|
bobfather posted:Well, for one when it’s a human at the wheel they’re forced to maybe pay attention and hopefully stop in time not to hit you. no, about half of the 6k pedestrian deaths by car every year happen at night, by someone crossing a road outside of a crosswalk i'm not defending the car here, just saying that it's an unsafe practice which will get you killed if you're not paying attention shortspecialbus posted:But that's pretty much an ideal setup for the self-driving car to see and avoid an unexpected obstacle in the road. If it couldn't even do that under what's effectively ideal conditions for an unexpected pedestrian, it's kind of a big problem. yeah, but imo this is a limitation of the technology in particular and not necessarily uber's design in general. everyone is using this design, and the thing about lidar is it lacks depth perception - there has to be some calculation going on to spot a hazard, and a human sized object moving at a walking pace may not have been easily detectable with that design or sensor boner confessor fucked around with this message at 03:23 on Mar 22, 2018 |
#
?
Mar 22, 2018 03:20
|
|
|
|
| # ? May 29, 2024 21:25 |
|
|
boner confessor posted:, and the thing about lidar is it lacks depth perception What? One of the things LIDAR is really really good at is telling how far away things are. DAR = Detection and Ranging.
|
|
#
?
Mar 22, 2018 03:26
|
|
|
The car should be BETTER then a human at seeing this. There's no excuse for the car here. Despite the safe driver claims I have doubts a car controlled going 40 could stop in time from when the video illuminates it. Besides point though since she's obviously on her phone. And how public and  this industry is driver gonna fry.
|
|
#
?
Mar 22, 2018 03:29
|
|
|
Phanatic posted:What? if you put two lidar next to each other, yeah. you get binocular lidar. most self driving cars dont have that, uber's doest https://techcrunch.com/2018/03/19/heres-how-ubers-self-driving-cars-are-supposed-to-detect-pedestrians/ they say it's a "3d" image but that's over time, we're taking 2d snapshots with a single laser and piecing them together - a slower moving object may not have been visible, especially if it were only visible for a short time
|
|
#
?
Mar 22, 2018 03:29
|
|
|
boner confessor posted:no, about half of the 6k pedestrian deaths by car every year happen at night, by someone crossing a road outside of a crosswalk THEN IT SHOULD NOT BE ON THE ROAD YOU COLOSSAL JACKASS.
|
|
#
?
Mar 22, 2018 03:29
|
|
|
Avenging_Mikon posted:THEN IT SHOULD NOT BE ON THE ROAD YOU COLOSSAL JACKASS. i agree with you buddy, settle down before you have a stroke boner confessor posted:people often run blindly into traffic or otherwise behave badly around traffic, and get killed, and it's no big surprise to me that self driving cars fail to properly account for this like human drivers also fail Bombadilillo posted:The car should be BETTER then a human at seeing this. There's no excuse for the car here. Despite the safe driver claims I have doubts a car controlled going 40 could stop in time from when the video illuminates it. Besides point though since she's obviously on her phone. they're gonna blame her for using the technology exactly how millions of americans are going to use it when it's released lol
|
|
#
?
Mar 22, 2018 03:30
|
|
|
boner confessor posted:if you put two lidar next to each other, yeah. you get binocular lidar. most self driving cars dont have that, uber's doest Then the technology is inadequate, and failed an important test. That it was dark was irrelevant - the lidar would have seen the same thing in broad daylight. If the algorithms can't handle unexpected situations, then they need to be rewritten.
|
|
#
?
Mar 22, 2018 03:32
|
|
|
boner confessor posted:i agree with you buddy, settle down before you have a stroke For the past 5 pages you’ve done nothing but shift blame to the pedestrian and away from the company alpha-testing its poo poo on public roads. Even your phrasing, "oh, it’s just the technology, and everyone uses it anyway" comes off as callous and toadying. You sure could have fooled me about being on my side.
|
|
#
?
Mar 22, 2018 03:34
|
|
|
Deteriorata posted:Then the technology is inadequate, and failed an important test. That it was dark was irrelevant - the lidar would have seen the same thing in broad daylight. yeah that's what i've been saying on these forums for months and people call me a luddite for questioning technology self driving cars aren't safer and probably can't be safer, cars are inherently dangerous, etc. so on simply because they have to inhabit a world full of humans, who make dumb mistakes constantly. this is the thread for humans making dumb mistakes around machines even Avenging_Mikon posted:For the past 5 pages you’ve done nothing but shift blame to the pedestrian and away from the company alpha-testing its poo poo on public roads. Even your phrasing, "oh, it’s just the technology, and everyone uses it anyway" comes off as callous and toadying. You sure could have fooled me about being on my side. sorry that you have trouble reading and not screaming at people
|
|
#
?
