Tempe Police Release Video of Uber Accident • Hacker News Board

Re: Tempe Police Release Video of Uber Accident

#341

Post by mkempe » Thu, Mar 22, 2018, 12:54 AM UTC

eric_b wrote: post #152

The Uber car was a Volvo XC90. They are supposed to have some of the best pedestrian avoidance safety features in existence. My understanding is that a stock Volvo would be expected to catch this situation and avoid or stop. Did Uber override these systems? Or does it just not work very well?

I had the same question, especially because several relatives and I own Volvos and pay attention to their safety promises. One would assume that Uber cannot disable Volvo's safety systems...

However, it appears that Volvo's "City Safety" detection of pedestrians or bicyclists requires clear visual identification. So it works in daytime, or at night with strong illumination, but not in relative darkness. [1]

One other thing that seems off in the video is that the headlights don't illuminate much distance ahead.

[1] https://support.volvocars.com/en-CA/cars/pages/owners-manual...

Re: Tempe Police Release Video of Uber Accident

#342

Post by cozzyd » Thu, Mar 22, 2018, 12:54 AM UTC

fudged71 wrote: post #251

I do wonder if perhaps Uber made the video darker before releasing. Or used a camera that is not the main sensing camera. Because I would hope the cars are smart enough to increase ISO to get more detail at night time.

I suspect the raw output of their main sensing sensors is proprietary somehow.

Re: Tempe Police Release Video of Uber Accident

#343

Post by CamperBob2 » Thu, Mar 22, 2018, 12:54 AM UTC

Something that's worth pointing out as well is what happens to low-light contrast sensitivity as we age. I believe there's something like a 50% loss in night vision between the ages of 20 and 50.

Having watched the video, I seriously doubt I'd have seen the pedestrian in time to stop for her. I don't know that I would even feel that sorry for her, frankly -- there is only so much we can (and should) do to protect drunks and idiots from hazards like crossing the street at night without looking.

Finally, had the driver been paying attention, at least some blame could probably be assigned to DOT's stubborn refusal to bring their lighting standards into the 21st century. Cars in other countries are equipped with noticeably better lighting tech than US cars are allowed to use. It looks like you'd be overdriving these headlights at any speed over 15 MPH or so.

Re: Tempe Police Release Video of Uber Accident

#344

Post by zamalek » Thu, Mar 22, 2018, 12:55 AM UTC

jbritton wrote: post #293

I would hope self driving cars have better sensors. There are high dynamic range cameras that simultaneously take multiple exposures and composite them in real time. The dynamic range in this video is horrible. An HDR camera should have made her visible much sooner. Further why not use larger CMOS sensors, something like on a DSLR camera. Finally, the headlights appear fairly ineffective too. I would hope these cars…

Xenon lights should be required on all new car sales. My new car has them and the difference is incredible. Not only can I see far more, but signage really pops. I've noticed that they are also much easier to look at when approaching. It is an incredible safety feature.

Re: Tempe Police Release Video of Uber Accident

#345

Post by TallGuyShort » Thu, Mar 22, 2018, 12:55 AM UTC

jonathanyc wrote: post #166

Treblemaker wrote: post #32

The first question that comes to my mind watching this, is, why didn't the _pedestrian_ see the _car_? It looks as though she didn't even look up until the car was only a few feet away from here. I'm not just trying to "blame the victim". The car "should" have responded sooner. But see-and-avoid works both ways. If you're choosing to cross a road in the dark, you, too, have a responsibility to understand the drivers'…

The pedestrian died, not the SDV developers. Calling this a “teachable moment ... also for pedestrians” is rather disrespectful.

The fact that the pedestrian died is precisely the point. As a frequent pedestrian, and as a person who likes people, and because people are often pedestrians, I would like to prevent pedestrian deaths. That happens by holding SDV developers to rigorous standards, but also by highlighting that crossing the street safely requires paying close attention and being careful, more so than many people do. I've seen several places where pedestrians have the right of way a lot, and many will just step directly into the street, legally mind you, when cars are traveling quickly nearby. No matter how you slice it, that yields an unacceptably high probability of death, and you can blame whoever you want but you will still be dead. There's a difference between victim blaming, and pointing out that you can avoid becoming a victim by taking sensible precautions. This is the latter. It doesn't absolve SDV developers of any carelessness they may have had in their work to point that out too.

Re: Tempe Police Release Video of Uber Accident

#346

Post by trhway » Thu, Mar 22, 2018, 12:55 AM UTC

The onus is on driver to see what the path ahead is clear. Uber clearly violated the 3 sec. rule [i.e. reaction time plus braking time with the braking time dominating the equation]. Uber was driving too fast for the conditions. If you have only 30m clear view, you can't drive faster than 10m/s. Driving faster, like Uber did here, is basically driving blind, reckless and negligent. The woman was in the headlights for only 1 second - thus Uber was driving 3 times faster than the maximum safe speed in those conditions ( specifically the power and angle of their low beam headlights)

>the pedestrian showed up in the field of view right before the collision

It takes AZ police and $70B corp together to conclude that the woman appeared too fast. Guess what? If people and objects appear too fast in the field of view then you're driving too fast.

Re: Tempe Police Release Video of Uber Accident

#347

Post by mehrdadn » Thu, Mar 22, 2018, 12:56 AM UTC

StavrosK wrote: post #329

fudged71 wrote: post #251

I do wonder if perhaps Uber made the video darker before releasing. Or used a camera that is not the main sensing camera. Because I would hope the cars are smart enough to increase ISO to get more detail at night time.

The cars don't have to increase ISO to get detail, that's for human eyes. Increasing ISO only works because we can't see details in shadows, but computers don't have that limitation. A neural network wouldn't have to add 100 to each pixel to make out detail, like a human would.

> The cars don't have to increase ISO to get detail, that's for human eyes. Increasing ISO only works because we can't see details in shadows, but computers don't have that limitation. A neural network wouldn't have to add 100 to each pixel to make out detail, like a human would.

I'm not an expert on this stuff but I think you're misunderstanding or mixing up the ISO value with exposure compensation. The ISO value corresponds to the sensor's sensitivity, and even RAW photos (which are more or less dumps of the raw sensor data) are captured with specific ISO values, and look different at different ISOs.

Re: Tempe Police Release Video of Uber Accident

#348

Post by osrec » Thu, Mar 22, 2018, 12:56 AM UTC

brokenmachine wrote: post #223

osrec wrote: post #83

That is utterly terrible. The car should have reacted to the pedestrian comfortably. I think something in the change in lighting confused the sensors (the road appears to go from having street lights to not having them at the point of contact).

I would have thought that it should be irrelevant if the street lights are there or not with LIDAR.

Not if the algo is making statistical inferences from a visual input as well. I feel it may have clouded the "perception" of what was actually there.

Re: Tempe Police Release Video of Uber Accident

#349

Post by hcknwscommenter » Thu, Mar 22, 2018, 12:56 AM UTC

freehunter wrote: post #42

jonknee wrote: post #30

Earlier quoted context omitted.

> there is no such thing as a too reflective vest for bicyclists And to not jaywalk a bike across a street in the dark!

Since I've seen this be a very controversial statement here in the past, I'd just like to point out to everyone that American drivers on four lane roads are not expecting pedestrians to be in the middle of the road and not at a crosswalk. That is not a normal thing to happen in America, and no one expects it to happen. It is a very rare situation. --edit: okay guys... no where did anyone argue that you shouldn't be w…

>the pedestrian has the right to the road over the car. Maybe it's true in Europe judging by comments on previous articles, but it's not common in the US. It's illegal.

You are conflating two different concepts. In every state in the U.S., it is illegal to hit a pedestrian whether or not they are where they are supposed to be. If you see a pedestrian jaywalking then you are required to avoid the pedestrian to the greatest extent possible. Anything less is at least manslaughter.

Re: Tempe Police Release Video of Uber Accident

#350

Post by bigiain » Thu, Mar 22, 2018, 12:56 AM UTC

burke wrote: post #285

ars wrote: post #74

Direct link: https://twitter.com/TempePolice/status/976585098542833664 The interior view tells you everything you need to know about why anything except perfection in self driving cars will lead to more accidents. The driver specially selected and trained for exactly this job, does not look at the road! If someone specially trained for that won't do it, the general public certainly will not. Also, in the video you ca…

> anything except perfection in self driving cars will lead to more accidents. Only if perfection is defined as "statistically better than humans". I do agree that these "assistive" technologies have the potential to do more harm than good, but the bar is actually far lower than "perfection".

I wonder what the "deaths per vehicle miles" stats for autonomous vs human-piloted cars look like right now? Humans kill, what, 30,000 other humans a year or so in The US? But there's a _lot_ of vehicle miles driven. My guess is this one incident has likely pushed the autonomous car death rate per vehicle mile way beyond the current human piloted death rate per vehicle mile...

(Having said that, there's not much doubt in my mind that stuff like antilock brakes and autonomous emergency braking have reduced the road toll (and will reduce it further as these recent-ish technology become more common i the aged vehicle fleet) - I think the difference there is then those technologies are not nearly so inclined to make driver pay less attention to what's going on around them. I _hope_ people don't think "It's OK, I've got emergency braking - I'll just watch this tv show as I drive to work!")