Shop OBEX P1 Docs P2 Docs Learn Events
First Self-Driving Car Fatality - Page 2 — Parallax Forums

First Self-Driving Car Fatality

2»

Comments

  • jmgjmg Posts: 15,175
    'Just watched the dashcam video. The bicyclist/pedestrian was obviously oblivious to oncoming traffic. ..
    -Phil
    I wonder if they were USA native, or from another country, that drives on the other side of the road. Hard to fathom otherwise ?

  • It was a her, and this person had a history of homelessness and other issues. Not to say this excuses the accident, but such people are pretty common in towns of any size. Cars/drivers need to be aware of them.

    The fact that it was "dark" is irrelevant. All self-driving cars have, or should have, sensors that operate outside the visible light range. A LIDAR requires no external light. It will work in the vacuum of the deepest, darkest space.
  • It was a her, ...
    You're aware, I hope, that this data point is irrelevant. The facts of the case are, and should be, gender-neutral.

    -Phil
  • evanhevanh Posts: 16,032
    The thing with automation is, any human supervision is always a delayed response before taking the controls. I've seen machinery jam up then twist until something breaks, and the operators just stand there watching in shock rather than hit the E-Stop.

    The so-called "safety" person can't realistically be expected to take over while still in the mindset of expecting the machine to react appropriately. It's all over before they can react to the situation.

    Such a person likely just reports on the feel/look of the ride and deals with any oddball stop conditions, again, after the fact.

  • evanhevanh Posts: 16,032
    So ... the software absolutely needed to handle this situation flawlessly ... and it didn't.
  • Cluso99Cluso99 Posts: 18,069
    edited 2018-03-22 08:01
    Just saw the video on TV. The bike/pedestrian came out of he shadows on the left lanes. Was obviously oblivious to any traffic. Clearly pedestrians fault.
    But the driver wasn't paying proper attention either. And the sensors should have picked the pedestrian up too.
    Like many accidents, a lot of. Ontributiing factors.
    That does not excuse the idiot pedestrian. And would you expect a pedestrian crossing that road???
    I would guess that if the driver was driving, any response could have been worse, like rolling the vehicle thy trying to swerve. What about any close vehicles as not enough tome to check
  • evanhevanh Posts: 16,032
    The important point is the pedestrian's actions is a perfectly valid scenario, and there is plenty more equally unexpected possibilities. An undistracted driver would've handled it without breaking a sweat, along with some swearing thrown in to show how easy it really was.

    A machine needs to be superior in every way with this sort of problem or its worthless. It'll be interesting to find out how come the Uber car failed so badly after so much development.

  • GordonMcCombGordonMcComb Posts: 3,366
    edited 2018-03-22 17:44
    It was a her, ...
    You're aware, I hope, that this data point is irrelevant. The facts of the case are, and should be, gender-neutral.

    In a post above someone repeatedly referred to the victim as a he. We should show respect in correctly referring to this real person, regardless of cold facts.

  • In the video (which was not taken with a camera that replicates the greater dynamic range of the human eye), the pedestrian comes from an empty lane on the left into the lane on the right before being struct by the Uber vehicle. There is NO WAY the sensors of the Uber car should have not seen the pedestrian. Forget shadow, forget jaywalking, forget everything else: the automation of the car failed to do its job.

    The dash cam is very low quality and is not used in the automation. There are other camera views that will be more telling, if Uber ever releases them. These include the visible light full color camera as well as the infrared camera. The data reconstruction from the LIDAR will be the most telling.
  • evanhevanh Posts: 16,032
    Here's a quote from NYT - 'Waymo, formerly the self-driving car project of Google, said that in tests on roads in California last year, its cars went an average of nearly 5,600 miles before the driver had to take control from the computer to steer out of trouble. As of March, Uber was struggling to meet its target of 13 miles per “intervention” in Arizona, ...'

    Rather damning contrast there.


    And their piccy of locations: autonmous-uber-pedestrian-graphic.jpg

    Eek, that looks like the car was changing lanes toward the pedestrian!
  • kwinnkwinn Posts: 8,697
    evanh wrote: »
    Here's a quote from NYT - 'Waymo, formerly the self-driving car project of Google, said that in tests on roads in California last year, its cars went an average of nearly 5,600 miles before the driver had to take control from the computer to steer out of trouble. As of March, Uber was struggling to meet its target of 13 miles per “intervention” in Arizona, ...'

    Rather damning contrast there.


    And their piccy of locations: autonmous-uber-pedestrian-graphic.jpg

    Eek, that looks like the car was changing lanes toward the pedestrian!

    OMG, RoadNet, possibly leading up to SkyNet?
  • Heater.Heater. Posts: 21,230
    When the first cars were allowed on public highways there were strict limits on speed, they required a guy with a red flag to walk in front of the car to warn the public that this dangerous contraption was coming.

    Just saying...

    I'm wondering about the kind of guys taking the job to be the minder, sitting in self-driving cars tasked with the impossible task of taking control in an instant when things go wrong. I would not want to be the minder in case, having to watch the machine you are responsible for kill somebody.


  • Dave HeinDave Hein Posts: 6,347
    edited 2018-03-24 23:57
    evanh wrote: »
    ...

    autonmous-uber-pedestrian-graphic.jpg

    Eek, that looks like the car was changing lanes toward the pedestrian!
    I don't believe the annotations on the picture are accurate. According to the dashcam video, the pedestrian was struck around the bottom of the picture, and not where the text and arrow indicate. The video shows the dashed lines on the right side of the road beginning at a point past where the victim was walking her bike. There is no indication in the video that the car was changing lanes prior to the accident. If the car did change lanes and move to the right, it was probably a result of the car sensing a collision.

  • I'm not sure what the annotations are supposed to be, but you are correct that they aren't anywhere near accurate.

    The dashcam video doesn't show a swerving -- it would be a mighty stupid computer to steer into the path of the moving object, so that would not be called a defensive maneuver. And is reveals at the time of impact the victim was positioned just before where the white line (right edge of the lane) transitions form solid to dashed.
  • evanhevanh Posts: 16,032
    My interpretation of the diagram is the car is tracking into the new lane to conduct an upcoming turn. Not a swerve. Not aiming at the pedestrian. Ie: The car is entirely oblivious to the obstacle ... but the accident may not have been fatal if the lane change wasn't also under way.

    Just another contributing factor but one I wasn't expecting to be.

    Heater,
    Yeah, the minder must be in a terrible state right now.
  • evanhevanh Posts: 16,032
    edited 2018-03-25 22:22
    Dave Hein wrote: »
    According to the dashcam video, the pedestrian was struck around the bottom of the picture, and not where the text and arrow indicate. The video shows the dashed lines on the right side of the road beginning at a point past where the victim was walking her bike.

    True. Quite a discrepancy there. A lane change, if any, doesn't seem very relevant now. Wonder how NYT managed to come up with that idea.

    Good to see a map of the accident location at least.

    EDIT: Fixed the spell checker's bad guess.
  • Too many things about the diagram are not reliable. It shows the car at least 25 feet beyond where it is in the video. The annotations don't provide enough distance between where the impact was said to have been and where the body was seen, given a nearly 2 ton car traveling at 40 MPH. There are other issues with the map, but you're right it is interesting to see a general map.
  • ercoerco Posts: 20,257
    Not necessarily related: fiery fatal Tesla Model X crash yesterday in NorCal. Unknown if autopilot was engaged. Reports say the battery caught fire.

    https://www.nbcwashington.com/news/national-international/Sig-Alert-Issued-After-Accident-Causes-Car-Fire-on-Hwy-101-477762763.html
  • To be clear, Tesla cars are not self-driving. The Autopilot feature (poorly named) is designed for driver safety assist.

    Elon wants to land on Mars, but he needs to be smarter what he calls things. "Autopilot" suggests the thing drives itself.
  • KeithEKeithE Posts: 957
    edited 2018-04-01 21:59
    There are a couple of things which have been in the news regarding the Tesla crash. Someone else crashed there about 10 days before, and the safety barrier had not yet been replaced apparently due to the recent storms.

    And autopilot in that particular Tesla allegedly kept swerving towards that barrier. Sounds like something out of a movie plot. Not clear why the driver kept engaging autopilot if he felt it was malfunctioning. I haven’t seen Tesla comment on that, but they did comment on the barrier and the driver.

    "Walter Huang's family tells Dan Noyes he took his Tesla to the dealer, complaining that -- on multiple occasions -- the auto-pilot veered toward that same barrier -- the one his Model X hit on Friday when he died."

    Edited to add - Tesla seems to be disputing the family’s statement:

    “We’ve been doing a thorough search of our service records and we cannot find anything suggesting that the customer ever complained to Tesla about the performance of Autopilot,”
  • ercoerco Posts: 20,257
    You can't fix stupid! Tesla driver jumps into passenger seat.

    http://www.newser.com/story/258541/driver-banned-from-roads-after-autopilot-seat-switch.html
Sign In or Register to comment.