Mar 22, 2018 03:35
|
|
|
boner confessor posted:if you put two lidar next to each other, yeah. you get binocular lidar. most self driving cars dont have that, uber's doest LIDAR systems don't need binocular vision - they typically use time-of-flight (or some kind of phase shift) to get distance information. If they don't, then it's not really LIDAR, just some kind of crappy laser scanner.
|
|
#
?
Mar 22, 2018 03:35
|
|
|
Avenging_Mikon posted:For the past 5 pages you’ve done nothing but shift blame to the pedestrian and away from the company alpha-testing its poo poo on public roads. Even your phrasing, "oh, it’s just the technology, and everyone uses it anyway" comes off as callous and toadying. You sure could have fooled me about being on my side. Your adding value judgment to him saying lidar is inadequate. Threads emotional.
|
|
#
?
Mar 22, 2018 03:35
|
|
|
I don't know enough about what the technology is to say for certain, but if you are correct that the tech in all cars is of the sort that wouldn't be able to deal with that incredibly straightforward obstacle moving at a predictable pace with radar-reflective materials then yes they all need to come off the road immediately. I still maintain that was pretty much an ideal situation for that car to see and avoid. If it can't do that basic thing because of tech limitations, the tech just isn't close to ready.
|
|
#
?
Mar 22, 2018 03:36
|
|
|
Hobnob posted:LIDAR systems don't need binocular vision - they typically use time-of-flight (or some kind of phase shift) to get distance information. If they don't, then it's not really LIDAR, just some kind of crappy laser scanner. they detect the "shadow" a moving object casts relative to the area around it, which probably isn't that pronounced for a human sized object moving at a slow speed directly in front of the vehicle https://www.youtube.com/watch?v=cyufiAyTLE0 shortspecialbus posted:I don't know enough about what the technology is to say for certain, but if you are correct that the tech in all cars is of the sort that wouldn't be able to deal with that incredibly straightforward obstacle moving at a predictable pace with radar-reflective materials then yes they all need to come off the road immediately. buddy i'm 100% with you in terms of all cars needing to come off the road immediately (i know that's not what you're saying but cars are inherently dangerous. they just are)
|
|
#
?
Mar 22, 2018 03:37
|
|
|
Hobnob posted:LIDAR systems don't need binocular vision - they typically use time-of-flight (or some kind of phase shift) to get distance information. If they don't, then it's not really LIDAR, just some kind of crappy laser scanner. I've worked extensively with Radar, but not Lidar. The "can't range" sounds odd. Because. It's in the name... been googling and the articles DO describe it as some sort of laser scanning. And not just bounce and return lidar. Articles might be written be people who don't know what they are talking about or it might be an adjacent technology just calling it lidar.
|
|
#
?
Mar 22, 2018 03:38
|
|
|
A human moving at a slow speed with a big bike and not a lot of background poo poo to sift.
|
|
#
?
Mar 22, 2018 03:38
|
|
|
boner confessor posted:buddy i'm 100% with you in terms of all cars needing to come off the road immediately (i know that's not what you're saying but cars are inherently dangerous. they just are) All self driving cars, as I would have expected to be obvious from context, you rear end. "all cars are dangerous" is the worst argument you've made yet.
|
|
#
?
Mar 22, 2018 03:40
|
|
|
Pull up thread. You're arguing with a half banana, half bird. And a corn-cob. It's obvious what really happened here. The car did exactly what it was made to do. Waymo is skipping the civilian market and going straight to automated, four wheel drive killbots. The military contracts will be worth billions and they will save millions in R&D by not having to avoid humans at all. For any other company, this incident is a bug. Now, it's a feature. Moto42 fucked around with this message at 03:44 on Mar 22, 2018 |
|
#
?
Mar 22, 2018 03:41
|
|
|
shortspecialbus posted:A human moving at a slow speed with a big bike and not a lot of background poo poo to sift. bikes aren't very "solid" and don't reflect much. like imagine that you're moving a laser pointer really fast across a bike, how much of the light hits the bike and how much passes through the spaces in it? shortspecialbus posted:"all cars are dangerous" is the worst argument you've made yet. they are, they move very fast around other humans. tens of thousands of people are killed by them every year. getting in a heavy metal object and accelerating it to high speeds without guidance other than a human is dangerous. i shouldn't have to explain this
|
|
#
?
