Uber Autonomous Car Death

[Jahulaye]

Would you have seen the cyclist and been able to brake in time?

http://www.dailymail.co.uk/news/article-5529453/Video-shows-moment-pedestrian-killed-self-driving-Uber-car.html

[ToraToraTora]

It seems the car made no attempt to brake, could it possibly be that the car did not see the pedestrian?

[Jahulaye]

The pedestrian came out of nowhere from the dark on a busy highway but the car should have had sensors and an infrared camera.

I’ll be interest in the findings to see whether the car carried on after the impact and was stopped by the driver or was the emergency brakes deployed by the car after the impact.

[naomi24]

If I hadn't managed to avoid hitting the pedestrian, I'm pretty sure I would have had some time to brake. I wouldn't have hit him at the speed of the driverless car. That's frightening!

[vulcan42]

The question has to be asked, what is the pedestrian doing in that position, she must have seen the car coming?

[Quoi]

There is a serious question about how to keep a supervising driver 'alert' enough. Not just the one in the video but for future supervisors.

[Zacs-Master]

'Would you have seen the cyclist and been able to brake in time?' who knows.

Looks like a CIA plot to make us all use public transport and buy more LED lights, to me. Eh, Jules?

[ludwig]

The car should have been able to react faster than a human in that situation, so it was obviously a major failure of the system.

There was clearly nothing the supervising passenger could have done. In any case, if you have to stay completely alert and ready to hit the brakes/steer at a split second's notice, there isn't much point the vehicle being autonomous - you might as well be driving the flipping thing in the first place.

[Hymie]

Although the video image reveals the pedestrian at the last minute (second) – such images may not be representative of what the human eye would be able to discern under the given lighting conditions.

Many colour video cameras are unable to capture detail where the light level is less than 0.1 lux – so a human driver may have seen the pedestrian much earlier than the video suggests.

[jno]

On the basis of that film, no.

[Malachite]

Probably not, bless her. I wouldn't put blame on anyone - just a sad tragedy.

[bhg481]

The car should have seen the pedestrian. Poor lighting is no excuse as the detectors will use wavelengths outside the human range, as well as radar. Having said that, even had the driver been fully alert, which she didn't seem to be, she would not have been able to avoid the accident.

[fiction-factory]

>In any case, if you have to stay completely alert and ready to hit the brakes/steer at a split second's notice, there isn't much point the vehicle being autonomous - you might as well be driving the flipping thing in the first place.
These are just trials so a driver is there. Depending on the success of the trials I think it is hoped that the cars will be self driving. Anyway, even if there is always a need for someone to be there the idea is that the process will be safer (proper stopping distances, less reliance on banned/drunk drivers, cars not distracted by road rage, phones , attractive pair of legs outside...) more efficient (less stop start)

[ToraToraTora]

Hear Hear Ludders

Hymie: “Although the video image reveals the pedestrian at the last minute (second) – such images may not be representative of what the human eye would be able to discern under the given lighting conditions.”
We see the woman when the headlights pick her out and most drivers would be able to avoid her. The sensors should have picked her up much earlier not being reliant on visible light. So either the sensor did not work or the software did not take account of the information. Both worrying, pretty pointless if you have to have a human overseer in these cars.

[fiction-factory]

>Both worrying, pretty pointless if you have to have a human overseer in these cars.

Again I think the point has been overlooked that these are trials. There is a human there now and will need to be until the technology is deemed to be ready. I can see a time when all cars on certain roads, maybe all roads, are autonomous - but it's a long way off. We already have some driverless trains and fork lift trucks, for example, are autonomous (unmanned)in this sense

[Hymie]

After viewing the video again – I would estimate that the time between the pedestrian first being visible and the impact being just over 1 second.

I doubt very much that a human driver would have been able to avoid this collision. By the time a driver had removed their foot from accelerator to brake and applied pressure – minimal vehicle braking would have occurred by the time of impact.

The only other option would be to steer around the hazard; at the speed the vehicle appears to be travelling, I think it highly likely that turning the steering wheel sharply to the left (sufficient to avoid the pedestrian) would result in the vehicle rolling – and still taking out the pedestrian.

[Spicerack]

I wonder if the driver was able to disable/mask certain sensors they considered annoying/superfluous?

[bhg481]

If Hymie's time estimate is right and the car was travelling at 40mph, or about 60ft/sec, that gives a car-to-pedestrian distance of 60 feet. The Highway Code claims that at 40mph the thinking distance is 40ft and the braking distance a further 60 feet; this means that by the time the driver has reacted there would only be 20 feet between the car and the pedestrian - not enough even to swerve, I think. There are those who would argue that the Highway Code stopping distances are out of date and that modern cars can stop in a shorter distance than those given but an article in the Independent (July 2017) claims that they are actually too short. The conclusion is that a human driver would have killed the pedestrian.

[jno]

That was also my conclusion, bhg, not nearly enough time to stop.

You might well argue the car’s systems failed, but I don’t think you can conclude that it performed worse than a car with a driver (though to be sure, some drivers on AB assess themselves as a lot smarter than me).

[bhg481]

I've just had a further look at the article in the Independent and realised it's the thinking distance that's increased, not the distance to come to a standstill after the brakes are applied. They now say that the average thinking time is 1.5 seconds (previousy 0.67 seconds) which means that in this particular case the average motorist would not even have had time to react.

[mibn2cweus]

The cyclist was most likely 'seen' by the ADS. The problem was likely with the autonomous system's inability to properly interpret and assess the hazardous nature of what it was 'seeing'.

https://www.bloomberg.com/news/articles/2018-03-21/for-self-driving-cars-seeing-everything-isn-t-always-enough

My assessment is that such systems are still a ways off from being a replacement for responsible human drivers as much as they do apparently serve a valuable role in augmenting and compensating to some degree for human error.