Forum menu
<div class="bbp-reply-author">
<div class="bbp-author-role">
<div class="">Member</div>
</div>
</div>
<div class="bbp-reply-content">Yeah, that is a really tough one to avoid. I don’t think any of us could have avoided that.
</div>
Maybe not avoided, but I reckon most would have got the brakes on at least and reduced the impact, **** me that was absolutely crap performance from the car. She should have been plainly visible to the LIDAR system and the RADAR system (assuming Uber have those like google do?)
It also looked like the human was looking down at their phone or something immediately prior as well.
I've been quite vocal in defending autonomous cars in general, but that really was a shocker. Uber's auto cars at least should not be back out on the road until the issue is has shown to have been reproduced and fixed.
And yet there was a human driver in that car.
It certainly shows the degree of alertness you can expect from a driver who has been ferried around town with no input required for a few hours. Certainly wouldn't be banking on it as a failsafe.
And yet there was a human driver in that car
The video inside the car shows that he wasnt paying attention. Which is what I would expect.
People have difficulty concentrating on monotonous tasks when they arent actually actively engaged in it. Watching a car drive I suspect would quickly fall into that category. I know my attention levels as a passenger rapidly vanish.
It would be interesting to get a driver instructors view of it. I am guessing their concentration on having their foot near the brake pedal decreases as the trainee gets more experienced.
From Volvo website:
"Collision avoidance systems are increasingly popular with motorists that spend a lot of time behind the wheel in stop-and-go commuting traffic where the risk for low speed collisions is quite high. Volvo Cars introduced City Safety as standard in new models from 2008. The first generation of the technology worked at speeds up to 18 mph. This was subsequently increased to 31 mph from 2013. In 2015, City Safety has been updated in the XC90 and now operates at all speeds*.
*Applicable only to new XC90s. In all other current Volvo models City Safety operates at up to 31 mph."
The second video is interesting, visibility looks good enough to me for an active human driver to have at least taken some kind of mitigating action.
Of course the car doesn't 'see' as we do, so I can't help thinking additional sensors are required, a thermal camera /sensor may well have provided the car with enough additional data to recognise the hazard?
If the tech is no better than the human what is the point of it?
The point of it is it will be much better than a human. It isn't there yet but is it evolving very quickly whereas the human most certainly isn't (and is arguably getting worse)
It certainly shows the degree of alertness you can expect from a driver who has been ferried around town with no input required for a few hours. Certainly wouldn’t be banking on it as a failsafe.
I agree. I can't find any real specific info on the car, it appears to be only level 2, so similar to a what Tesla and Audi currently offer for sale, if that had been any other state the driver would be 100% to blame but the laws in Arizona seem very lax; so much so that Uber apparently have a fleet of autonomous lorries on the state's roads and another taxi app company, Waymo, have plans to introduce a fully autonomous taxi company there this year.
It seems like the real drive (sorry) here is the rush to be first to market and things are progressing too quickly.
A lot of money at stake for whatever company achieves this.
I'm absolutely sure the tech can and will get there, but it's clear to me that it's not there yet.
A lot of money at stake for whatever company achieves this.
Yeah you have to imagine that the company who develops the first reliable level 5 car will be able to license their tech to many other car manufacturers. The market for delivery / goods vehicles alone must be massive.
a thermal camera /sensor may well have provided the car with enough additional data to recognise the hazard?
I believe Uber uses Lidar so the camera, in terms of what the car would see, would be irrelevant. Tesla go for cameras only but most of the other biggies go for a mix of cameras and Lidar.
Tesla go for cameras only but most of the other biggies go for a mix of cameras and Lidar.