Mar 22, 2018 03:43
|
|
|
Bombadilillo posted:Except for crossing a street in the dark. Not at a crosswalk, at a walking pace, not watching traffic run into her. A walking pace is the correct way for a pedestrian to cross the street. Running is wrong. Watching traffic run into her? When she started crossing the street that car would have been more than far enough away to see her and avoid her. A sober, attentive driver should have been able to see her. If she was riding that bike she could have been in violation for not having lights and reflectors at night, but those are usually not required for walking. I didn't see a crosswalk in the 7 seconds of video, nor is there one within sight up ahead, so it isn't like she ignored a nearby crosswalk. So how far out of their way should a pedestrian go? How far can traffic planners reasonably expect pedestrians to go? For a car going an extra block or two is no big deal, but it is a big deal when you are tired and out in the weather. Having a crosswalk every 40 yards could eliminate jaywalking. Would you be in favour of this safety change?
|
|
#
?
Mar 22, 2018 03:47
|
|
|
The fact that all cars are dangerous isn't particularly relevant to this situation, though, and so can be seen as an attempt at distraction. We aren't talking about all cars. We're talking about this car.
|
|
#
?
Mar 22, 2018 03:49
|
|
|
I'm on a phone and can't seem to get a link to it but if you Google search for "Radar Cross Section Measurements of Pedestrian Dummies ... - JRC Publications Repository - Europa EU" that is the name of a pdf that has the radar cross section of a pedestrian with a bike. It's pretty big, at least in comparison to pedestrians in general. I don't know enough to compare it to other likely obstructions, but it doesn't look at all hidden or anything.
|
|
#
?
Mar 22, 2018 03:54
|
|
|
Imagined posted:The fact that all cars are dangerous isn't particularly relevant to this situation, though, and so can be seen as an attempt at distraction. We aren't talking about all cars. We're talking about this car. the question is if self driving cars can reduce the inherent danger factor of cars, which are large objects that travel along the ground at high speeds. i'm not sure that's possible in any sort of near term timeframe Facebook Aunt posted:I didn't see a crosswalk in the 7 seconds of video, nor is there one within sight up ahead, so it isn't like she ignored a nearby crosswalk. she was about 500 feet from a crosswalk and passed near a sign that says "use crosswalk". here's the intersection in question https://www.google.com/maps/place/N+Mill+Ave+%26+E+Curry+Rd,+Tempe,+AZ+85281 go south a bit, where there's a big X in the median. the car was traveling north people are going to jaywalk, it happens, i've done it many times and i'll do it again. but you've got to exercise caution while doing so because even though drivers often have a legal and always a moral obligation to yield for pedestrians in the roadway, that won't stop someone who can't see you from killing you. seriously it kills thousands of people a year in the us - arizona and maricopa county are notably high for pedestrian deaths because of heavy emphasis on automobile travel over pedestrians https://www.ghsa.org/resources/spotlight-pedestrians18
|
|
#
?
Mar 22, 2018 03:56
|
|
|
boner confessor posted:bikes aren't very "solid" and don't reflect much.
|
|
#
?
Mar 22, 2018 03:58
|
|
|
Phanatic posted:The radar in a police radar gun can see a bike just fine, and that's a 10GHz beam, wavelength about 2 centimeters. A longer wavelength might not reflect well from something that's mostly empty space, like a bike frame, but LIDAR's using a frequency thousands of times higher than that, it's got accordingly higher resolution and accuracy. The pulses will absolutely reflect off the frame of the bike with very good resolution. the frequency of the beam doesn't matter compared to currently unknowable factors such as the sampling rate (number of "dots") and the algorithm uber was using to detect objects relative to the background behind those objects. like i said, if you shone a laser pointer across a bike, what proportion of the laser will hit the bike and what proportion will hit the area behind the bike? that determines how visible it is, not if the bike will absorb the light or whatever. it's a question of permeability boner confessor fucked around with this message at 04:03 on Mar 22, 2018 |
|
#
?
Mar 22, 2018 04:00
|
|
|
It’s really hard to judge how fast a car is going at night, this lady misjudged how much time she had to cross and now she’s dead. gently caress Uber.
|
|
#
?
