Self-Driving Vehicles
03-21-18, 06:05 PM
#676
She didn't emerge out of shadows, its the lack of contrast in the video makes it seem this way. You can see many street lights in the video, so the street was very well lit. The cars speed wasnt very fast either, only 38mph. And the pedestrian didn't just dart out of nowhere like originally claimed - she was actually crossing slowly and there was plenty of time for an attentive human driver to see her and react appropriately.
03-21-18, 06:08 PM
#677
Lexus Champion
She didn't emerge out of shadows, its the lack of contrast in the video makes it seem this way. You can see many street lights in the video, so the street was very well lit. The cars speed wasnt very fast either, only 38mph. And the pedestrian didn't just dart out of nowhere like originally claimed - she was actually crossing slowly and there was plenty of time for an attentive human driver to see her and react appropriately.
This doesn't address my points about the fact that the car's LIDAR should have spotted this woman even if she was on the median in the bushes, which goes to your point. Between LIDAR and cameras the hardware/software should have figured out that this is a jaywalker intent on being in the middle of the road. There will be more information emerging because the two main agencies and the companies are looking into it.
03-21-18, 06:24 PM
#679
How do you know this? Have you viewed the original source footage that the police have seen? Or the data that the car's systems recorded? And do you expect a vehicle to stop just two seconds before a pedestrian steps in its path? 38 mph fyi is 55 ft/sec.
This doesn't address my points about the fact that the car's LIDAR should have spotted this woman even if she was on the median in the bushes, which goes to your point. Between LIDAR and cameras the hardware/software should have figured out that this is a jaywalker intent on being in the middle of the road. There will be more information emerging because the two main agencies and the companies are looking into it.
This doesn't address my points about the fact that the car's LIDAR should have spotted this woman even if she was on the median in the bushes, which goes to your point. Between LIDAR and cameras the hardware/software should have figured out that this is a jaywalker intent on being in the middle of the road. There will be more information emerging because the two main agencies and the companies are looking into it.
03-21-18, 06:41 PM
#681
Lexus Champion
You can see in the video that she was walking across the street very slowly, and the car hit her in the right lane. It took her several seconds to get from the median across the left lane and into the right lane - plenty for an attentive human driver to react. Judging from the video there was no apparent reaction from the car at all - no attempt to swerve or brake.
My bet is the other way. The car's LIDAR and radar should have spotted her at roughly 300 ft away if she placed herself in the median in the bushes and resolved that it was a pedestrian, especially with a bicycle which is metal. Once she started moving, the algorythim should have told the car's brain that there was an object in motion moving into the car's path.
The car is moving roughly 55 ft/sec and that means that it should have braked without human intervention a minimum of 120 ft away from her but even much farther away than that. As others say, the point is moot because radar/lidar doesn't care if its dark outside or the noonday sun.
Your point is that a human driver would have outperformed the onboard systems because that human would have better vision and detection capabilities. Humans are just as fallible because your average driver would not necessarily have reconciled what was going on in front of them any better than the Uber software/hardware.
03-21-18, 06:48 PM
#682
03-21-18, 06:53 PM
#683
So basically you are betting that a human driver in that same set of circumstances on that road with those visibility conditions would have or could have reacted faster than the tech in detecting this person and stopped or mitigated this collision.
My bet is the other way. The car's LIDAR and radar should have spotted her at roughly 300 ft away if she placed herself in the median in the bushes and resolved that it was a pedestrian, especially with a bicycle which is metal. Once she started moving, the algorythim should have told the car's brain that there was an object in motion moving into the car's path.
The car is moving roughly 55 ft/sec and that means that it should have braked without human intervention a minimum of 120 ft away from her but even much farther away than that. As others say, the point is moot because radar/lidar doesn't care if its dark outside or the noonday sun.
Your point is that a human driver would have outperformed the onboard systems because that human would have better vision and detection capabilities. Humans are just as fallible because your average driver would not necessarily have reconciled what was going on in front of them any better than the Uber software/hardware.
My bet is the other way. The car's LIDAR and radar should have spotted her at roughly 300 ft away if she placed herself in the median in the bushes and resolved that it was a pedestrian, especially with a bicycle which is metal. Once she started moving, the algorythim should have told the car's brain that there was an object in motion moving into the car's path.
The car is moving roughly 55 ft/sec and that means that it should have braked without human intervention a minimum of 120 ft away from her but even much farther away than that. As others say, the point is moot because radar/lidar doesn't care if its dark outside or the noonday sun.
Your point is that a human driver would have outperformed the onboard systems because that human would have better vision and detection capabilities. Humans are just as fallible because your average driver would not necessarily have reconciled what was going on in front of them any better than the Uber software/hardware.
I'm not saying that a human car always outperform onboard systems, but in this particular scenario there was plenty of time for an attentive human drive to react. Maybe onboard systems could outperform the human driver if they were working properly, but in this particular case they did absolutely nothing and encouraged the driver not to pay attention to the road.
03-21-18, 07:12 PM
#684
Lexus Champion
Well as a minimum, IMHO this poor unfortunate woman triggered this event by making a very bad decision. As most police collision reconstructionists would say, it's easier for a person who is walking to come to an abrupt stop nearly instantly vs a two ton vehicle that is hurtling towards them at those speeds. Physics doesn't really care about daytime or nightime and fault or no fault. I said that I wondered about her state of mind in crossing like that. She didn't appear to have detected the vehicle that was headed towards her. And that's what is unfortunate.
03-21-18, 07:20 PM
#685
Well as a minimum, IMHO this poor unfortunate woman triggered this event by making a very bad decision. As most police collision reconstructionists would say, it's easier for a person who is walking to come to an abrupt stop nearly instantly vs a two ton vehicle that is hurtling towards them at those speeds. Physics doesn't really care about daytime or nightime and fault or no fault. I said that I wondered about her state of mind in crossing like that. She didn't appear to have detected the vehicle that was headed towards her. And that's what is unfortunate.
03-21-18, 07:23 PM
#686
Driver School Candidate
Join Date: Mar 2018
Location: MA
Posts: 4
Likes: 0
Received 0 Likes
on
0 Posts
The Lidar/radar should have spotted her, but apparently it didn't. Or maybe it did, but the software didn't react. Or maybe it did, but the mechanical system that is responsible for applying brakes and steering did work properly. A lot of ifs, but the bottom line there was no reaction from the car whatsoever, and someone got killed as a result.
I'm not saying that a human car always outperform onboard systems, but in this particular scenario there was plenty of time for an attentive human drive to react. Maybe onboard systems could outperform the human driver if they were working properly, but in this particular case they did absolutely nothing and encouraged the driver not to pay attention to the road.
I'm not saying that a human car always outperform onboard systems, but in this particular scenario there was plenty of time for an attentive human drive to react. Maybe onboard systems could outperform the human driver if they were working properly, but in this particular case they did absolutely nothing and encouraged the driver not to pay attention to the road.
Video is out:
https://www.axios.com/uber-self-driv...85e12a145.html
It is clearly an uber software issue and safety driver was simply not attentive at the time.
This is where sensors should have detected the person on the street.
Uber is going to lose a lot of money over this.
Most safety systems out there today, with way lesser sensors would have attempted to brake at some point...
https://www.axios.com/uber-self-driv...85e12a145.html
It is clearly an uber software issue and safety driver was simply not attentive at the time.
This is where sensors should have detected the person on the street.
Uber is going to lose a lot of money over this.
Most safety systems out there today, with way lesser sensors would have attempted to brake at some point...
03-21-18, 07:37 PM
#687
Couldn't agree more. Slow moving pedestrian crossing straight road in front of relatively slow moving car, no obstructions, no inclement weather, no other traffic. At 38mph, even a 100ft range of lidar/radar sensing would have given almost 2s to react. The video quality is rubbish, but a human driver with good eyesight, properly adjusted headlights and paying attention IMO would have had a better than 50% chance to at least slow down the car significantly. I'm sure that the Uber engineers would have sworn up and down that this precise scenario would pose no problem for their car at all. Obviously, the Uber "safety drivers" are of little to no use.
03-21-18, 07:52 PM
#688
Well as a minimum, IMHO this poor unfortunate woman triggered this event by making a very bad decision. As most police collision reconstructionists would say, it's easier for a person who is walking to come to an abrupt stop nearly instantly vs a two ton vehicle that is hurtling towards them at those speeds. Physics doesn't really care about daytime or nightime and fault or no fault. I said that I wondered about her state of mind in crossing like that. She didn't appear to have detected the vehicle that was headed towards her. And that's what is unfortunate.
Obviously car should have stopped and it did not. It is a very clear, wide street.
There are many tests on every new vehicle with auto brake system done by IIHS, EuroNCAP and JNCAP, and at these speeds, system in your midrange vehicle would autobrake to full stop and not hit a woman.
In fact, those industry tests are done at much harder situations for systems because pedestrian is partially hidden from view by another vehicle thats stationary, and then they "jump" from that side into the street, to test how well the system works.
This on the other hand, was very textbook situation, there should have been no problems here by a system in Camry, let alone system with much better lidar. She is even crossing next to the street lights, I would guess that it is due to the poor dashboard cam that it looks so dark, as we can see that crash occurs just next to the streetlights.
In any case, system did not detect a problem at any point, it continued going 38 mph even after running over the person. It is clearly a system failure.
03-21-18, 07:58 PM
#689
Welcome to CL. I find it surprising that Uber hired this "safety driver" in the first place, and that she managed to keep her job for as long as she did. You would think that Uber monitors all the activity of its self driving cars and safety drivers, and should terminate drivers that are caught not paying attention to the road.
And I can understand human test driving not being attentive since they just sit around doing nothing or looking at their computer for info which is very distracting...
If we lived in serious world, there would be rules about autonomous testing... things like distractions, hands on wheel, how long they can drive before they have to rest, passing some kind of attention tests, etc.
But we live in silly world where lawmakers were trying to get a jump start on other states and hence we have AZ where there are no rules at all when it comes to autonomous testing. Now it looks very silly, doesnt it? But it is also very likely that if state attempted to add more rules, that media would criticize it for stopping the technology.
Most manufacturers dont test in California anymore since it requires them to report of failures.
03-21-18, 08:02 PM
#690
why in the world would you assume that everything is perfect? As if Uber is utopian company for which perfect utopian people work for? Your car has had multiple recalls, your laptop and cellphone likely had few issues this month, at least... why do you think Uber is run by people smarter than everyone else? If anything by all the evidence, they are not.
And I can understand human test driving not being attentive since they just sit around doing nothing or looking at their computer for info which is very distracting...
If we lived in serious world, there would be rules about autonomous testing... things like distractions, hands on wheel, how long they can drive before they have to rest, passing some kind of attention tests, etc.
But we live in silly world where lawmakers were trying to get a jump start on other states and hence we have AZ where there are no rules at all when it comes to autonomous testing. Now it looks very silly, doesnt it? But it is also very likely that if state attempted to add more rules, that media would criticize it for stopping the technology.
Most manufacturers dont test in California anymore since it requires them to report of failures.
And I can understand human test driving not being attentive since they just sit around doing nothing or looking at their computer for info which is very distracting...
If we lived in serious world, there would be rules about autonomous testing... things like distractions, hands on wheel, how long they can drive before they have to rest, passing some kind of attention tests, etc.
But we live in silly world where lawmakers were trying to get a jump start on other states and hence we have AZ where there are no rules at all when it comes to autonomous testing. Now it looks very silly, doesnt it? But it is also very likely that if state attempted to add more rules, that media would criticize it for stopping the technology.
Most manufacturers dont test in California anymore since it requires them to report of failures.