Volvo and the victim's bike and placing the bike where it was hit and see if the Volvo could brake in time before hitting the bike. Said they ran it 5 times and it was able to brake in time each time. Test was done at night with Uber Volvo coming from down by the bridge at the same speed. Videos in the thread. And the test driver could see the bike just fine. The dashcam video the police initially saw which I think you did too makes the dark of the night look blacker than it is to human eye. Typical for many dashcams. The lady was walking across 5 traffic lanes from left to right and got hit in the last lane over.
 
  • Informative
  • Like
Reactions: zmarty and Tam
Didn't look up AZ road rules but here's what Calif road rules are with respect to crosswalks and pedestrians in the roadway:

"A crosswalk is the part of the roadway set aside for pedestrian traffic. Most intersections have a pedestrian crosswalk whether or not lines are painted on the street. Most crosswalks are located at corners, but they can also be located in the middle of the block. Before turning a corner, watch for pedestrians about to cross the street. Pedestrians have the right-of-way in marked or unmarked crosswalks. Although pedestrians have the right-of-way, they also must abide by the rules of the road. A pedestrian should not suddenly leave a curb, or other place of safety, and cross into the path of a vehicle as this creates an immediate hazard. Furthermore, a pedestrian must not stop unnecessarily or delay traffic while in a crosswalk."

Given that the area she crossed from had those paved walking paths that ended at the roadway's curb, I guess you could view that as a legitimate crosswalk area, at least in California.
 
I was surprised there was a settlement so quickly but had read initially someone saying Uber should make their best offer right off the bat and maybe they did. It's possible the family also doesn't have the stomach for dealing with lawyers and protracted litigation. I know if I had been killed and it was up to my mother she would think to hire a lawyer but wouldn't likely want to spend years in court.
 
Last edited:
  • Like
Reactions: ℬête Noire
I was surprised there was a settlement so quickly but had read initially someone saying Uber should make their best offer right off the bat and maybe they did. It's possible the family also doesn't have the stomach for dealing with lawyers and protracted litigation. I know if I had been killed and it was up to my mother she would think to hire a lawyer but wouldn't likely want to spend years in court.
Also, if their attorney is working on spec (paid only by a % of the settlement, which is typical of family of modest means vs a corp) then the deeper they go into the process the higher the percentage the attorney gets. So there are diminishing returns. No doubt Uber's council was giving marching orders of "here's a ridiculous bag of money, make this go away before discovery", maybe even "make this go away before any digging that risks the politicians in our pocket get pulled into this". Oops on the later but the former is likely worth a LOT of $ in PR terms and potentially fueling info for other potential future lawsuits down the road.
 
  • Like
Reactions: zmarty
The other video you haven't found that is also linked in this thread is the video taken by someone else using their phone driving the same chunk of road at night showing that that video is quite deceptive in just how dark that area is. It's a typical urban road, lit up like a Christmas tree. Something is VERY wack with the gamma on that Uber video that the police released.

I don't see off-hand where that is but here's a still shot of the area. Any details on headline - Arizona pedestrian is killed by Uber self-driving car



P.S. The driver isn't a "he".

Of all the things in this thread that could blow my mind, the fact that the driver is a woman is what did it. I swore that was a man.