Tesla use a different kind of sensor in addition to the cameras, ultrasonic maybe.
edit: yes ultrasonic https://www.tesla.com/en_GB/autopilot
Human driver could have avoided or mitigated that impact. Probably wouldn’t have, but definitely could have. But then, lidar equipped autonomous system could have too, so I’m back to square one in my opinion, really. Computer can make mistakes too, shocker. It was probably checking it’s twitter feed or randomly rebooting/updating for no apparent reason or something...
computers are a very long way from infallible. This appears to have been a case of ‘unexpected bag lady in the driving area...’ :-/
edit: yes ultrasonic
Thats for very close, so slow speed, stuff rather than general self driving. That said they do have limited radar as well. Didnt realise they had that. They are a lot more camera focused than the others.
The more I look at it, in the context of the local's video taken with a Pixel XL phone, the more it seems there was ample chance for vehicle camera, radar and IR to pick up the presence of the woman and bike, and plenty of time to at least reduce speed if not completely avoid a collision. Something was not working, turned off, or ignored in the decision making process. That car looked like it would have driven into practically anything in the middle of that road, a stranded car, dropped load from a truck, bollards and barriers for construction, cycle and rider side on and moving slowly, it just didn't react at all. Police also said there was no braking at all.
I suspect someone is going to be for the high-jump based on that video. The guy in the car supposedly monitoring the car was spending nearly all his time looking at something on the dashboard (is there a video screen or something there?) and was not looking at the road for significant periods. That the car didn't make any attempt to brake is going to be a big problem because that pedestrian plus bike - side on - must be a bigger target than a bike and rider would be if the car had been coming up behind them. So there will be a lot of explaining to do.
Personally I think it's entirely possible that a driver paying attention would have at least started to slow down before the collision - although as others have mentioned that's on the assumption that the camera view isn't anything like as good as a human eye would be for anything outside the area lit by the car lights.
For me I'd have said that accident is bad enough that further tests using those vehicles need to cancelled until further development is done and at least a clear explanation of why the car didn't "see" the person it killed.
The video is damning
https://twitter.com/davidshepardson/status/976587694066323456
My foot was, on the brake well before impact, try it. And that was with a tiny vido screen. In real life even the dipped headlights would have made the cyclist visible from a long way off.
Edukator
The video is damning
My foot was, on the brake well before impact, try it. And that was with a tiny vido screen.
Well you definitely win today's award for Best Driver of a Video Clip.
I'd also like to know what happens post video, does the car continue down the road as if nothing has happened? Did the operator have to stop it? There seems to be a bit of 'hey look you can hardly see her' spin in the media, while to me there's questions as to why the array of sensors and on board processing wasn't enough for the vehicle to react in any way.
There are pictures at the scene where you can see her crumpled body lying by the side of the road right beside the car. They'll ruin your day, very tragic. Obviously the driver reacted eventually. Apparently there are fully autonomous cars driving around Arizona with no human occupants at all, which makes me wonder if the tech on this car is such that it requires human supervision. I still can't find any info on that though.
dissonanceThats for very close, so slow speed, stuff rather than general self driving. That said they do have limited radar as well. Didnt realise they had that. They are a lot more camera focused than the others.
D'oh, you're totally right, bit of a brain fart on my part there. It was the radar one I meant, don't know why I had the ultrasonic thing in my head.
They bounce the radar signal under the car in front to detect cars stopping/slowing ahead. Clever stuff.
Just watched the video and I'm stunned.
I'd imagined the victim lunging out from beside the road but she was actually crossing at slow speed from the opposite lane.
How could the vehicle possibly be considered to be traveling at an appropriate speed if it couldn't see/detect her and take any form of action?
stunned?
it happens all the time. Person in dark clothes, in a poorly lit area, walks out into the middle of a main road without looking, or misjudges the speed of the traffic on that road.
The difference is that with a normal human driver the "SMIDSY" is now old News so no one cares about it (except the victims family of course....)
How could the vehicle possibly be considered to be traveling at an appropriate speed if it couldn’t see/detect her and take any form of action?
Logs will be downloaded conditions assessed etc. This does not happen after a typical human controlled accident/incedent The area according to laws etc should have been free of pedestrians.
How fast would you drive on an 8 lane road with no crossings?
I like how someone says, after watching a video where the events were known "i'd have managed to avoid her". Sorry, but, no you wouldn't
I did an interesting driver evaluation exercise for Warwick University about 15 years ago. We hired the Millbrook Proving ground, and set up some air cannons in small enclosures are the side of the track. Those air cannons would fire a small child sized dummy into the path of the oncoming test vehicle, but only after a randomised amount of time, and at a random location on the approx 3km loop.
The idea was to test drivers reactions when drunk and distracted, but our "control group" actually showed an interesting anomaly. After around 45min of driving round the loop, the drivers stopped concentrating, and even though initially they would slow down and cover the brake as they approached the potential "kid in the road" locations, after a while of not finding a kid in the road, even in a test where they KNEW it was CERTAIN that they would find one in the road eventually, they started to ignore the risk, and pretty much 95% of them eventually "ran over" the kid.
This regression to the norm is how we drive. We drive hundreds of thousands of miles without a kid running into the road, so that is the "Normal" state. When unfortunately for some unlucky driver they do eventually and unexpectedly run out, i'd say that 99% or more of drivers would be unable to react fast or well enough to avoid such an impact.
FFS Maxtorque and Jimjam, if you don't think you'd have done better than the Uber car hand in your driving licences now. Check the second counter for the time the bike is in view in the video and consider that a human with eyes adadpted to the night would have seen the bike much earlier. If you hadn't been on the brakes for enough time to save that woman's life you shouldn't be on the road, and if you must drive with such slow reaction times the 2 second minimum distance on the motorway should be at least five for you to compensate for your handicap.
If the supervising driver hadn't been on board, do you think any of us would ever have heard about this?
I like how someone says, after watching a video where the events were known “i’d have managed to avoid her”. Sorry, but, no you wouldn’t
Sorry Maxtorque. I’m often impressed by your posts, as they often seem to demonstrate a considered opinion from an impressive base of knowledge. But I’m going to have to at least partially disagree with you here. Even on the fairly poor dash cam footage, the pedestrian appears from the inky blackness of beyond the dipped beam a whole one and a half seconds before impact. A human driver who is actually looking at the road would definitely react in that time, and at the quoted 38mph, the reaction would likely have mitigated, or even avoided a collision. The Uber car SHOULD have been able to see the ped far outside the range of the lights, as it’s sensors are not reliant on the visible light spectrum. It completely failed to react, which is either a sensor or a logic failure, and is a massive fail on Uber’s behalf. This is exactly the sort of situation that autonomous cars are meant to best humans. Incidentally, the road in question actually has a good level of ambient light, and there is footage available that would suggest that the human eye would have been able to see considerably further than the dipped beam range too, had he (she? Can’t decide) actually been looking forward out of the windscreen, rather than down at something (phone?).
Maxtorque is the biggest motorsport ' making progress" specialist-language-baffles-brains fantasist on this forum and this thread proves it.
Thank you for confirming my thoughts on the 1.5 seconds V8ninety, my Casio made it a fraction more but with half a second to get your foot on the brake that's still a second of braking time, which is less than the time a human driver with night-adapted human eyes would have had.
The car ****s up, simple as.
Edukator I don't think anyone is saying that something didn't go wrong here but people are disputing that humans would on average have done better, you are relying on the driver concentrating on the right thing at the start of the 1.5s. How many people are maintaining that level of concentration? The stats tell us not enough.
The car can be improved, the sensorsed improved, the decision making and assesment refined. The meat sack can't be. If somebody had not been paying attention and hit the person this would probablynot have even made the local evening news, that is sad but it's how normalised people have become to road deaths.
Funnily enough, ive actually measured typical "reaction times" for drivers.
