Author Topic: EEVblog #1066 - Uber Autonomous Car Fatality - How?  (Read 59593 times)

0 Members and 1 Guest are viewing this topic.

Offline timgiles

  • Regular Contributor
  • *
  • Posts: 236
  • Country: se
  • Programmer, DB architect
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #100 on: March 22, 2018, 08:20:55 pm »
Seemed to me the person walked out in front of the car. I doubt the car with headlights on was hard to see / notice - they chose to cross at the point where there was no light, in front of a moving vehicle...

So long as automated cars can be shown to drive as safe as your average human driver, I dont seem why we should hold them more accountable. The insurance companies will point us in the direction as to how safe they are, and it is likely in the not too distant future that they will start pricing non automated cars off the roads as they become significantly safer. Indeed it will become almost impossible to justify driving a car when the automated pilot can do so.

For now, it is a tragic accident, it needs investigating and lessons learnt, but every car company deciding to stop testing of autonomous cars seems extremely short sighted.
 

Offline Bob Sava

  • Regular Contributor
  • *
  • Posts: 98
  • Country: us
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #101 on: March 22, 2018, 08:27:03 pm »
Based on line markings (10ft per line segment) camera footage seems to either show that car lights are improperly adjusted (pointing down) or that the camera is not showing how dark it really was. 

Either way LIDAR does not depend on headlight illumination (or daylight) as it's projecting it's own light (laser beam).  Therefore this was definitively a system failure and program should be halted and system retested on closed circuit.

Personally I think they should make all autonomous vehicles emit a warning sound or light so pedestrian know that this car will not behave or see like human drivers do (I bet this pedestrian expected the car to slow down as visibility was good - to human). 



« Last Edit: March 22, 2018, 08:31:21 pm by Bob Sava »
 

Offline Monkeh

  • Super Contributor
  • ***
  • Posts: 7990
  • Country: gb
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #102 on: March 22, 2018, 08:30:36 pm »
Personally I think they should make all autonomous vehicles emit a warning sound or light

It has an engine and the lights are on. If you're stupid enough to think it's going to stop for you in that situation you deserve it.
 

Online wraper

  • Supporter
  • ****
  • Posts: 16792
  • Country: lv
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #103 on: March 22, 2018, 08:31:22 pm »
(I bet this pedestrian expected the car to slow down as they did many times before).
I bet the only thing that could be expected from car driver is swearing at such dumbfuck.
 

Offline Bob Sava

  • Regular Contributor
  • *
  • Posts: 98
  • Country: us
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #104 on: March 22, 2018, 09:02:05 pm »
Without sensor data this video is quite useless.

The video shows the car to be driving on a road having no visibility ahead, and with no long range lights enabled, can we agree on this?
I also wonder if you drive a car yourself. :| Bonus question.
Sure, we can agree on that, of course. Long range lights may have helped in this situation, but in the city area using long lights is not always possible due to other traffic or pedestrians. In this situation human vision might have been somewhat better than the video camera (cannot know how much better, this is just speculation), but nevertheless the self-driving cars are using sensor data to detect obstacles and other vehicles / pedestrians / just name it, so the sensor data is the key here.

Sure, I drive. We have dark fall/winter when there is no snow yet, and driving in the poorly lit areas is pretty demanding, especially when the roads are wet and absorb all light - or if someone is driving towards me with headlights reflecting from wet asphalt straight into eyes. Hitting a moose in the country side in dark roads is not uncommon either. Quite often I do hope to have some sensor fusion available.

The reason why long range lights are not needed in the city is that there are street lights and speed limit is lower.  In this case weather conditions were excellent and there no external circumstances preventing lidar from operating. 

Bottom line, the system *probably* failed to detect object on the road and *definitively* failed to react.
 

Offline ez24

  • Super Contributor
  • ***
  • Posts: 3082
  • Country: us
  • L.D.A.
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #105 on: March 22, 2018, 09:18:01 pm »

Although this film has been around a long time, when an autonomous car can do this, then it might be ready for public release.

Exciting video - anyone know what car it was?  I hope the driver got 20 years in jail for this.  My bet is in 200 years an autonomous car will be able to do this but with electric motors, but I will also bet I will not be around to see it.
YouTube and Website Electronic Resources ------>  https://www.eevblog.com/forum/other-blog-specific/a/msg1341166/#msg1341166
 

Online wraper

  • Supporter
  • ****
  • Posts: 16792
  • Country: lv
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #106 on: March 22, 2018, 09:43:05 pm »
though this film has been around a long time, when an autonomous car can do this, then it might be ready for public release.

This was 7 years ago:


« Last Edit: March 22, 2018, 09:45:28 pm by wraper »
 

Offline ez24

  • Super Contributor
  • ***
  • Posts: 3082
  • Country: us
  • L.D.A.
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #107 on: March 22, 2018, 09:48:33 pm »
FYI

From what I know - the victim was homeless, maybe this is why she was not at a crosswalk?

- The driver was an ex-felon female (not male).  And I think with a violent conviction and I think served 4 years.  This was the biggest surprise for me.  I thought Uber were using trained techs to do this job, not felons  (in general I am not against felons but I have met many (my tenants) and none would be qualified for this).   

If Uber uses felons to test their cars then, in my opinion, they should not be in business.  I am sure this is going to set back things for many years.  Regulators will have a field day on this.

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. 

If I need to correct my facts let me know and I will.

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

Offline Smokey

  • Super Contributor
  • ***
  • Posts: 2536
  • Country: us
  • Not An Expert
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #108 on: March 22, 2018, 10:33:40 pm »

It's a two lane road, and an object in the other lane as a car is driving along isn't unusual. I'm sure the LIDAR saw it but how is the car supposed to know the object is about to move in front of an approaching car? It makes no sense. Somewhere in the software there has to be some assumptions that other road users simply don't do that. The car would be constantly braking otherwise and could never overtake anything.

The real question is: How could the pedestrian not see the approaching car? They just walked straight across the road without even looking. No human would have expected that either.

This... This, this, this, this!

This isn't about if the Lidar was working, it is about what predictions were made based on the available data from all systems and the information about the surroundings.
We ALL make predictions about the world around us based on the rules of reality.  How many times would you have been in a terrible accident and likely killed on your way into work this morning if the car next to you on the highway went crazy (broke the rules) and turn hard into the side of your car at 65MPH?  Likely 100% of the time.  What extra sensors would save you?  How many times were you actually in that possible situation in just this one drive to work?  100s of times if not more, in just this one ride?  With something that potentially deadly, why is that not the greatest ever present fear in all of us all the time every time we get behind the wheel???  Because we make the prediction that people will stay in their lane and follow the rules of the road.  Is it a good prediction?  Well at least most of us reading this made it where we were going today (and the vast majority of days like it), but some small amount of people actually did get hit/killed in that same time in this exact way.  This is the fault of the person that broke the rules by which we make predictions, not the person expecting the rules to be followed. Just because it's a computer and not a person, and just because they have more sensors, doesn't mean they aren't making the exact same predictions, and relying on everything around them to fundamentally follow the rules.

Now in this particular case you need to look at the situation and the reasonable predictions being made.
The car does have an impressive set of sensors, and under optimal conditions those sensors all contribute to the total perception of the surroundings.  When multiple sensors contribute the same feedback [camera sees and identifies a person, lidar sees an object of some sort (lower rez), radar sees an object (again low rez)] then confidence can be high what some thing is and how you can expect it to act.
Based on the video, the visual light cameras could not see the person in the shadows, at least not well.  Mark that off as one of the sensor contributions to the feedback.  The camera based computer vision is probably the most capable of determining what something actually is since it has the highest resolution representation of the details of a thing.  Colors, faces, textures, etc are all the domain of the computer vision that the other sensors can't "see". 
The camera system, even multiple cameras in an array, is less good at distance and velocity determination.  That's where the Lidar comes in.  Excellent quick distance measurement but terrible resolution.  It gives you a point cloud, which some computer has to put back together and.. .you guessed it... make a prediction of what that object could be.  Did the Lidar detect there was some sort of object over there?  We don't know yet, but I would guess it probably did.  The question now becomes, without the visual camera data what did the Lidar see and what assumptions did the computer make based on that sensor data about the probable object?

Question1: What can be reasonably expected to be in the left lane of a two lane highway in the middle of the highway far between street crossings? 
Answer1: If you said a person with no regard for their own surroundings wandering between lanes at night dragging a bike loaded up with stuff with no lights on, then you are crazy.  That is a poor prediction by person or computer.  Probability of that is low.  Very low.  It would be a better prediction to think it was a couch or some other junk that fell off someone's truck that's blowing around in the street, which is more probable.
A much better prediction would be a motorcycle of some sort.  Probability of that is fairly high.  You would expect a motorcycle to be in the street, and expect it to be in the left lane, possibly stopped or moving slowly in the beginning of a left turn lane.  In the place where the accident took place, the highway opened up from 2 lanes into 4 lanes with 2 left turn lanes on the left.
Question2: What actions can you reasonably expect a motorcycle to take while in a left turn lane?
Answer2: If you said purposefully swerve into a car driving at a normal highway speeds in the right lane without any regard or hesitation, then again, you are crazy.  Probability is incredibly low.  A much better prediction would be merging back into the left through lane from the beginning of the left turn lane.  Probably of that is good.  This would give you the profile of a side view of a motorcycle moving at slow speeds, which is something the Lidar could tell on it's own and matches what we see in the video.

Now I'm not saying this is a perfect exoneration of the computer.  For example; is it a good prediction for a motorcycle to have no lights on at night?  Probably not, but it's also not a terrible prediction.  Vehicles are pretty routinely found driving at night with no lights on.

Should the computer have realized it has sub-optimal sensor information without the full camera detail, and there was a questionable condition of a vehicle with no lights on in the adjacent lane, and therefore proceeded with more caution?  Maybe, but not 100% certainly does that mean it needs to take evasive action.  Maybe it did reduce it's speed to what it considered optimal based on the predictions in the seconds before the video.   I'm sure we will hear from the Uber team eventually, but if the prediction was a vehicle such as a motorcycle (which I think is a reasonable prediction by human or computer standards given the conditions and probabilities) that doesn't warrant the follow up prediction people are making here that the object will try to jump in-front of the car and demands panic breaking.  Now why the breaks didn't lock up once it was clear the object was going to get hit is another question, but it might come out that they did and it's just not obvious from the video.  I'm willing to give them the benefit of the doubt on that one since it's not clear.  Dave was the one that synced the driver with the front video and that could be off.

Here is a completely incomplete list of things we (or a computer) would never attempt if we had to be 100% certain of every prediction or be able to take 100% successful evasive action:
1) Never enter a building of any kind.  It could collapse, it could light on fire, an airplane could crash into it, etc. 
2) Never turn on a light switch.  You could be electrocuted, you could start a fire, etc...
3) Never walk on the sidewalk.  A car could hop the curb, you could fall in a hole, a tree could fall on you, etc...
4) Never drive on a highway.  65MPH is too fast to react to all possible situations.
5) Never drive on a city street.  35MPH is too fast to react to all possible situations.
6) Never get in your car at all.  Breaks could fail, steering wheel could pop off, gas could explode, etc....

All of these things WILL happen, but that doesn't mean we need to account for them in everyday predictions.  Self driving cars will kill people.  Period.  It will happen fairly often when the technology is more prevalent.  There is no way around that with 2ton steel boxes moving at high speeds.  The question is whether they kill people less than people drivers kill people, which will most certainly be the case. 

TL;DR
Don't break the rules and walk in-front of moving cars.  More sensors doesn't ensure perfect predictions 100% of the time.
« Last Edit: March 22, 2018, 10:38:48 pm by Smokey »
 

Offline Bud

  • Super Contributor
  • ***
  • Posts: 6877
  • Country: ca
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #109 on: March 22, 2018, 10:37:56 pm »
It is difficult to read such long posts. Chances are people will skip at least 50% of it.
Facebook-free life and Rigol-free shack.
 

Offline Smokey

  • Super Contributor
  • ***
  • Posts: 2536
  • Country: us
  • Not An Expert
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #110 on: March 22, 2018, 10:40:26 pm »
Ya, that didn't start out that way.  Sorry.  Slow day at work :)
 

Offline tpowell1830

  • Frequent Contributor
  • **
  • Posts: 863
  • Country: us
  • Peacefully retired from industry, active in life
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #111 on: March 22, 2018, 11:28:29 pm »

Although this film has been around a long time, when an autonomous car can do this, then it might be ready for public release.

Exciting video - anyone know what car it was?  I hope the driver got 20 years in jail for this.  My bet is in 200 years an autonomous car will be able to do this but with electric motors, but I will also bet I will not be around to see it.

 Ferrari 275 GTB
Agreed on the electric motor part, but I think 50 years we will see autonomous vehicles (they may not be cars as we know them).
« Last Edit: March 22, 2018, 11:30:29 pm by tpowell1830 »
PEACE===>T
 
The following users thanked this post: ez24

Online wraper

  • Supporter
  • ****
  • Posts: 16792
  • Country: lv
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #112 on: March 22, 2018, 11:47:28 pm »

Although this film has been around a long time, when an autonomous car can do this, then it might be ready for public release.

Exciting video - anyone know what car it was?  I hope the driver got 20 years in jail for this.  My bet is in 200 years an autonomous car will be able to do this but with electric motors, but I will also bet I will not be around to see it.
With the same (=high) probability of accident they can do it right now.
 

Offline ez24

  • Super Contributor
  • ***
  • Posts: 3082
  • Country: us
  • L.D.A.
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #113 on: March 22, 2018, 11:51:32 pm »
Ferrari 275 GTB

Thanks for the video - exciting.  Since there is a red car in your profile and you knew what car it was  - is it your car ?

If not, do you know if the driver was a professional?  Now to start a new topic - How to make my 2000 Corolla sound like a Ferrari 275 GTB.    Uber needs to make their cars sound like a 275, then I would use them, felons or not.

FYI  I once had a taxi drive like this across Taipei in the 70s.  I told him I was in a big hurry,  the ride was just like the video so it brought back memories.  But it was not a 275.



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

Offline tpowell1830

  • Frequent Contributor
  • **
  • Posts: 863
  • Country: us
  • Peacefully retired from industry, active in life
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #114 on: March 23, 2018, 12:02:38 am »
Ferrari 275 GTB

Thanks for the video - exciting.  Since there is a red car in your profile and you knew what car it was  - is it your car ?

If not, do you know if the driver was a professional?  Now to start a new topic - How to make my 2000 Corolla sound like a Ferrari 275 GTB.    Uber needs to make their cars sound like a 275, then I would use them, felons or not.

FYI  I once had a taxi drive like this across Taipei in the 70s.  I told him I was in a big hurry,  the ride was just like the video so it brought back memories.  But it was not a 275.

I wish I had a Ferrari 275 GTB, but the pic is my 1966 Ford Mustang GT.
This was done in 1976 and the driver was a professional Formula 1 driver, however, it is rumored that the actual vehicle was a Mercedes 450 SEL 6.9 with the Ferrari soundtrack dubbed over it. Who knows?

http://www.marotprodriving.com/film-ferrari-275-gtb-paris/
PEACE===>T
 
The following users thanked this post: ez24

Offline SilverSolder

  • Super Contributor
  • ***
  • Posts: 6126
  • Country: 00
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #115 on: March 23, 2018, 12:51:30 am »
This isn't about if the Lidar was working, it is about what predictions were made based on the available data from all systems and the information about the surroundings.

You are on to something here, it seems to me.  The software maintains a model of what it thinks the reality around the car looks like.  Perhaps it mistook the sideways facing pedestrian/bicycle combination for a larger vehicle with its front pointing towards the autonomous vehicle...  perhaps the absence of lights made it think it was a stationary/parked object...  suddenly, the vehicle started to move sideways...  does not compute.

What the system seems to be missing is a "gut feeling that something is not right" (sensor inputs are becoming inconsistent with the model of the surrounding world) which should have led it to slow down, or at least be prepared to slow down very quickly.

« Last Edit: March 23, 2018, 12:54:00 am by SilverSolder »
 

Offline tpowell1830

  • Frequent Contributor
  • **
  • Posts: 863
  • Country: us
  • Peacefully retired from industry, active in life
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #116 on: March 23, 2018, 02:21:45 am »
This isn't about if the Lidar was working, it is about what predictions were made based on the available data from all systems and the information about the surroundings.

You are on to something here, it seems to me.  The software maintains a model of what it thinks the reality around the car looks like.  Perhaps it mistook the sideways facing pedestrian/bicycle combination for a larger vehicle with its front pointing towards the autonomous vehicle...  perhaps the absence of lights made it think it was a stationary/parked object...  suddenly, the vehicle started to move sideways...  does not compute.

What the system seems to be missing is a "gut feeling that something is not right" (sensor inputs are becoming inconsistent with the model of the surrounding world) which should have led it to slow down, or at least be prepared to slow down very quickly.

Yes, to put it another way, the friggin crap failed and killed a human... this is no way to move forward. And yet, the corporate engine will prevail and the testing will continue. The marketing fucks will spin it positive and the arsehole board will say (in their own minds) 'We can profit from a positive spin'. WTF!

Sorry about rant, but the wheel keeps turning. We do need a different mode of ground travel that is expedient, cost effective and safe (not necessarily in that order). The autonomous car could be the answer (and electric EZ24), but not now.
PEACE===>T
 

Online wraper

  • Supporter
  • ****
  • Posts: 16792
  • Country: lv
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #117 on: March 23, 2018, 02:31:51 am »
Yes, to put it another way, the friggin crap failed and killed a human... this is no way to move forward.
That stupid human (who actually deserved it by own recklessness) would be killed regardless if it was autonomous car or driven by human driver. How about banning vehicles as such? So many people die under the tires. Also lets ban elevators and trains as people get killed there too. Heck, lets ban electricity, it sometimes causes fire or electric shock with human casualties.
 

Offline Monkeh

  • Super Contributor
  • ***
  • Posts: 7990
  • Country: gb
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #118 on: March 23, 2018, 02:33:14 am »
Yes, to put it another way, the friggin crap failed and killed a human...

The friggin' crap failed to protect the human from itself.
 

Online Rasz

  • Super Contributor
  • ***
  • Posts: 2616
  • Country: 00
    • My random blog.
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #119 on: March 23, 2018, 03:00:56 am »
I took the video that was released and just asked them before playing the video to say "BRAKE" whenever they saw the person. I might have converted a few people.

good thing drivers arent observing the road thru $10 chinese dashcraps then, right?

That view of stepping out from the shadows may be misleading. A human may have had better vision than the vision shown by the camera footage.


I really doubt it.

this is what human sees at same spot same time of night:



Who logs in to gdm? Not I, said the duck.
My fireplace is on fire, but in all the wrong places.
 

Offline james_s

  • Super Contributor
  • ***
  • Posts: 21611
  • Country: us
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #120 on: March 23, 2018, 03:10:04 am »
This is another reason why I don't think fully atonymous cars will be put into mainstream use any time soon. Some politician has to sign off on that, and they like covering their arse. Look at how the virtually never want the raise the speed limits or remove speed humps etc once they have lowered them, no one wants to take responsibility.

As I've said for a long time, I suspect fully autonomous cars are at least 20 years behind where a lot of enthusiasts think. The companies making these things are going to have to watch out for arrogance and use an abundance of caution when deploying them because what can go wrong, will. Even if they do cause fewer deaths overall than individual drivers, it's easier to assign blame to a specific individual while it will only take a few incidents like this before the whole concept of an autonomous car has a reputation for killing people. It could turn out to be a classic textbook case like that of the Therac-25 incidents.
 

Online wraper

  • Supporter
  • ****
  • Posts: 16792
  • Country: lv
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #121 on: March 23, 2018, 03:13:58 am »
this is what human sees at same spot same time of night:
This is what good camera sees if long exposure is used, not human eye. I dod not notice anything like this when streets are lit with similar lights.
 

Online Rasz

  • Super Contributor
  • ***
  • Posts: 2616
  • Country: 00
    • My random blog.
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #122 on: March 23, 2018, 03:28:43 am »
this is what human sees at same spot same time of night:
This is what good camera sees if long exposure is used, not human eye. I dod not notice anything like this when streets are lit with similar lights.

it was taken by someone who specifically selected exposure to produce same picture he saw with his own eyes, plus you have YT clip of that road at night and someone elses pictures on twatter https://twitter.com/thekaufaz/status/976686336877871104

_this is a well lit road_
Who logs in to gdm? Not I, said the duck.
My fireplace is on fire, but in all the wrong places.
 

Offline Fungus

  • Super Contributor
  • ***
  • Posts: 16560
  • Country: 00
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #123 on: March 23, 2018, 04:32:29 am »
Off topic, but...


Although this film has been around a long time, when an autonomous car can do this, then it might be ready for public release.

Exciting video - anyone know what car it was?

It was a Mercedes S-class. The tire squeals and sound of a Ferrari engine were dubbed on afterwards (listen for the occasional mismatch/crossfade in the sound).

If you take away the camera angle and sound of a Ferrari going through the gears then it's not as fast as you think (mute the sound and look at the objects going past, not at the road surface...)

People have calculated real speeds:

https://en.wikipedia.org/wiki/C%27%C3%A9tait_un_rendez-vous#Route

Ferrari 275 GTB

Nope.

Admittedly, the streets were fairly empty in this film.

It was the 1970s, very early Sunday morning. It would be impossible to do these days - there's a lot more cars in the world now.

FYI  I once had a taxi drive like this across Taipei in the 70s. 

I've done 150mph/250kph in a Taxi in Germany.

I don't know what the fuel consumption was, I wonder if it was profitable?
« Last Edit: March 23, 2018, 05:57:31 am by Fungus »
 

Offline hendorog

  • Super Contributor
  • ***
  • Posts: 1617
  • Country: nz
Re: EEVblog #1066 - Uber Autonomous Car Fatality - How?
« Reply #124 on: March 23, 2018, 05:07:52 am »
So:

Failure of the LIDAR
and failure of the system which detects failures in the LIDAR
and failure of the human backup driver
and failure of the camera to capture the scene
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf