Driverless Uber car kills pedestrian

[Caernarfon_Claret]

It's impossible to have a 100% perfect system. But computers in driverless cars would bring us far nearer to that than human drivers ever would.

In which case you have a balanced view.

A lot of the younger generation seem to think computers are perfect.

[Imploding Turtle]

In which case you have a balanced view.

A lot of the younger generation seem to think computers are perfect.

I'd love to know what you're basing that opinion on.

[thatdberight]

I'm sure computers are more reliable than computers...

I'm sure they're not.

[Caernarfon_Claret]

I'm sure they're not.

There's a glich in my system.

[Caernarfon_Claret]

I'd love to know what you're basing that opinion on.

Just anecdotal evidence, nothing concrete so it's only enough for me to form my own opinion, I'm not expecting others to share that opinion nor am I expecting others to be able to provide anything imperical to support either view as I doubt such information exists.

[dsr]

Just anecdotal evidence, nothing concrete so it's only enough for me to form my own opinion, I'm not expecting others to share that opinion nor am I expecting others to be able to provide anything imperical to support either view as I doubt such information exists.

If you want more anecdotal evidence to back up what you already have, I know quite a few young people who see no problem at all with a cashless society and can't imagine anything that might go wrong.

[Imploding Turtle]

Police release the dashcam footage on Twitter. It doesn't show the impact but it shows up to the split second before the impact and i think anyone who watches it will agree that even a human driver wouldn't have been able to prevent the accident.

https://twitter.com/TempePolice/status/ ... technology" onclick="window.open(this.href);return false;

[ClaretDiver]

Good grief just watched it.....wow....

[Sidney1st]

A human was never going to stop in time.

I would assume a car with cameras/sensors would BUT it's still under development, so the human in the car should've been watching the road more closely.

[aggi]

That's pretty unpleasant.

It doesn't really explain anything though, the dashcam footage is pretty poor which enhances the "came out of nowhere" narrative but the car itself should have had LIDAR which builds up an image like this regardless of light or dark:



If it was a human driver you'd expect their view would have been better than the low contrast dashcam footage or, if they couldn't see ahead of them, to be driving at an appropriate speed (I know that one is a bit unrealistic).

[UpTheBeehole]

Police release the dashcam footage on Twitter. It doesn't show the impact but it shows up to the split second before the impact and i think anyone who watches it will agree that even a human driver wouldn't have been able to prevent the accident.

https://twitter.com/TempePolice/status/ ... technology" onclick="window.open(this.href);return false;

They would if they weren't looking at the screen, like this driver was.

[Sidney1st]

They would if they weren't looking at the screen, like this driver was.

40 plus mph, pedestrian crossing the road in a very dark spot, wearing no reflective clothing and outside of a designated crossing zone...

Sorry but there are very few people who would've avoided that, paying attention or not.

[BennyD]

Pilots only carry out landings manually for the practice. Most commercial airliners can land themselves at most major airports.

Cr4p. Whilst most commercial aircraft have an auto land system, it is only used under certain specific conditions as the ground environment has to be closely controlled to prevent corruption/ interruption of the cat 3 ILS signal. If these conditions aren’t active the aircraft is always manually landed by the pilots. As for pilots not doing anything, again that’s cr4p. Whilst the aircraft is almost always flown by the autopilot the pilots are always trying to get ‘directs’, or short cuts, in order to reduce fuel burn and flight time. Also, a pilot always has to be aware which is the nearest airfield, and the prevalent weather conditions there, in case of an in flight emergency. These airfields change as the flight progresses. Google AirTransat flight 236 and ask yourself if a computer would have achieved a similar outcome (from the point of engine failure). If it was an easy job, it wouldn’t cost over £100k to get a licence and pilots wouldn’t have to be examined more than brain surgeons in order to keep their jobs.

[UpTheBeehole]

40 plus mph, pedestrian crossing the road in a very dark spot, wearing no reflective clothing and outside of a designated crossing zone...

Sorry but there are very few people who would've avoided that, paying attention or not.

The human eye is much more light sensitive than a camera lens. What is a dark spot for a camera isn't necessarily a dark spot for the human eye.

[Sidney1st]

The human eye is much more light sensitive than a camera lens. What is a dark spot for a camera isn't necessarily a dark spot for the human eye.

Even so, at 40 mph plus, in the dark and outside of a crossing zone, there's absolutely no way you can say with certainty that you'd avoid hitting her.

The average driver doesn't pay enough attention to the road to enable them to act that quickly, especially if they know where the crossing zones are etc.

[UpTheBeehole]

Even so, at 40 mph plus, in the dark and outside of a crossing zone, there's absolutely no way you can say with certainty that you'd avoid hitting her.

The average driver doesn't pay enough attention to the road to enable them to act that quickly, especially if they know where the crossing zones are etc.

It was a 35 zone wasn't it? Why's it going at 40?

[BennyD]

Cr4p. Whilst most commercial aircraft have an auto land system, it is only used under certain specific conditions as the ground environment has to be closely controlled to prevent corruption/ interruption of the cat 3 ILS signal. If these conditions aren’t active the aircraft is always manually landed by the pilots. As for pilots not doing anything, again that’s cr4p. Whilst the aircraft is almost always flown by the autopilot the pilots are always trying to get ‘directs’, or short cuts, in order to reduce fuel burn and flight time. Also, a pilot always has to be aware which is the nearest airfield, and the prevalent weather conditions there, in case of an in flight emergency. These airfields change as the flight progresses. Google AirTransat flight 236 and ask yourself if a computer would have achieved a similar outcome (from the point of engine failure). If it was an easy job, it wouldn’t cost over £100k to get a licence and pilots wouldn’t have to be examined more than brain surgeons in order to keep their jobs.

Also, what would happen in the event of a medical emergency? Who would you inform to get the aircraft diverted if there are no pilots? Who would be responsible, and therefore sue-able, for on board deaths?

[Sidney1st]

It was a 35 zone wasn't it? Why's it going at 40?

I've seen reports it was a 45 zone.

Car was driving 38mph along Tempe road when it hit a pedestrian on Sunday
The speed limit was 45mph and the car made no attempt to slow down or stop

My comment of 40 mph plus was incorrect, but the car wasn't speeding.
It's an unfortunate accident with both sides at fault in this instance.

I've seen another better quality clip and even in that one there's still a high chance of a human driver hitting her.

[Jeffbfc]

Can you truly trust technology or human nature.
In all your years of driving, how many times have you had to brake suddenly at a zebra crossing due to someone walking up to a crossing and just walking out, no matter how close a car maybe, they go because its a crossing.
Not lets wait until the cars stop.
Other drivers when stopped, opening the door into traffic as your going past (complete tools).
Kids running out from behind cars.
It happens, all our reaction times are different, it would make sence that a computers reaction will only be as good as the information it receives.
Maybe all driverless cars should be fitted with the following in the drivers seat.

Capture.PNG (381.97 KiB) Viewed 1848 times

At least you might think that won't stop for me.

[starting_11]

Video of the incident.

I dont know why the radar or IR cameras didn't catch her but a human driver wouldn't have braked until after they hit her anyway.

http://www.dailymail.co.uk/news/article ... r-car.html" onclick="window.open(this.href);return false;

[Sidney1st]

Daily mail articles aren't allowed on this forum, it makes certain people froth at the mouth.

[starting_11]

Sorry. It was linked to me by someone else.

I swear!

[Geoff]

They would if they weren't looking at the screen, like this driver was.

Quoted from elsewhere...

"So doing some basic math here....

White center lines = 10' each

Empty space between lines = 30'

Distance traveled in FPS "Feet Per Second" at 35 MPH = 52.5'

Road gradient roughly 0%

Reaction distance = 34'

Breaking distance at 35 MPH - 62'

Stopping distance at 35 MPH - 100'

"When the pedestrians shoes first become visible in the video there is approximately 59' between the car and the pedestrian, in 1 second the car will have already covered 52.5' of that gap leaving 6.5' left to stop the car.

"In order for a human driver, or the driver in this car to have avoided this collision by merely hitting the brakes and traveling in a straight line, "as is the reaction when startled by something on the road" there would have needed to be at least another 127.5' of distance between the car and the pedestrian.

"For all the posts and articles that I have seen bashing the driver and Uber because this could have been avoided, it really couldn't have, the laws of physics would not have even allowed this to have been avoided in the best possible scenario.

"And to touch on the subject of their being street lights there, in many scenarios those street light reflections on your windshield can actually be quite obstructing depending on the glare, anyone living in an area with many street lights can probably attest to that as well.

"Lesson of the day - Don't jaywalk in the dark."

[UpTheBeehole]

Quoted from elsewhere...

"So doing some basic math here....

White center lines = 10' each

Empty space between lines = 30'

Distance traveled in FPS "Feet Per Second" at 35 MPH = 52.5'

Road gradient roughly 0%

Reaction distance = 34'

Breaking distance at 35 MPH - 62'

Stopping distance at 35 MPH - 100'

"When the pedestrians shoes first become visible in the video there is approximately 59' between the car and the pedestrian, in 1 second the car will have already covered 52.5' of that gap leaving 6.5' left to stop the car.

"In order for a human driver, or the driver in this car to have avoided this collision by merely hitting the brakes and traveling in a straight line, "as is the reaction when startled by something on the road" there would have needed to be at least another 127.5' of distance between the car and the pedestrian.

"For all the posts and articles that I have seen bashing the driver and Uber because this could have been avoided, it really couldn't have, the laws of physics would not have even allowed this to have been avoided in the best possible scenario.

"And to touch on the subject of their being street lights there, in many scenarios those street light reflections on your windshield can actually be quite obstructing depending on the glare, anyone living in an area with many street lights can probably attest to that as well.

"Lesson of the day - Don't jaywalk in the dark."

That's all well and good, but as I've said, the human eye reacts better in the dark than a cheap dashcam lens.

[Sidney1st]

Not always it doesn't as proven by numerous accidents over the years.

Even those with 20/20 vision could hit that woman dependant on their reaction times.
As pointed out, there wasn't much time to react due to various circumstances.

I also suspect it isn't going to be a cheap camera on an automated vehicle..

[UpTheBeehole]

What I'm saying is the human eye would have seen her well before she comes into view on the camera, because the eye is so superior in dark light.

Therefore the reaction time is much longer, and that woman might be alive.

[Sidney1st]

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2564438/" onclick="window.open(this.href);return false;

That's worth a read, apparently more fatalities at night than during the day.

Yes the human eye should in theory be better than the camera but you're assuming a driver has had their eyes checked, their window is clean, the road is illuminated properly for starters.

You can dismiss it again if you like, it's entirely your choice, but it isn't just a case of someone's eyesight.

The woman stepped out on to a road, in a dark spot whilst pushing a bike and wearing what appear to be dark clothes and outside of a designated crossing area.
The woman? in the car isn't watching the road but as Geoff has shown, it would take exceptional driving ability/reactions to miss that woman.

[UpTheBeehole]

She didn't just step out in the road, she was 4/5ths of the way across the road until a silent vehicle smashed into her at 38mph.

She's wearing light blue jeans and pushing a bright pink bike. She's crossing the road at a section which is lit. You can see that from the video. She's quite visible.

Crash.png (126.69 KiB) Viewed 1810 times

[Spijed]

The problems with humans is that they can't apply the same accuracy. A driverless can can be programmed to drive at 70 mph on a motorway. It's virtually impossible for a human to keep their foot in the accelerator as precisely. That's why they will never be as good. Same with the human eye. Even though the eye has millions more sensors than a camera it relies on which way the person is looking and without distraction. An automated machine is devoid of any emotion and therefore won't make the same mistakes.

Yes, machines do go wrong, and there will be fatalities but they will be far less than a driver at the wheel.

[UpTheBeehole]

The problems with humans is that they can't apply the same accuracy. A driverless can can be programmed to drive at 70 mph on a motorway. It's virtually impossible for a human to keep their foot in the accelerator as precisely. That's why they will never be as good. Same with the human eye. Even though the eye has millions more sensors than a camera it relies on which way the person is looking and without distraction. An automated machine is devoid of any emotion and therefore won't make the same mistakes.

Yes, machines do go wrong, and there will be fatalities but they will be far less than a driver at the wheel.

A car with a driver can be programmed to drive at 70; it's called cruise control.

The human is in charge of the brakes, and the human eye is superior to a lens/sensor in dark light, so a human is better placed in the dark.

[aggi]

The disengagement reports (i.e. when the driver has to take control from the car) for California make interesting reading. Google is way ahead of the others in terms of miles travelled and miles per disengagement. They drove over 350,000 miles and needed to take over 63 times which seems pretty impressive. Obviously the issue is how low they can get that figure before it is acceptable to use on the roads without someone overseeing the driving

https://www.dmv.ca.gov/portal/wcm/conne ... OD=AJPERES" onclick="window.open(this.href);return false;

[aggi]

A car with a driver can be programmed to drive at 70; it's called cruise control.

The human is in charge of the brakes, and the human eye is superior to a lens/sensor in dark light, so a human is better placed in the dark.

The human eye may be superior to a lens/sensor (although that's not really true when you consider night vision systems) but it definitely isn't superior to LIDAR/RADAR in the dark.

[Sidney1st]

She didn't just step out in the road, she was 4/5ths of the way across the road until a silent vehicle smashed into her at 38mph.

She's wearing light blue jeans and pushing a bright pink bike. She's crossing the road at a section which is lit. You can see that from the video. She's quite visible.

Crash.png

In a still frame yes, at 38mph not so much and that's the part you're ignoring.

38mph, poorly lit road, outside of a crossing section.
Said pedestrian could clearly see the cars headlights before the car saw her yet she made the choice to cross anyway.

[UpTheBeehole]

At 38mph, from that distance, you'd be expected to stop. That's a good couple of car lengths.

[Sidney1st]

It's been a long time since I did an emergency stop.

You've then got to factor in things like human reaction time.

We don't know what the system on the car was like in regards to capability either.

Like I've said, both parties were at fault and it's resulted in a death unfortunately.

[starting_11]

Why are you arguing with this joker Sid?

He's clearly trolling.

Either that or he'd be superman. Since he can see in the dark and has reaction times quicker than the speed of light.

[starting_11]

She didn't just step out in the road, she was 4/5ths of the way across the road until a silent vehicle smashed into her at 38mph.

She's wearing light blue jeans and pushing a bright pink bike. She's crossing the road at a section which is lit. You can see that from the video. She's quite visible.

Crash.png

at 7 seconds you can see the soles of her shoe...which you wouldn't be able to work out if what it was anyway.

By 8 seconds, she's been hit.

I know you're an argumentative luddite but you're wrong on this one.

[Sidney1st]

Interestingly I'm not sure it was a silent car.
It was potentially a Hybrid, but at 38 mph I don't know if would be on electric or normal engine mode.

[Imploding Turtle]

At 38mph, from that distance, you'd be expected to stop. That's a good couple of car lengths.

33m is the stopping distance for 38mph

http://www.brakingdistances.com/38Mph" onclick="window.open(this.href);return false;

Even with a fast reaction time it's 31 metres.

But i suppose that's fine for a 15.5 metre long car

[aggi]

As an illustration on how poor quality the dashcam is when saying they couldn't be seen, this is the image from the same place on a better dash cam



It's a pretty brightly lit road, see the big lights up above.

[Rick_Muller]

As an illustration on how poor quality the dashcam is when saying they couldn't be seen, this is the image from the same place on a better dash cam



It's a pretty brightly lit road, see the big lights up above.

That illustration is pretty damning in my opinion. I think the human element in the car is in error, especially considering that their eyes should have been on the road ahead, not looking down at a screen.

[Sidney1st]

The human in the car is at fault then based on those images and should be prosecuted as such.
The car is still under development.

[Hipper]

Surely this sort of thing is exactly what you would hope a driverless car would be capable of dealing with.

I don't know what the person in the car's instructions were but clearly they had to monitor something.

I would have thought a good driver concentrating as he should, would have spotted this woman crossing the road in those circumstances (based on the second picture above). As this appears to be a dual carriageway, depending on what was around him, successful avoidance was surely possible.

Ultimately the woman crossing the road is to blame for her demise.

[dsr]

If that's a fair view of what the car could see, it looks like it wouldn't get a driving licence because of poor eyesight. The woman may have been foolish crossing a dual carriageway, but the car had plenty of time to see her if its "eyesight" had been better.

[Sidney1st]

I don't think the car would use that camera to 'drive' with.
It should be using sensors to pick up what's ahead of it on the road.

Lane Assists uses 'cameras' in a sense but they're designed to pick up road markings to keep you in the lane.

To spot objects on the roads it needs different sensors.

[UpTheBeehole]

As an illustration on how poor quality the dashcam is when saying they couldn't be seen, this is the image from the same place on a better dash cam



It's a pretty brightly lit road, see the big lights up above.

This is exactly what I was saying.

So many people pinning their decisions on what you can see from a crap dashcam instead of taking a step back and thinking about things in real life.

[Paul Waine]

What I'm saying is the human eye would have seen her well before she comes into view on the camera, because the eye is so superior in dark light.

Therefore the reaction time is much longer, and that woman might be alive.

Agree. If you've had the pleasure of doing the speed awareness course you will be familiar with the acronym COAST.

Concentration, Observation, Anticipation, Speed, Time

OK. The computer should always be concentrating, but to pass your test you need to be looking around, not just ahead. Was the computer always observing? Did the computer know that it was dark and visibility was low? Was it equipped with "visual aids" that can see whatever the light conditions - and is looking out for the unexpected? If the computer saw the pedestrian with her bike, or any shape/object, did it anticipate that the object might be heading towards a point where there might be a collision? Was the speed appropriate for the conditions? Would the computer have been driving at 45 mph if it was clear daylight? Was the slightly lower speed an adjustment to the conditions? Did the computer slow down when it should have been possible to see an object at the side of the road? Or did it not see the woman and bike? If the computer had slowed down it would have given itself the time to avoid the collision.

Why think that the pedestrian is partly responsible? It was dark. Yes, she could see the headlights, but how easy is it to judge speed and distance when it's dark? In some countries the motorist is always at fault if they collide with a pedestrian or a cyclist. Shouldn't we be aiming to build driverless cars where this is the minimum standard? What if it had been a visually impaired person? or a child? or a drunk? How about dogs and cats?

Remember the Tesla the ran into the side of the truck because it was the same colour as the sky?

How many driverless vehicle manufacturers are testing their computer programs on the roads? Should there be a requirement to prove that they operate safely rather than testing them with "live fire" where the public can be harmed?

[Sidney1st]

This is exactly what I was saying.

So many people pinning their decisions on what you can see from a crap dashcam instead of taking a step back and thinking about things in real life.

In real life the woman crossed the road in the wrong place and got run over by a car that wasn't being monitored properly.

[Sidney1st]

Agree. If you've had the pleasure of doing the speed awareness course you will be familiar with the acronym COAST.

Concentration, Observation, Anticipation, Speed, Time

OK. The computer should always be concentrating, but to pass your test you need to be looking around, not just ahead. Was the computer always observing? Did the computer know that it was dark and visibility was low? Was it equipped with "visual aids" that can see whatever the light conditions - and is looking out for the unexpected? If the computer saw the pedestrian with her bike, or any shape/object, did it anticipate that the object might be heading towards a point where there might be a collision? Was the speed appropriate for the conditions? Would the computer have been driving at 45 mph if it was clear daylight? Was the slightly lower speed an adjustment to the conditions? Did the computer slow down when it should have been possible to see an object at the side of the road? Or did it not see the woman and bike? If the computer had slowed down it would have given itself the time to avoid the collision.

Why think that the pedestrian is partly responsible? It was dark. Yes, she could see the headlights, but how easy is it to judge speed and distance when it's dark? In some countries the motorist is always at fault if they collide with a pedestrian or a cyclist. Shouldn't we be aiming to build driverless cars where this is the minimum standard? What if it had been a visually impaired person? or a child? or a drunk? How about dogs and cats?

Remember the Tesla the ran into the side of the truck because it was the same colour as the sky?

How many driverless vehicle manufacturers are testing their computer programs on the roads? Should there be a requirement to prove that they operate safely rather than testing them with "live fire" where the public can be harmed?

Several companies are testing, some have been in the UK.
Some one else provided a stat about the number of times Google cars needed driver intervention and it was incredibly low.

In this instance it's a combination of pedestrian error coupled with useless human in the car not paying attention herself so she could intervene.

They'll get the hang of it though and in the next 10years I suspect autonomous cars will become a regular thing on the road.

[UpTheBeehole]

In real life the woman crossed the road in the wrong place and got run over by a car that wasn't being monitored properly.

In real life a woman is dead because of unsafe testing