Author Topic: EEVblog #1066 - Uber Autonomous Car Fatality - How?  (Read 37198 times)

0 Members and 1 Guest are viewing this topic.

Offline FrankBuss

  • Supporter
  • ****
  • Posts: 2348
  • Country: de
    • Frank Buss
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #150 on: March 23, 2018, 02:21:50 pm »
Someone who is an adviser for the robotics cars team at Google says, there is a rumour that LIDAR was turned off (search for "rumour", but the rest of the analysis is very good as well) :

Don't these cars perform a self test, and simply don't drive (at-least warn) when systems are turned-off / not working ?

Checklist comes to mind  ::)

If the rumour is right, they turned it off intentionally to test operation with just the camera and radar.
So Long, and Thanks for All the Fish
Electronics, hiking, retro-computing, electronic music etc.: https://www.youtube.com/c/FrankBussProgrammer
 

Offline Fungus

  • Super Contributor
  • ***
  • Posts: 12606
  • Country: 00
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #151 on: March 23, 2018, 02:28:09 pm »
I agree with what you said. When I said simulated miles I meant the endless clocking up of miles is worthless.

Checking for slowly accumulating memory leaks/fragmentation...?  :popcorn:

 

Offline Brumby

  • Supporter
  • ****
  • Posts: 11154
  • Country: au
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #152 on: March 23, 2018, 02:29:08 pm »
I'd say that reaction is either from emergency braking or from the impact. Unfortunately the video is cut off right at that frame so I can't decide which (maybe both?)

What happens after the cut is irrelevant to the point I was making.

Two points:
 1. The person is clearly looking to the left of the car.  The impact would not have happened.
 2. There is no body movement in the seat that braking or evasive action would cause.  The car's velocity did not change during the video.


The whole way the 'interior' video is cut makes me think Uber is hiding something,
The cut may not be for that reason.  It may have been to prevent further stress on the safety driver.

Quote
eg. Did the car brake at all? :popcorn:
THAT, however, is a very good question.
 
The following users thanked this post: TheDane

Offline Rasz

  • Super Contributor
  • ***
  • Posts: 2546
  • Country: 00
    • My random blog.
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #153 on: March 23, 2018, 02:32:21 pm »
If the rumour is right, they turned it off intentionally to test operation with just the camera and radar.

Tesla drove for over a year on that combination using repurposed mobileye line following tech, with occasional driver decapitation :'(
Who logs in to gdm? Not I, said the duck.
My fireplace is on fire, but in all the wrong places.
 

Offline apis

  • Super Contributor
  • ***
  • Posts: 1667
  • Country: se
  • Hobbyist
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #154 on: March 23, 2018, 02:36:18 pm »
To be clear, do you think anything you said is in conflict with what I said?

The latest videos I've seen from Google still seem to focus on their high res maps being correlated with lidar and radar images to fine tune the GPS guidance. I still can't find much about how they deal with a world they didn't expect, like roadworks or a massive crash blocking the way.
When I hear GPS, I think of the satellite system and a receiver providing coordinates and time. Not one of those talking navigator units with built in maps etc, although they might be called simply GPS colloquially. Those things rely solely on GPS for localisation and often have outdated maps and crappy software. That is not how these cars work, although no doubt the cars have path planning algoritms and maps of course.

There really isn't anything "solely" about the use of the GPS [in google's self driving cars]. The cars can use the other sensors to precisely predict it's position and drive for long periods without GPS [they have to in order to be able to drive in tunnels and cities where GPS reception is not always available]. (If the other sensors alone are more precise than the GPS that in itself indicate they don't need the GPS.) I'm pretty sure they can drive entirely without GPS but they might have problem finding their initial location after a reboot without the GPS, if the car was moved from it's last known position while turned off, or something like that. I don't know enough details to speculate about that though.

Being able to read road signs, plan a new route if a road is blocked or handle a situation where the map doesn't agree with reality doesn't really have anything to do with GPS. I don't know how they handle that actually, but it's my impression they do not only rely on static data (in fact it's hinted at in that video as well, although indirectly). If you know of a source with details about that it would be much appreciated. Finding alternative routes is something all route planning software do though, can't imagine it would be a deal-breaker.
« Last Edit: March 23, 2018, 03:28:07 pm by apis »
 

Offline Fungus

  • Super Contributor
  • ***
  • Posts: 12606
  • Country: 00
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #155 on: March 23, 2018, 02:58:22 pm »
The whole way the 'interior' video is cut makes me think Uber is hiding something,
The cut may not be for that reason.  It may have been to prevent further stress on the safety driver.

a) Is somebody forcing them to watch it in a loop?

b) Having that face posted all over the Internet isn't stressful?

I suspect that the only reason they posted the interior shot at all was to show that the "safety driver" wasn't watching the road. What other reason is there for releasing it?


 

Offline AF6LJ

  • Supporter
  • ****
  • Posts: 2902
  • Country: us
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #156 on: March 23, 2018, 03:44:40 pm »
I wouldn't get near Uber, they don't do background checks on their drivers.
Some of us have standards.
 :popcorn:
Sue AF6LJ
Test Equipment Addict, And Proud Of It.
 

Offline SilverSolder

  • Super Contributor
  • ***
  • Posts: 5146
  • Country: 00
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #157 on: March 23, 2018, 05:32:57 pm »

The airline failure model isn't really the same as cars, if there's a mechanical failure on an airliner you can't simply pull over and call a tow truck.

Well, what is being investigated here is the reasons that a fatal accident occurred.  There were multiple factors that enabled this accident to happen, including the bicyclist crossing the road unsafely...  everyone had a part to play here. 

The issues around safety are reminiscent of the issues around data security:  adaptive security controls at various layers is essential to securing systems in today's environment.

 

Offline darrellg

  • Supporter
  • ****
  • Posts: 103
  • Country: us
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #158 on: March 23, 2018, 06:01:35 pm »
If anyone has a Toyota with TSS-P - would their car avoided this?  TSS-P has pedestrian avoidance.
I had a distracted accident (100% my fault) and someday I want to get a car with AEB  but only Toyota seems to be the one I can get someday. 
Some Honda models are available with Honda Sensing, which has the Collision Mitigation Braking System and is supposed to brake when it senses pedestrians. I haven't "tested" mine yet.
 

Offline ez24

  • Super Contributor
  • ***
  • Posts: 3082
  • Country: us
  • L.D.A.
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #159 on: March 23, 2018, 06:06:00 pm »
Was he paying attention and doing his job responsibly in this situation?  If the babysitter is found to have not been suitable for the task (right personality, right training), this is an area where Uber is vulnerable too.

FYI  Both the victim and Uber driver were female.  The victim was homeless and the driver was a felon.  I think both of these facts will make the investigation more difficult, especially if the victim does not have any family to sue.  I feel really sorry for her.

Does anyone know of any official site that would be good to go to for the ongoing facts?  Who will be the lead in the investigation - I hope it is not the police department.



YouTube and Website Electronic Resources ------>  https://www.eevblog.com/forum/other-blog-specific/a/msg1341166/#msg1341166
 

Offline coppice

  • Super Contributor
  • ***
  • Posts: 6504
  • Country: gb
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #160 on: March 23, 2018, 06:08:32 pm »
Some Honda models are available with Honda Sensing, which has the Collision Mitigation Braking System and is supposed to brake when it senses pedestrians. I haven't "tested" mine yet.
For the current European Honda CR-V the sensing package is a costly option, which you can only add to the top spec version. With the new CR-V, appearing in the summer, sensing will be standard across the range. I find that a interesting data point for how much collision avoidance features are working their way into the market. I guess the NCAP rating system is a significant factor pushing this. They can't keep those 5 star ratings on new models if they don't have at least some automated safety kit.
 

Offline coppice

  • Super Contributor
  • ***
  • Posts: 6504
  • Country: gb
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #161 on: March 23, 2018, 06:13:12 pm »
Was he paying attention and doing his job responsibly in this situation?  If the babysitter is found to have not been suitable for the task (right personality, right training), this is an area where Uber is vulnerable too.

FYI  Both the victim and Uber driver were female.  The victim was homeless and the driver was a felon.  I think both of these facts will make the investigation more difficult, especially if the victim does not have any family to sue.  I feel really sorry for her.

Does anyone know of any official site that would be good to go to for the ongoing facts?  Who will be the lead in the investigation - I hope it is not the police department.
Who but the desperate would take a job babysitting an autonomous car? It has to be one of the most boring, soul destroying jobs around. It is implausible that whoever you put behind that wheel, with nothing much to do all day but look ahead, is going to maintain full attention hour after hour.
 

Offline apis

  • Super Contributor
  • ***
  • Posts: 1667
  • Country: se
  • Hobbyist
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #162 on: March 23, 2018, 06:18:26 pm »
If anyone has a Toyota with TSS-P - would their car avoided this?  TSS-P has pedestrian avoidance.
I had a distracted accident (100% my fault) and someday I want to get a car with AEB  but only Toyota seems to be the one I can get someday. 
Some Honda models are available with Honda Sensing, which has the Collision Mitigation Braking System and is supposed to brake when it senses pedestrians. I haven't "tested" mine yet.
Found a video of such a system being tested:
http://www.bosch-presse.de/pressportal/de/en/emergency-braking-in-two-blinks-of-an-eye-121600.html
It's not clear at what speed the car is driving and the max speed it's able to stop in time, but visibility clearly doesn't matter to the radar here.
 

