Follow Slashdot blog updates by subscribing to our blog RSS feed

 


Forgot your password?
Close
typodupeerror
×
AI Transportation

Blocked Traffic, Disrupted Firefighters: Why San Francisco Wants to Slow Robotaxi Rollout (nbcnews.com) 93

Posted by EditorDavid from the golden-tailgating dept.
"San Francisco is trying to slow the expansion of robotaxis," reports NBC News, "after repeated incidents in which cars without drivers stopped and idled in the middle of the street for no obvious reason, delaying bus riders and disrupting the work of firefighters." The city's transportation officials sent letters this week to California regulators asking them to halt or scale back the expansion plans of two companies, Cruise and Waymo, which are competing head-to-head to be the first to offer 24-hour robotaxi service in the country's best-known tech hub.

The outcome will determine how quickly San Francisco and possibly other cities forge ahead with driverless technology that could remake the world's cities and potentially save some of the 40,000 people killed each year in American traffic crashes.... Neither vehicles from Cruise or Waymo have killed anyone on the streets of San Francisco, but the companies need to overcome their sometimes comical errors, including one episode last year in which a Cruise car with nobody in it slowly tried to flee from a police officer.

In one recent instance documented on social media and noted by city officials, five disabled Cruise vehicles in San Francisco's Mission District blocked a street so completely that a city bus with 45 riders couldn't get through and was delayed for at least 13 minutes. Cruise's autonomous cars have also interfered with active firefighting, and firefighters once shattered a car's window to prevent it from driving over their firehoses, the city said....

"A series of limited deployments with incremental expansions — rather than unlimited authorizations — offer the best path toward public confidence in driving automation and industry success in San Francisco and beyond," three city officials wrote Thursday in a letter to the utilities commission, the state agency that decides if a company gets a robotaxi license. A second letter expressed concerns about Waymo....

Cruise has argued that its service is safer than the status quo.
A Cruise spokesperson also provided letters of support "written by local San Francisco merchants associations, disability advocates and community groups." And U.S. Transportation Secretary Pete Buttigieg told Quartz last year that "it would be hard to do worse than human drivers when it comes to what we could get to theoretically with the right kind of safe autonomous driving."

But in 2021 CBS reported that dozens and dozens of Waymo's robo-taxis kept mistakenly driving down the same dead-end street. And in 2018 a self-driving Uber test vehicle struck and killed a woman in Arizona.

More stories from the Verge: In July, a group of driverless Cruise vehicles blocked traffic for hours after the cars inexplicably stopped working, and a similar incident occurred in September. Meanwhile, a driverless Waymo vehicle created a traffic jam in San Francisco after it stopped in the middle of an intersection earlier this month. The National Highway Traffic Safety Administration opened an investigation into Cruise last December over concerns about the vehicles blocking traffic and causing rear-end collisions with hard braking... [San Francisco] city officials also express concern over the way driverless vehicles deal with emergency vehicles. Last April, officials say an autonomous Cruise vehicle stopped in a travel lane and "created an obstruction for a San Francisco Fire Department vehicle on its way to a 3 alarm fire...."

Other incidents involve Cruise calling 911 about "unresponsive" passengers on three separate occasions, only for emergency services to arrive and find that the rider just fell asleep.... Officials say companies should be required to collect more data about the performance of the vehicles, including how often and how long their driverless vehicles block traffic.
This discussion has been archived. No new comments can be posted.

Blocked Traffic, Disrupted Firefighters: Why San Francisco Wants to Slow Robotaxi Rollout More

Blocked Traffic, Disrupted Firefighters: Why San Francisco Wants to Slow Robotaxi Rollout

Comments Filter:

  • It isn't just robotaxis stopping in the middle of the street for no reason. Teslas are pulling over and stopping [theintercept.com] in traffic causing accidents as well as deciding to brake for no reason [cbsnews.com] while on the highway.

    But yeah, all that great software.

    • Re: (Score:3, Funny)

      by mangomama ( 1671024 )
      FSBC, or Full Self-Brake Checking, was included in a subsequent release after Elon promised a more aggressive autopilot.
  • What a pathetic attempt to derail robotaxis.

    • Re: (Score:3)

      by Entrope ( 68843 )

      It's good (in some minimal sense) that these bugs have been a fail-safe kind of bug, but there are good reasons we don't let ten-year-olds drive, and why many states require adult supervision in a vehicle when a new driver is at the wheel.

      • there are good reasons we don't let ten-year-olds drive

        +1 informative (here in Brazil, the age to get driver's license is 18)

        • It's more or less 18 here as well. Not sure if 16 can still get a permit, but that's how it was 20 years ago.

          Of course, if someone isn't allowed to drive a car until 18, why would we consider letting them vote?

          • It's completely different things: minors than 18 can't drink alcohol and be jailed too, here (but can work, formally, from 14 [only apprentice jobs: normal work needs 16])

            • I agree!!! voting is much more important and has significantly more consequences then letting a 16 year old have a license.

              • I agree (but a 16 year old is biologically/mentally fine to vote, but not to drive a car or drink alcohol [even very common in this age - we've saw it in the past, drink and drive can't be punished with jail...]) I agree more with this legal condition than that in the US (which, I presume, you are comparing: US is a big mess to me [even trowing guns in the mix...])

                • We will just have to disagree on the matter. A 16 year old is extremely easy to sway with social pressure. They don't tend to vote Democrat because they generally understand the platform or agree with it. Sure, some really might and do. Those individuals grew up on politics, probably have discussions around the dinner table, etc. Same as my family. The average person just doesn't care and will vote how their friends insist they must or you aren't my friend, blah blah blah.

                  I likely cost myself numerous dates

            • Also, if we can send an 18 to war, why can't they have a drink? You think they have the mental capacity to kill and deal with that but not have a drink? Sorry, but completely disagree. Or maybe we should raise recruiting to 21, so their brains have more time to develop. Not likely to happen.

      • Yes, and of course no adult human driver ever killed anypne.

        • Yes, and of course no adult human driver ever killed anypne.

          So, your answer is to let buggy/unproven software kill people instead?

    • Or maybe urban driving is an AI hard problem and even with on demand remote control autonomous vehicles are just an economic liability for cities.

    • Or maybe they should fix their bugs before pretending their cars are production ready.

    • Indeed. Human drivers do stupid, disruptive and reckless shit on a very regular basis. But there was this one time when a robotaxi blocked the road for a few minutes, therefore these things need to be banned.

  • In conclusion ... (Score:3)

    by AlanObject ( 3603453 ) on Sunday January 29, 2023 @03:07PM (#63249245)

    This must be Elon Musk's fault somehow. I just know it.

  • Big Tech Will Save Us! (Score:5, Interesting)

    by Cbs228 ( 596164 ) on Sunday January 29, 2023 @03:13PM (#63249253)

    The outcome will determine how quickly San Francisco and possibly other cities forge ahead with driverless technology that could remake the world's cities and potentially save some of the 40,000 people killed each year in American traffic crashes....

    Per the NHTSA, there were 38,824 traffic fatalities in the United States [dot.gov] [p. 5] in 2020. Assuming a population of 329.5 million, that's 11.8 fatalities per 100,000 population. Meanwhile, during that same timeframe in the Netherlands [itf-oecd.org] [p. 3], there were 3.8 fatalities per 100,000 population. Both fact sheets cover road users other than automobiles, including pedestrians and cyclists.

    If increasing road safety is the ultimate goal, that can be done today using technology, infrastructure, and methods which are well-known and proven. There is no need to wait around for Big Tech or anyone else to reinvent the automobile—only with greater expense and more points of failure.

    • The outcome will determine how quickly San Francisco and possibly other cities forge ahead with driverless technology that could remake the world's cities and potentially save some of the 40,000 people killed each year in American traffic crashes....

      Per the NHTSA, there were 38,824 traffic fatalities in the United States [dot.gov] [p. 5] in 2020. Assuming a population of 329.5 million, that's 11.8 fatalities per 100,000 population. Meanwhile, during that same timeframe in the Netherlands [itf-oecd.org] [p. 3], there were 3.8 fatalities per 100,000 population. Both fact sheets cover road users other than automobiles, including pedestrians and cyclists.

      "The Netherlands recorded 661 road fatalities in 2019, representing a 2.5% decrease when compared to 2018."

      The Netherlands calls 661 fatalities a year's worth of road statistics.

      The United States calls 661 fatalities a bad month for Chicago.

      I believe these country "comparisons" fall into are you fucking kidding me category.

      If increasing road safety is the ultimate goal, that can be done today using technology, infrastructure, and methods which are well-known and proven. There is no need to wait around for Big Tech or anyone else to reinvent the automobile—only with greater expense and more points of failure.

      You seem to be overlooking the most obvious behavior on the road today. When an electronic distraction device is strapped to the hand of damn near every American driver over the age of social-media-addict, it's not the damn drunks I'm worried about. Big Tech is being called upon for two reas

      • You must not have children anywhere near drivers age. The drivers ed training that US youngsters get nowadays is light years better and more extensive than what I had when I was learning to drive. I see more middle aged people swerving down the road while they look at their cell phone than kids. Nowadays, drivers ed hammers that sort of lesson into the kids heads like you wouldn’t believe.

        It’s true that 16-17 year olds are still the most dangerous drivers, cause they’re immature, but t

        • Re: (Score:3, Interesting)

          by geekmux ( 1040042 )

          You must not have children anywhere near drivers age. The drivers ed training that US youngsters get nowadays is light years better and more extensive than what I had when I was learning to drive. I see more middle aged people swerving down the road while they look at their cell phone than kids. Nowadays, drivers ed hammers that sort of lesson into the kids heads like you wouldn’t believe.

          Oh, the assumptions. Been driving with a full license since I was 15 years old due to drivers ed. Raised and am raising kids still. All the drivers ed training in the damn world does nothing to address tech addiction that takes a foothold years prior. You don't merely "teach" that away. We know this based on damn near every other addiction in the world.

          I suspect that road ragers and drunk drivers are a much bigger part of the equation.

          There are tens of millions of smartphone addicts who drive every day distracted, and multiple times a day. Road ragers are not always raging, and drunk

        • In the UK, while you can theoretically get a licence at 17, the cost of insurance for anyone under 25 makes it basically unaffordable. You are looking at a minimum of £6,000 per year for insurance. For comparison, my insurance is about £300.

      • Re: (Score:2)

        by Cbs228 ( 596164 )

        One; young drivers are social media addicts, and don't have time for paying attention.

        I'm sure that social media is still a thing in the Netherlands and anywhere else with a lower per-capita traffic fatality rate.

        Two, they don't even remotely want the responsibility.

        Or the cost, which may be a substantial driver of behavior in today's economy. Owning a car comes with a substantial up-front cost and a recurring expense. My current vehicle—which is a very basic, inexpensive car—still costs me about $0.45 USD/mile to operate. Larger or older vehicles are even more expensive per-mile. I didn't own a car as a teenager, or drive very much

      • You seem to be overlooking the most obvious behavior on the road today.

        America (and Canada) have shockingly poorly designed roads. The only thing that saves them from being worse death traps than they already are is that they are also very poor at traffic flow and so the speed is quite low due to being incredibly prone to traffic jams.

        But sure, blame it on the youth of today, not insane zoning and roadbuilding regs put in place by people you voted for.

      • Re: (Score:2)

        by spitzak ( 4019 )

        You do know they have cellphones in the Netherlands, don't you?

    • Yep. Agree. Effective transport solutions for large cities tend to be light rail, tram, metro, bus, & regular taxis. Once you get the privately owned cars off the streets, pedestrianise some of them & make some nice public areas, you end up with a nicer, cleaner, greener city to live in & it's easier, quicker, & cheaper to get around. Everyone wins, except the car manufacturers & sales reps. The rest of the world is gradually going that way. The trouble is, 'Muricans don't wanna be like

      • Everyone wins, except the car manufacturers & sales reps.
        And people who want to carry groceries home.

        • Do you buy car loads of groceries every day? If so, you'd be better off ordering most of them online & then shop occasionally for the stuff you want to pick out for yourself.

          For everyone else who maybe does a big shop once a week, there's nothing to stop them from using their car. For people who live downtown, within easy walking distance of local supermarkets, markets, shops, etc., there are these marvelous shopping carts that people around the rest of the world use. They even go up & down stairs

      • Nice fantasy, but in how many places does this actually work in real life?

        • Do you mean in the rest of the world? Check it out for yourself. They're investing like mad in public transport infrastructure.
    • Even in the US, as in the EU, rates very. Massachusetts is 4.9 per 100k. Romania is at 9.3.

    • Re: (Score:3)

      by uncqual ( 836337 )

      (Stats from current CIA World Factbook unless otherwise noted and dates of stats vary from country to country)

      One should factor in demographics and geographic differences when comparing traffic deaths per capita. I've not found good stats for "distance traveled in passenger car annually" for the USA vs. the Netherlands, but...

      Land area of USA (excluding territories): 9.8M sq km
      Land area of the Netherlands: 41.5K sq km

      Roadways in USA: 6.5M km
      Roadways in the Netherlands: 139K km

      Population of USA: 337M
      Populati

      • Re:Big Tech Will Save Us! (Score:4, Interesting)

        by serviscope_minor ( 664417 ) on Monday January 30, 2023 @04:03AM (#63250315) Journal

        Why not go for Sweden in your comparison? I think it has a lower population density than the US. Of course one could argue that the local, not country wide density matters. Then you could include smaller countries. Sure people drive less far in the Netherlands, but the Netherlands have also put a LOT of effort into getting people out of cars into other transport modes whereas America has done the opposite.

        Netherlands is perhaps more comparable to a large American city not the whole country, but you will find it still has a rather better record on road safety.

        Ultimately every country is different. If you focus on the differences excessively you end up with "well that's just the way it is" and miss the cause problems, or maybe even that America has a shockingly poor road safety record. Even by its own standards it's been getting worse.

        America used to have compact, walkable town centres, amenable to the Dutch style transportation systems. They were ripped out after massive lobbying and propaganda efforts from the car industry. This isn't because Americans are idiots, Europeans had similar plans too. But with half of the continent being in ruins from WW2 meant there just wasn't the money for reengineering cities for the supposed future. One of the ones that almost took hold in London was defeated by a massive backlash from residents across the city (this also happened in the 1667 too. It's a London thing).

        End result though, transportation wise America is in a pickle and has unsafe roads.

    • Re: (Score:3)

      by khchung ( 462899 )

      If increasing road safety is the ultimate goal, that can be done today using technology, infrastructure, and methods which are well-known and proven.

      What tech cannot change is culture, namely, the car culture of the US.

      The easiest way to reduce traffic fatalities is having less road traffic. Have you considered that perhaps the much lower traffic fatalities in Netherlands was due to more people using public transportation to go to work, and hence reducing the miles driven?

      https://en.wikipedia.org/wiki/... [wikipedia.org]

      Yep, using per-billion-km-driven metric, the US/Netherlands ratio was reduced from 11.8/3.8=3.1 to 7.3/4.7=1.55. So half of the difference was simply

    • So you are suggesting that all we need to do is rebuild all our cities to be denser so that public transit will be a reasonable alternative to driving. That is much, much harder than making self-driving cars. It's probably not even possible without significantly revising our environmental laws and adding a healthy dose of tort reform (though we should probably do that anyway, the best time to plant a tree is 10 years ago and all that). Even if that is politically achievable, it would cost tens, if not hundr

      • Re: (Score:2)

        by flink ( 18449 )

        We don't need to rebuild cities. We just need to rebuild the infrastructure that the oil, car, and tire companies conspired to buy up and destroy [wikipedia.org].

        We need to abandon the idea that regional and light rail needs to turn a profit or break even. We don't apply that bar to public roads, so there is no reason that it should apply to public transit.

  • Real solution (Score:4, Insightful)

    by Retired Chemist ( 5039029 ) on Sunday January 29, 2023 @03:23PM (#63249269)
    Create large fines for driverless vehicles that break the law. Say $100,000 per incident. The problem would go away or the robotaxis would.

    • Incorporate every vehicle as its own operating company, with a total worth less than $100,000. If fined, the corporation is immediately insolvent. The assets are then auctioned off to another single-vehicle operating company for less than purchase price.

  • The one human involved may learn but the rest make the same mistakes time and again.

    When they fix the bug in the car software, all cars in the fleet are fixed.

    It seems to me that automated taxis are still the better solution. He sooner we get rid of human drivers the better.

    • When they fix the bug in the car software, all cars in the fleet...

      have a new bug.

    • Re: (Score:3)

      by HiThere ( 15173 )

      Sorry, but that's an argument for a slow, staged, roll-out. Get it working properly in one set of circumstances, and then expand it SLIGHTLY. Repeat until done.

      • The issue is they already work and need additional training that will improve faster based on mass economy. The sooner they proliferate, the sooner they far surpass human drivers.

        The additional argument that robotaxis will always receive service because they are members of organized managed fleets seems like a major plus as well

        • The issue is they already work and need additional training that will improve faster based on mass economy.

          Apparently they don't already work in a large number of situations.

          There's no reason to fanboy about these things. If they work, put them on the streets. If they don't, then improve the technology.

  • I called Kaiser's emergency and advice Nurse at 1am to discuss some symptoms that could have been related to a stroke. I satisfied myself that I was fine, and told the nurse I was done and wanted to hang up and go to bed. She refused to let me go saying she wanted to consult with a specialist. There aren't many specialists on duty at that hour, so while waiting on hold for another 10+ minutes, I did fall asleep. She called 911 to resolve the "emergency". They showed up.

    Fortunately I had a guest visiti

    • When you call the medical line, the people on the phone have a liability to address the problem fully. If you had a stroke, you could have sued Kaiser for not adequately addressing your concern. From her point of view, you could have been on drugs and not stay awake. From your story, it looks like she was really worried about someone whom she knew nothing about. Where I live, they would not have cared if I had a stroke, as long as I got off the phone and left them alone. It's 1am for them too. I hope

      • Re: (Score:2)

        by lpq ( 583377 )

        If it is 1am for them, they are likely working on a shift where their body clock says 1pm or such. I told nurse I was fine, and that I needed to get off the phone and not be on hold, she though that I should be forced to wait on hold, when I told her that I was falling asleep as I'd been up over 48 hours and that I didn't want to wait on hold. Told her that I was on m y way to falling asleep (they would have had a recording of that as they record such calls (for training purposes, of course). When she re

  • There is surely a directive in the software that says "when in doubt, stop". Seems logical at first sight, until you consider that you are not the only vehicle on the road.

    And it seems difficult to change, except by incrementally reducing the "doubt" situations, which will take time.

    • And it seems difficult to change, except by incrementally reducing the "doubt" situations, which will take time.

      If the cars were controlled by carefully engineered expert systems, this would be true. Incremental improvements would reduce doubt over time. The bug-space would slowly shrink.

      Using trained AI, as they all are, there is no guarantee that this will take place. Reducing doubt in one area might increase it in another; especially if the system resources haven't changed. And there is no realistic way to prevent regressions, because the internal processes are not knowable, and not immutable over time, and so the

    • Except it doesn't always stop when it should [instagram.com]. I have also personally witnessed Waymo blocking crosswalks, which means they should have stopped a bit earlier.

    • > There is surely a directive in the software that says "when in doubt, stop"

      And yet, human drivers are taught "stop somewhere safe".

      As someone noted above, robocars stopping dangerously should be fined as a human might be for similar offences. Self-driving is hard, if the companies doing it don't like it, then get out of the way and let someone else have a crack at it.

  • Okay I made up the number but it's up there. For just about anything that interacts with the real world in some way, the "core functionality" is the easiest part to design and implement and the "interaction with the real world" is the hardest and most time consuming to get right.

    When I was in grad school, I wrote about 150k lines of code as part of my dissertation. Of those, maybe 5k was doing the calculations I actually needed to do, and the other 145k was ingesting and processing real world data from not

    • This is the problem with "percentages" in general.

      Let's say your software reliability is measured in uptime.

      We code for 9 days:
      1min of downtime every 10 minutes is 90% reliability.
      Great just need to code for 1 more day to get the last 10%.

      90 days later:
      1 min of downtime every 100 minutes. Wait now it's 99% reliable... We spent 10x as long coding and we're less than 10% closer to 100% reliable.

      3 years later...
      1 min of downtime every 1,000 minutes. WTF!? we only picked up 1% of reliable even though we put i

      • One definiton of chaotic systems is that n times the information now only increases your understanding* by log(n) or worse. Life is a chaotic system and information in software is roughly proportional to the amount of time spent writing it.

        *Understanding can be quantified as a prediction horizon for the time evolution of something or the amount of cases covered as in your hypothetical. Point is that magical computation can't fix the fundamental uncertainty of the real world.

  • I'd rather regulate a required method of forcing the automated car to move as directed. Remote control input, like an RC car. Forward, left, right, etc...

    Then add a way to direct the car to relocate to where directed. Sort of a smarter version of the above, where they can ignore the precise steps if the system is mostly functional.

    Then create a truly public and reliable electronic ID system. Otherwise people will hack into these capabilities.

    I'm amazed I've heard of zero credible internet capable govern

  • And that way is to have dedicated roads for robots vs humans. Unless and until there's a general purpose AI that matches human sense processing and response, the needs of a robo-road are different from those of a human road. And from a cost benefit perspective, there's only certain routes that make sense to automate anyway (until you reach the long tail, decades from now, when most people don't expect to be driving). Build a few dedicated interstates with robot-friendly signage and lane markers.

    • Sounds like a great plan for San Francisco. With the endless surplus land we have here in the second most densely populated city in the USA this should be easy to try out.

      • Re: (Score:2)

        by larwe ( 858929 )
        For the literally dozens of people who live and work in places that are not San Francisco, your sarcasm falls flat. Simply because a fair number of tech companies are HQd in the area doesn't mean it's the right place to test or even use that tech.

    • Re: (Score:2)

      by flink ( 18449 )

      Yeah, and to make navigation an easier problem, we could have the wheels of the vehicle engage with the road. We could call these "rails". And since it's a dedicated right of way with limited access, it makes sense to share one prime mover between multiple cars. Since there are multiple cars entrained behind a lead vehicle, we could call such an arrangement a "train". And for efficiency's sake, let's keep "trains" assembled and run them on a regular schedule, that way we could have centralized automated

      • Re: (Score:2)

        by larwe ( 858929 )
        Snarkasm aside - the difference is that a road vehicle can go the last mile to wherever goods or people actually need to be - a train needs a station, and then road transportation from that station. So in that sense a road for robo-vehicles only make some sense - it's much cheaper to build an off-ramp to local roads than an entire train station. In a sociological sense of "people want personal transportation, not shared transportation" sense, the robo-road also wins, to some degree (though I feel most of th

  • How can these cars not have remote control? Robotaxis should be remotely human monitored at all times. One person can manage about 10 cars at a time.

  • Comment removed based on user account deletion
  • The GPS satellites will come down in the first minutes of a modern war.

  • Hindquarter pounding (Score:3)

    by NagrothAgain ( 4130865 ) on Sunday January 29, 2023 @09:15PM (#63249883)

    causing rear-end collisions with hard braking...

    So when it's a human in front it's the fault of the other driver for "following too close" but when it's driverless suddenly it's not?

  • a city bus with 45 riders couldn't get through and was delayed for at least 13 minutes.

    Never in the history of San Francisco has a human driver done anything boneheaded that blocked traffic for an extended period of time.

  • Instead of cars, automation needs to be applied to delivery trucks, such as semi-tractors, and then allow them only in good weather and from say 2200 until 0600. IOW, in the middle of the night when roads are not heavily used. This is far better for starting true robotic's driving.

  • It's OK to being careful with rollout of robotaxi's, but it's better to supervise the rollout and fixing of 'bugs'. So make sure the companies putting out these robotaxi's can explain why the cars stopped, and if there is fix for the unnecessary stops within a certain amount of time.
    Let's not forget, human drivers also stop for no reasons in the middle of the road and also block firefighters or busses, so don't act as if it's something new and extra due to robotaxi's. I even think robotaxi's will be able to

  • Any robocar that's just idling in the road, haul off and crush, and charge the owner of record for that.

Slashdot Top Deals

Some people manage by the book, even though they don't know who wrote the book or even what book.

Close