I'd have to go find the reports i wrote, but iirc, typically, for an 'observant but unprepared driver' ie one who is not expecting to have to react, but is at least observing their surroundings it took about 200ms to realise they needed to react, around 400ms to get their foot off the accelerator and onto the brake, and well over 1 second to get to full threshold braking. ie a total of 1.6 seconds to get to -1g decel (which is -9.98 m/s/s)
Travelling at 40 mph (~18m/s) that means the vehicle has traveled 29m.
Seriously try it. Next time you have a long journey ask your front seat passenger to suddenly shout "STOP NOW" at some random point in the journey (they need to make sure it's safe to do so of course! ie no one behind you etc) and see how long it takes you to get from "completely unready to stop" to a full ABS stop. IME, you maybe shocked about how far you've traveled before you even start to brake.
PS. the two second gap to a car in front is for vehicles travelling at the same speed!
That two seconds is not there to allow you to stop, but simply to allow you enough time to react so you can start to stop without hitting the car ahead! (both cars will take time/distance to stop, but if you are too close, you will hit the braking car ahead before you can brake enough yourself)
If you are two seconds away from a stationary object and travelling at any more than roughly walking pace, you're going to hit that object!
Seriously try it.
I have, Maxtorque.
There's a mass of information on reaction times in the public domain. At 4 in the morning people are slower than in the middle of the day but the recognised average reaction time from brakes hard on is 1 second and your theshold braking is irelevent specialist language bollocks to ry and mask the fact the car ****ed up and you're still try to deny it.
You'l forgive me for doubting anything you claim to have done but you claimed rally experience on several threads but have failed to come up with any results list in the public domain.
The car failed miserably by human standards unless one is as lousy as Jimjam or Maxtorque who claim to be miserably below average drivers. I'm aware why the gap is 2seconds, so does everybody else who has aver passed a driving test, but you clearly need more given that you wouldn't have ben hard on the brakes in that Uber car, Maxtorque
The car didn't brake at all despite the fact the cyclist had made it 2/3 the way across a two lane road in good weather conditions, ambient light and dipped headlights.
Who is trying to say the car didn't **** up?
You are trying to say that human drivers are way better, stats don't back that up. You are relying on it being a good driver paying attention, some news here they don't they don't look, they dont scan for everything and they get easily distracted.
Who is trying to say the car didn’t **** up?
Maxtorque and Jimjam, read back.
Go on post the quote..
Edukator
FFS Maxtorque and Jimjam, if you don’t think you’d have done better than the Uber car hand in your driving licences now.
I didn't make any statement as to my ability one way or the other but (I think Maxtorque and I are on the same page here) I believe it's pointless to watch a video then re-watch and speculate as to how you might have reacted. I'm pretty skeptical that anyone would have reacted (as much as you can react to a video) the very first time they see it, unless you're watching with the specific intention of anticipating the crash which is not how we naturally watch videos. Hence my comment about you being the best video driver on the internet today.
Honestly, please stop making straw men or ascribing intentions, motives or opinions to people that they don't have. That's twice you've done it to me in a couple of pages, it's as if you think other people are incapable of following your posts and my posts and they'll just blindly follow your misrepresentation of my opinion.
Edukator
Maxtorque and Jimjam, read back.
Nope. You managed to do it again while I was posting.
OK, Jimjam. Make the following statement clearly:
" the UBer car completely ****ed up. In those condtions a very average human driver would have seen and had time to react to the crossing cyclist, and been braking hard at impact even if they hadn't managed to stop completely".
Sign that statement and I'l back off, till them you're just an aplogist for a dangerous driverless car.
Eh what? You still feeling under the weather there. The 2 parts of your statement are not exclusive that is the subtly of the situation you seem to be missing.
OK, Jimjam. Make the following statement clearly:
I would but here's the thing, it's not clear (or at least I can't find out) what level of autonomy the car is supposed to have, and what the legality of that tech is in Arizon where the crash happened. As far as I can tell the car is only level 2 AV, and thus the driver should have been ready at any point to intervene, such as with Tesla's current autopilot system.
If the car is level 2, then the driver should have been paying attention and is just as guilty as someone who rear ends another driver because they had cruise control on a dumb car and didn't react. If you can clarify what level of autonomy the car has, and by extension how alert the driver should have been then I can better agree with (or disagree with) your statement. Is that clear enough or did I muddle my explanation?
Sign that statement and I’l back off, till them you’re just an aplogist for a dangerous driverless car.
Is this the Autonomous vehicle thread equivolent of "I'm going to call you a Nazi? Where's my laugh emoji.
Feeling fine, I had a nice afternoon playing guitar in good company in a sunny park with a view over snowy mountians.
I'd like Jimjam and Maxtorque to admit any human could reasonably be expected to do better than the Uber car by signing my statement that I'l kepp posting til I get banned, they sign it or **** off off the thread as their petrol-head support for a murduring car isn't welcome on this forum where there are real people who ride bikes and might wish to use the road without getting mown down by an Uber car that doesn't react to their presence.
Go on Jimjam and Maxtorque sign. You know it's the truth.
” the UBer car completely **** up. In those condtions a very average human driver would have seen and had time to react to the crossing cyclist, and been braking hard at impact even if they hadn’t managed to stop completely”.
More emaly mouthed bollockes form Jimjam I see while I was posting - just admeit it Jimjam and sign.
Feeling fine, I had a nice afternoon playing guitar in good company in a sunny park with a view over snowy mountians.
I’d like Jimjam and Maxtorque to admit any human could reasonably be expected to do better than the Uber car by signing my statement that I’l kepp posting til I get banned, they sign it or * off off the thread as their petrol-head support for a murduring car isn’t welcome on this forum where there are real people who ride bikes and might wish to use the road without getting mown down by an Uber car that doesn’t react to their presence.
Go on Jimjam and Maxtorque sign. You know it’s the truth.
” the UBer car completely * up. In those condtions a very average human driver would have seen and had time to react to the crossing cyclist, and been braking hard at impact even if they hadn’t managed to stop completely”.
Read my post right above yours, have a think about it, then post back.
This might help too - https://www.carmagazine.co.uk/car-news/tech/autonomous-car-levels-different-driverless-technology-levels-explained/
Yes you muddled your explanation and refused to sign.
Anyone else in here apart from Jimajam and Maxtorque think this statement is unreasonable?
” the UBer car completely **** up. In those condtions a very average human driver would have seen and had time to react to the crossing cyclist, and been braking hard at impact even if they hadn’t managed to stop completely”.
I’d like Jimjam and Maxtorque to admit any human could reasonably be expected to do better than the Uber car by signing my statement that I’l kepp posting til I get banned, they sign it or **** off off the thread
OK then. Happy trolling perhaps you could learn something
Yes I do, your assessment of the very average driver is a bit off though
Wel somone needs to call out the petrol heads on this forum when they see a cyclist killed and claim the Uber car wasn't responsible.
I's not trolling.
Jimjam and Maxtorque have been advocating a drivng style that is illegal and inappropriate on this forum for years. The contest the validity of speed limits. They encourage driving fast and all this on roads they sharte with STW members on those same raods with their bikes. When there are collisions they are the first to blame the cyclsit (the case here by default) .
Maxtorque calimas experinec as a tech at Prodrive that I don't doubt but having spent years in motorsport I also know tha the most dangerous drivers in therally team were the service barge drivers and those given a car to "test" on public roads. Was it 230mph in Saudi, maxtorque and how much rally experience? (which I have yet to see any prooff of)
If you're going to set yourself up as an expert at least recognise that an average driver would have ben on th ebrakes when th eUber car failed. so sign this:
” the UBer car completely **** up. In those condtions a very average human driver would have seen and had time to react to the crossing cyclist, and been braking hard at impact even if they hadn’t managed to stop completely”.