Offline apis

  • Super Contributor
  • ***
  • Posts: 1667
  • Country: se
  • Hobbyist
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #163 on: March 23, 2018, 06:25:56 pm »
Was he paying attention and doing his job responsibly in this situation?  If the babysitter is found to have not been suitable for the task (right personality, right training), this is an area where Uber is vulnerable too.

FYI  Both the victim and Uber driver were female.  The victim was homeless and the driver was a felon.  I think both of these facts will make the investigation more difficult, especially if the victim does not have any family to sue.  I feel really sorry for her.

Does anyone know of any official site that would be good to go to for the ongoing facts?  Who will be the lead in the investigation - I hope it is not the police department.
Who but the desperate would take a job babysitting an autonomous car? It has to be one of the most boring, soul destroying jobs around. It is implausible that whoever you put behind that wheel, with nothing much to do all day but look ahead, is going to maintain full attention hour after hour.
Yeah, and one reason you are sitting there is just to take the blame if something goes wrong... pretty thankless job.
« Last Edit: March 23, 2018, 06:29:12 pm by apis »
 

Offline Fungus

  • Super Contributor
  • ***
  • Posts: 12606
  • Country: 00
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #164 on: March 23, 2018, 06:38:28 pm »
Found a video of such a system being tested:
http://www.bosch-presse.de/pressportal/de/en/emergency-braking-in-two-blinks-of-an-eye-121600.html
It's not clear at what speed the car is driving and the max speed it's able to stop in time, but visibility clearly doesn't matter to the radar here.

Here's another one:


And another:


 

Offline ez24

  • Super Contributor
  • ***
  • Posts: 3082
  • Country: us
  • L.D.A.
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #165 on: March 23, 2018, 06:38:34 pm »
Some Honda models are available with Honda Sensing, which has the Collision Mitigation Braking System and is supposed to brake when it senses pedestrians. I haven't "tested" mine yet.

In my area (So Calif)  Honda Fit EV also has AEB  (about 40k !).  I believe Edmunds rates Honda Sensing as poor.  One thing about AEB is that it is not a standard.  For example Sentra with AEB tops out at 50 MPH.  Toyota makes the most makes with full AEB.  In general you have to read the owners manual to get the facts.  The dumbest people to ask are the salesmen.  The technology is so new the salespeople cannot keep up with it.

Consumer Reports is the best place to start looking for AEB.   My bet is a Toyota with TSS-P would not have hit the victim,  TSS-P  is full range AEB with pedestrian avoidance.  Someday I hope to get a car with AEB.

https://www.consumerreports.org/car-safety/cars-with-advanced-safety-systems/

YouTube and Website Electronic Resources ------>  https://www.eevblog.com/forum/other-blog-specific/a/msg1341166/#msg1341166
 

Offline darrellg

  • Supporter
  • ****
  • Posts: 103
  • Country: us
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #166 on: March 23, 2018, 06:52:41 pm »
Some Honda models are available with Honda Sensing, which has the Collision Mitigation Braking System and is supposed to brake when it senses pedestrians. I haven't "tested" mine yet.
For the current European Honda CR-V the sensing package is a costly option, which you can only add to the top spec version. With the new CR-V, appearing in the summer, sensing will be standard across the range. I find that a interesting data point for how much collision avoidance features are working their way into the market. I guess the NCAP rating system is a significant factor pushing this. They can't keep those 5 star ratings on new models if they don't have at least some automated safety kit.

I have a US 2018 Honda Civic Hatchback, and it's available at all trim levels as a $1000 option. I didn't buy it for the emergency braking, but for the adaptive cruise control.
 

Offline apis

  • Super Contributor
  • ***
  • Posts: 1667
  • Country: se
  • Hobbyist
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #167 on: March 23, 2018, 07:06:20 pm »
Found a video of such a system being tested:
http://www.bosch-presse.de/pressportal/de/en/emergency-braking-in-two-blinks-of-an-eye-121600.html
It's not clear at what speed the car is driving and the max speed it's able to stop in time, but visibility clearly doesn't matter to the radar here.

Here's another one:
[snip]

And another:
[snip]
The second one is pretty interesting, the Volvo is doing great but it uses both a camera and a single radar sensor, while the Mercedes one uses three radar sensors but still fails miserably. According to this http://ideas.4brad.com/it-certainly-looks-bad-uber that FrankBuss found, radar sensors doesn't have enough resolution to be reliable on their own. And apparently Uber might have turned off the lidar to test how their system worked without it.
« Last Edit: March 23, 2018, 07:08:58 pm by apis »
 

Offline tpowell1830

  • Frequent Contributor
  • **
  • Posts: 821
  • Country: us
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #168 on: March 23, 2018, 08:19:17 pm »
Found a video of such a system being tested:
http://www.bosch-presse.de/pressportal/de/en/emergency-braking-in-two-blinks-of-an-eye-121600.html
It's not clear at what speed the car is driving and the max speed it's able to stop in time, but visibility clearly doesn't matter to the radar here.

Here's another one:
[snip]

And another:
[snip]

One of these videos is from 2010 and the other is from 2013. I wonder if there are more recent improvements? The fellow with the Tesla video on one of the previous pages shows the Tesla several hundred feet back when the alarm goes off.

The video of the uber accident seems to show that the car did not alarm or brake at all, but it would be nice to see video from the actual system camera rather than the dinky dashcam video. If the vehicle had even started braking the nose would have plunged down, as you see in these videos.
PEACE===>T
 

Offline gearshredder

  • Contributor
  • Posts: 36
  • Country: us
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #169 on: March 23, 2018, 08:54:45 pm »
In the video the cyclist appears from the darkness at the very last moment, as though the dipped headlights were dipped way too far. If a human were driving on that dark road they would have had their high beams on, and it looks like they would have seen the cyclist reasonably early, when it was still realistic to stop. I don't think the "human driver would have hit this cyclist anyway" argument, which I have seen, holds water.

It probably wouldn't have mattered due to that extra reaction time by the driver but I agree, the headlights look poorly adjusted. You can't even see the guys feet before it's too late to stop. It's a HID system with projectors, they'd have no problems seeing them if they were properly set. Looks like the car at factory has them poorly adjusted. I have mine set A lot higher, Just below the mirrors on a small car. I retrofitted some Morimoto Mini d2s projectors into my 93 Impreza.
« Last Edit: March 23, 2018, 09:12:49 pm by gearshredder »
 

Offline taydin

  • Frequent Contributor
  • **
  • Posts: 515
  • Country: tr
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #170 on: March 23, 2018, 09:12:58 pm »
As a safety critical system, that car MUST HAVE and probably does have multiple means of detecting an obstruction (LIDAR, LASER, thermal camera, maybe others). And the design of the sensor grid would be such that a fault in one sensor won't bring down the other sensors. Based on these assumptions, I'm concluding that AT LEAST ONE sensor system has reacted to the pedestrian's presence and reported it to the main computer.

Whatever happened, must have happened in the algorithm of the main computer that looks at all the sensor data and makes car control decisions.
Real programmers use machine code!

My hobby projects http://mekatronik.org/forum
 

Offline Fungus

  • Super Contributor
  • ***
  • Posts: 12606
  • Country: 00
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #171 on: March 23, 2018, 09:16:23 pm »
It probably wouldn't have mattered due to that extra reaction time by the driver but I agree, the headlights look poorly adjusted.

If only there was a way to automatically select high beams based on ambient lighting and oncoming traffic.
 

Offline r3bers

  • Contributor
  • Posts: 26
  • Country: ru
    • solovjov.net
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #172 on: March 23, 2018, 09:26:02 pm »
All kangaroos in AU must have GPS now...
 

Offline maginnovision

  • Super Contributor
  • ***
  • Posts: 1962
  • Country: us
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #173 on: March 23, 2018, 09:27:43 pm »
It probably wouldn't have mattered due to that extra reaction time by the driver but I agree, the headlights look poorly adjusted.

If only there was a way to automatically select high beams based on ambient lighting and oncoming traffic.

Pfft, maybe in 100 years. If uber and the safety driver get off scot free something bad is happening with this situation. I can't believe the POLICE are the ones releasing such a deceptive video. Also the driver is a man, long hair and manboobs don't change that because you add an A to your name.
 

Offline pickle9000

  • Super Contributor
  • ***
  • Posts: 2216
  • Country: ca
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #174 on: March 23, 2018, 09:56:40 pm »
I'd like to see some laws giving the cars some minimum standards. First LIDAR is a critical system. No LIDAR pull over and stop the vehicle. Decision time on critical events. and so on. This could be a big selling point as well.

As for the driver. Sitting behind a wheel for say 2-3 hours with nothing to do and expecting him to react to a situation like this? Can't see it. I for one would not expect him to have any kind of reaction time close to a driver operating a real vehicle. He has little situational awareness.

If there was a critical system being tested (disabled) was the driver made aware so he could pay more attention?

There is a great deal unknown.

 
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf