Autonomous Vehicles

Entai said:
Have they not heard of take it out of gear and turn the ignition off ?????? Then slowly drift to a stop on the brakes, yes you will have no assistance on the brakes but they will work you just press the pedal a bit harder.

Although the car in question will automatically slow down to about 20mph and turn off the auto cruise when you open the car door and takeoff your seatbelt, so there should have been zero issue and danger to anyone, perhaps some of these people should actually read the handbooks given with their cars.
Click to expand...

All good points but I think there is a phenomenon that occurs when you tell a human that his/her car has autonomous features, even basic ones. That is, after a while, the human begins to trust the autonomous features and tunes out, creating really slow reaction times (even assuming they read the handbook!).

I have pointed this out on numerous occasions but this is something that Google/Waymo discovered a number of years ago when they were testing AVs with their employees in the driver seat. This is the main reason that Google skipped entirely the concept of depending on human input behind the wheel and going straight to Level 5 autonomy as their desired goal.

By the way, on your input previously about human safety drivers looking at screens showing the cars performance rather than possibly at their mobile devices (something any reasonable human would be inclined to do after hours of tedious, boring screen staring), are your human safety drivers specifically prohibited from carrying their mobile devices into the vehicle while they are testing?
 
Nasher said:
........

The problem right now with AVs is you'd be constantly on edge as the driver, poised to take over at any moment if it looks like it's going to crash. That's more stressful than just driving the car yourself.
Click to expand...

I continue to see the Uber car as the main potential defect until proven otherwise. The whole point of a Level 4/5 vehicle is to not have to depend primarily on a safety driver. Waymo's vehicles in Chandler Arizona have no safety drivers in the front seat.

I found the video to contradict the original reporting where the pedestrian was supposed to have stepped out suddenly from "the bushes" and from a parked car blocking any view of the "driver". The video seems to show a pedestrian that had already entered the traffic lane and had already crossed at least one lane of traffic on an empty street before the accident occurred. The pedestrian did not dart out from behind an obstruction. A "something" at least should have been detected as moving perpendicular to the direction of travel of the AV and the LIDAR alone should have had a number of frames to detect the collision hazard.

This just seems to be so basic to the workings of any Level 4 AV that I continue to be alarmed at how poorly Uber's AV performed. It never slowed down.
 
Entai said:
Have they not heard of take it out of gear and turn the ignition off ?????? Then slowly drift to a stop on the brakes, yes you will have no assistance on the brakes but they will work you just press the pedal a bit harder.

Although the car in question will automatically slow down to about 20mph and turn off the auto cruise when you open the car door and takeoff your seatbelt, so there should have been zero issue and danger to anyone, perhaps some of these people should actually read the handbooks given with their cars.
Click to expand...

Your probably only going to notice it's screwed up when you need to slow down, but by then it's to late.
 
sesevans said:
All good points but I think there is a phenomenon that occurs when you tell a human that his/her car has autonomous features, even basic ones. That is, after a while, the human begins to trust the autonomous features and tunes out, creating really slow reaction times (even assuming they read the handbook!).

I have pointed this out on numerous occasions but this is something that Google/Waymo discovered a number of years ago when they were testing AVs with their employees in the driver seat. This is the main reason that Google skipped entirely the concept of depending on human input behind the wheel and going straight to Level 5 autonomy as their desired goal.

By the way, on your input previously about human safety drivers looking at screens showing the cars performance rather than possibly at their mobile devices (something any reasonable human would be inclined to do after hours of tedious, boring screen staring), are your human safety drivers specifically prohibited from carrying their mobile devices into the vehicle while they are testing?
Click to expand...


We are all specifically denied taking our mobile devices on site at all.

They are locked away in lockers as soon as we enter the site and returned to us as we leave at the end of the day.
 
Just watched the video on bbc and that woman on bike would have been hit by a human driver. By the time the car's lights illuminate her it is already too late.
 
danlightbulb said:
Just watched the video on bbc and that woman on bike would have been hit by a human driver. By the time the car's lights illuminate her it is already too late.
Click to expand...

I agree, I don’t think any human drivers would’ve avoided that.

But...should a computer with IR cameras and sensors see that?
 
danlightbulb said:
See it yes, take precautionary action? maybe not. cant second guess every possible pedestrian move otherwise the car would never get above 5 mph.
Click to expand...

I believe you are setting too low a bar for what an AV on public roads today should be able to do. Simply put, Uber's self driving car failed a very basic scenario. I have talked about "corner cases" (really challenging hard to imagine edge instances) but this seems a really bread-and-butter, test track common variety situation.

I continue to hold the belief that Uber's technology is sub-standard, unless proven otherwise. Remember they just "lost" a trial to Waymo (Uber paid $ 245 million to settle and agreed not to use any Waymo LIDAR trade secrets). Could it be that they did remove some element of their LIDAR as part of that settlement that too closely infringed upon Waymo and that resulted in the accident?

Any smart lawyer acting for the family of the pedestrian will be seeking answers.
 
danlightbulb said:
If thats true then fault lies with Waymo. Since when should corporate IP and vested interests override public safety.
Click to expand...

Hardly. Uber's reputation for a "take no prisoners", "second place is for losers", and well-documented instances of questionable practices by Uber's former CEO and the head of autonomous vehicle development at Uber that he hired from Waymo and who is now the subject of a his own legal investigation, speaks volumes. Here in the UK they have had their license to operate called into question and are appealing a decision to revoke the license in several UK cities including London.

Uber is totally responsible for its own LIDAR/AV development and if it infringed on Waymo's patents and had to remove anything from their hardware/software configuration recently, they should have grounded their fleet of AVs purely on safety grounds. Waymo cannot be held accountable for the actions taken purely by Uber in pursuit of Uber's aggressive ambitions.

The point I made about whether they did have to remove any infringing hardware or software is my own speculation and not based on any factual information. However the lawyer for the plaintiff in the Uber crash would surely be pursuing this line of questioning.
 
danlightbulb said:
Just watched the video on bbc and that woman on bike would have been hit by a human driver. By the time the car's lights illuminate her it is already too late.
Click to expand...

Raymond Lin said:
I agree, I don’t think any human drivers would’ve avoided that.

But...should a computer with IR cameras and sensors see that?
Click to expand...

I disagree for a few reasons.

The camera does not have close to the sort of dynamic range or 'night vision' as a human eye which is why you don't see the pedestrian until the very end. The head lamps from and street lights will be making the camera adjust its levels to stop the directly lit sections looking blown out at the expense of detail in the shadows.

The person is walking quite slowly across quite a wide road but they were 3/4 the way across before contact was made.

There are street lights, old yellow ones too which spread the light out more and she is crossing the road within a few meters of them so would have been within their light splash. It would not have been pitch black and the driver would have been able to see her had his eyes been on the road.

The headlights would have lit her up up long before the camera shows due to the dynamic range point above.

The car was not going quickly, only 40mph.

I have been in a similar situation before, I was confronted with a cow walking across an unlit section of the A1 in the middle of the night and there was no other traffic. I was able to stop in time from 70mph (just). I did need new trousers mind.... but the difference is that I was paying attention.

I accept not everyone would react in time to stop but I would expect they would be scrubbing a lot of speed by the time of making contact.

The car clearly failed in its duty and all 3 systems failed to 'spot' the individual (radar, lidar and camera) but the driver clearly was busy monitoring the car and not looking where he was going. Is that also the fault of Uber for putting the driver in that situation we will probably never know. When you are concentrating on a screen and glancing up periodically as the driver was you would not 'see' anything in your peripheral vision where the pedestrian would come from. Looking down at a back lit screen (especially if it is showing white) will also mean it takes longer for you to 'see' dark objects correctly.

I completely accept there is a large amount of culpability on the part of the pedestrian and we will never know what was going through their mind.
 
Last edited:
Yea it's not as dark as it looks on camera with the light having a sudden cut-off like that. He would have been able to see her if he was looking and been able to stop if he was in control of the car.

Uber will do everything they can to try and pin this on the guy. But they just can't get out of this one, their AV killed someone, the end.
 
Last edited:
b0rn2sk8 said:
I disagree for a few reasons.

The camera does not have close to the sort of dynamic range or 'night vision' as a human eye which is why you don't see the pedestrian until the very end. The head lamps from and street lights will be making the camera adjust its levels to stop the directly lit sections looking blown out at the expense of detail in the shadows.

The person is walking quite slowly across quite a wide road but they were 3/4 the way across before contact was made.

There are street lights, old yellow ones too which spread the light out more and she is crossing the road within a few meters of them so would have been within their light splash. It would not have been pitch black and the driver would have been able to see her had his eyes been on the road.

The headlights would have lit her up up long before the camera shows due to the dynamic range point above.

I have been in a similar situation before, I was confronted with a cow walking across an unlit section of the A1 in the middle of the night and there was no other traffic. I was able to stop in time from 70mph (just). I did need new trousers mind.... but the difference is that I was paying attention.

I accept not everyone would react in time to stop but I would expect they would be scrubbing a lot of speed by the time of making contact.

The car clearly failed in its duty and all 3 systems failed to 'spot' the individual (radar, lidar and camera) but the driver clearly was busy monitoring the car and not looking where he was going. Is that also the fault of Uber for putting the driver in that situation we will probably never know. When you are concentrating on a screen and glancing up periodically as the driver was you would not 'see' anything in your peripheral vision where the pedestrian would come from. Looking down at a back lit screen (especially if it is showing white) will also mean it takes longer for you to 'see' dark objects correctly.

I completely accept there is a large amount of culpability on the part of the pedestrian and we will never know what was ground through their mind.
Click to expand...

Your first sentence is wrong, soooo kinda skimmed the rest, sorry.

Night vision cameras. IR cameras. Radar.

Your eyes can see in the dark? You can see with your eyes closed?

Cameras, or shall we say, technology can see much better than we can. Especially for a walking pace pedestrian with a bike.
 
Last edited:
Raymond Lin said:
Your first sentence is wrong, soooo kinda skimmed the rest, sorry.

Night vision cameras. IR cameras. Radar.

Your eyes can see in the dark? You can see with your eyes closed?

Cameras, or shall we say, technology can see much better than we can. Especially for a walking pace pedestrian with a bike.
Click to expand...

Obviously not lol
 
Raymond Lin said:
Your first sentence is wrong, soooo didn’t read the rest, sorry.

Night vision cameras. IR cameras. Radar.

Your eyes can see in the dark? You can see with your eyes closed?
Click to expand...

No.

A)
I was specifically referring to the camera recording the footage released to the media not the AV system, that clearly failed.

The footage is not from a night vision camera, its from standard colour camera you can see the grass is green, street lamps are orange/yellow. Every article I have seen reports it as a dash cam and not a part of the AV system. IR cameras are monochrome.

B) It wasn't pitch black, far from it in fact. There are headlamps and street lamps a human can see quite far in that kind of light especially if the object is moving.

I'd suggest re-reading what I actually wrote.
 
b0rn2sk8 said:
No.

A)
I was specifically referring to the camera recording the footage released to the media not the AV system, that clearly failed.

The footage is not from a night vision camera, its from standard colour camera you can see the grass is green, street lamps are orange/yellow. Every article I have seen reports it as a dash cam and not a part of the AV system. IR cameras are monochrome.

B) It wasn't pitch black, far from it in fact. There are headlamps and street lamps a human can see quite far in that kind of light especially if the object is moving.

I'd suggest re-reading what I actually wrote.
Click to expand...

I don’t care what the dash cam shows,all that tech on the roof are not dash cams. The car don’t see through via the dash cam but radars and sensors.

I don’t care if isn’t pitch black either, it could be pitch black and it should still see the same image as if it’s midday. Both dash cam or amount of light are irrelevant, or they are meant to be irrelevant.

That’s the point.
 
You are completely missing the point of what I wrote.

I was dis-agreeing with 2 posters saying a human wouldn't have been able to avoid the accident reacting to the dash cam footage.

My first point is the dash cam footage is not representative of what a human can see in the light levels shown in the video.

My second point is a human would have been able to see the pedestrian long before they were visible in the footage released to the media and potentially reacted to it to reduce the severity or completely avoid the incident in the first place.

I was not commenting at all on the AV system and what it saw, could have seen or should have seen that clearly failed to 'see' anything.

The 'safety' driver was not concentrating on the road or looking at it most of the time and spend most of the time looking at either the in car telemetry or something else equally as detracting.
 
b0rn2sk8 said:
You are completely missing the point of what I wrote.

I was dis-agreeing with 2 posters saying a human wouldn't have been able to avoid the accident reacting to the dash cam footage.

My first point is the dash cam footage is not representative of what a human can see in the light levels shown in the video.

My second point is a human would have been able to see the pedestrian long before they were visible in the footage released to the media and potentially reacted to it to reduce the severity or completely avoid the incident in the first place.

I was not commenting at all on the AV system and what it saw, could have seen or should have seen that clearly failed to 'see' anything.

The 'safety' driver was not concentrating on the road or looking at it most of the time and spend most of the time looking at either the in car telemetry or something else equally as detracting.
Click to expand...

Without knowing the sensor, ISO, aperture, exposure compensation and being there, it’s a guess how much a driver could’ve seen.

A guess.
 
You must log in or register to reply here.
Back
Top Bottom