Mar 22, 2018 04:02
|
|
|
Phanatic posted:The radar in a police radar gun can see a bike just fine, and that's a 10GHz beam, wavelength about 2 centimeters. A longer wavelength might not reflect well from something that's mostly empty space, like a bike frame, but LIDAR's using a frequency thousands of times higher than that, it's got accordingly higher resolution and accuracy. The pulses will absolutely reflect off the frame of the bike with very good resolution. Right, the pdf I can't seem to get the url from on mobile has data showing that. Here's a screen shot of some of it. 
|
|
#
?
Mar 22, 2018 04:02
|
|
|
boner confessor posted:the frequency of the beam doesn't matter compared to currently unknowable factors such as the sampling rate (number of "dots") and the algorithm uber was using to detect objects relative to the background behind those objects Sounds like you're saying that it should have been able to detect her, just that uber might have programmed it to not be able to. I don't know how that's any better than "it did see her, but the computer didn't act on that information"?
|
|
#
?
Mar 22, 2018 04:04
|
|
|
Facebook Aunt posted:A walking pace is the correct way for a pedestrian to cross the street. Running is wrong. The question of how many crosswalks to have is irrelevant, to the situation. How far cars are away when she started and if the can stop? If you as a pedestrian see a car and say "they have time to stop" and start not watching the cars coming. I don't know what to tell you. You are correct, the car should stop and the sober human avoid you. You are also hit by a car. It is reckless. It being reckless doesn't mean she deserved to get hit. Certainly didn't deserve to die. Especially THIS car of all cars should have stopped. Still reckless.
|
|
#
?
Mar 22, 2018 04:04
|
|
|
i hope self driving cars are all nuked so ican look at more jpgs and gifs of people doing dumb poo poo on work sites
|
|
#
?
Mar 22, 2018 04:05
|
|
|
boner confessor posted:yeah, but imo this is a limitation of the technology in particular and not necessarily uber's design in general. everyone is using this design, and the thing about lidar is it lacks depth perception - there has to be some calculation going on to spot a hazard, and a human sized object moving at a walking pace may not have been easily detectable with that design or sensor i hope you would agree that any sensor that cannot detect a human sized object moving at any speed between 0 and, let's say, 200 mph (relative to the sensor) has absolutely no goddamned place as the primary obstacle detection mechanism, or even a redundant one, on a self-driving car. but wait! there's more! this wasn't just a human-sized object -- it was a human-sized object pushing a 30 pound steel tube structure, and it had been on the roadway for several seconds and passed directly in front of the vehicle! from a detection standpoint this is significantly easier than, say, picking up a bicyclist from the rear, or identifying a construction flagman, or locating a pylon that's closing off one lane, and only slightly more difficult than reading a brick loving wall. every type of sensor used on autonomous cars (LIDAR, doppler radar, thermal camera, infrared camera, visual camera, ultrasonic) should be able to see that no problem. i would posit that the various sensors actually did see the woman and the bicycle, and the software told them to ignore it as a false positive, because we are so far from having reliable object-discrimination technology in this context that the only way Uber could possibly get something on the road and not have it stop every ten seconds when a bug flies in front of the emitter is to crank up the threshold and lie through their teeth about the system's effectiveness.
|
|
#
?
Mar 22, 2018 04:06
|
|
|
boner confessor posted:the question is if self driving cars can reduce the inherent danger factor of cars, which are large objects that travel along the ground at high speeds. i'm not sure that's possible in any sort of near term timeframe Yeah, 500 feet away is way too far. If I want to go to the place directly across the street from me and going to nearest crosswalk would add 1,000 feet to my trip, I probably wouldn't do it either. Traffic would basically need to be super heavy and physically impossible to cross before I would accept defeat and add nearly a quarter mile to my journey. Pedestrians aren't going to treat a flat road like an impassible river that can only be crossed at bridges. If you make it too difficult to cross safely, they will cross unsafely. City planners need to plan for actual human behavior or death is an inevitable result of their designs.
|
|
#
?
Mar 22, 2018 04:07
|
|
|
boner confessor posted:the frequency of the beam doesn't matter compared to currently unknowable factors such as the sampling rate (number of "dots") and the algorithm uber was using to detect objects relative to the background behind those objects. like i said, if you shone a laser pointer across a bike, what proportion of the laser will hit the bike and what proportion will hit the area behind the bike? that determines how visible it is, not if the bike will absorb the light or whatever. it's a question of permeability If they haven't thought to train their algorithm/computer brain to detect someone riding or pushing a bike perpendicular to their vehicles line of travel they should be charged with criminally negligent manslaughter (or whatever the appropriate term is in AZ).
|
|
#
?
Mar 22, 2018 04:07
|
|
|
Blast of Confetti posted:i hope self driving cars are all nuked so ican look at more jpgs and gifs of people doing dumb poo poo on work sites Be the change you want to be. By doing dumb poo poo on work sites, not posting.
|
|
#
?
Mar 22, 2018 04:07
|
|
|
i learn by example so you first
|
|
#
?
Mar 22, 2018 04:08
|
|
|
ekuNNN posted:They released the self-driving car footage, and it's clear the car should have easily seen her with its' lidar and radar: Was that Divine at the wheel? I thought she was dead?
|
|
#
?
Mar 22, 2018 04:09
|
|
|
tangy yet delightful posted:If they haven't thought to train their algorithm/computer brain to detect someone riding or pushing a bike perpendicular to their vehicles line of travel they should be charged with criminally negligent manslaughter (or whatever the appropriate term is in AZ). It's called, "blame the backup driver" this is Uber after all.
|
|
#
?
Mar 22, 2018 04:09
|
|
|
Jabor posted:Sounds like you're saying that it should have been able to detect her, just that uber might have programmed it to not be able to. i'm saying a few things -the bike is less visible than a human because the bike is less likely to be hit by and reflect a laser beam than a human, since a human is more solid. so the fact she had a bike doesn't really matter -given her speed and path relative to the vehicle, as well as the amount of time she was visible to the laser, it may have been more difficult to detect her than anticipated using lidar -there may be baked in assumptions about how humans would behave around cars, the designers of self driving cars may not anticipate people just casually walking into the roadway despite this being something that happens in reality with shocking frequency there's a lot of people to blame here - the pedestrian, the driver, the car designers, the people who designed the road (it practically invites jaywalking, and there are clear desire lines leading away from the eastern side of the roadway). and i want to emphasize here i dont think the woman deserved to die due to her behavior, just that her behavior was reckless my main gripe is i don't believe in the fundamental assumption that self driving cars can be made safer than manually driven cars in any meaningful timeframe, as in, during the next 5-10 years of development during which they will gradually be sold to the public, where they will kill more people Facebook Aunt posted:Pedestrians aren't going to treat a flat road like an impassible river that can only be crossed at bridges. If you make it too difficult to cross safely, they will cross unsafely. City planners need to plan for actual human behavior or death is an inevitable result of their designs. yes, and the road design is bad, and there is visible proof of frequent jaywalking at this point (the signs that say not to do it, the trails demonstrating people have done it) but do you agree with me or not that when jaywalking, it is the obligation of a pedestrian to exercise caution? i believe it is boner confessor fucked around with this message at 04:12 on Mar 22, 2018 |
|
#
?
Mar 22, 2018 04:09
|
|
|
While the video is damning, what did the lidar see, if anything? If I registered something ahead of the vehicle and tried to do something but couldn’t in time, then there’s an obvious limit to how quickly/far away lidar can react at certain speeds. In this case, autonomous cars should therefore be speed limited until they can be proven to avoid a situation like this at higher speeds. If it registered something and did nothing, that’s on Uber, their code, and tech and their vehicles should absolutely be pulled from public roads. If it registered nothing, why? Then it’s why was Uber technology allowed on the street if it couldn’t detect a pedestrian walking at a normal pace into the path of the vehicle? On top of all that, according to Ars Technica, the pedestrian is only visible (and that’s only her shoes) 1.4 seconds before impact. It’d require an attentive driver making a split second correct decision to avoid the same thing happening in a normal vehicle. Not blaming the victim but even in the video, you can see her until you’re nearly on top of her. The headlights didn’t reveal her until the last possible moment. Thwomp fucked around with this message at 04:15 on Mar 22, 2018 |
|
#
?
Mar 22, 2018 04:10
|
|
|
Put all the cars into tunnels underground, turn the surface roads into parks. Never have a kid chase a ball into traffic again.
|
|
#
?
Mar 22, 2018 04:12
|
|
|
|
| # ? May 29, 2024 21:25 |
|
|
Facebook Aunt posted:Yeah, 500 feet away is way too far. If I want to go to the place directly across the street from me and going to nearest crosswalk would add 1,000 feet to my trip, I probably wouldn't do it either. Traffic would basically need to be super heavy and physically impossible to cross before I would accept defeat and add nearly a quarter mile to my journey. Fine, but maybe look for a car coming to kill you.
|
|
#
?
Mar 22, 2018 04:12
|
|
