• vocornflakes@lemmy.world
    link
    fedilink
    English
    arrow-up
    23
    ·
    1 year ago

    I read in a book that the current system of drivers acting on their own without something coordinating their every move is actually 75% as efficient as a fully coordinated system.

    Therefore, the benefit obtained with all people using self driving cars is nothing compared to just improving public transit or improving car infrastructure.

    • duffman@lemmy.world
      link
      fedilink
      arrow-up
      20
      ·
      1 year ago

      I don’t know what book that was or what metrics its using, but my local intersections could easily pass 3x the current number of cars per green light if they accelerated together, and right away.

      The number of people who poorly merge and cause traffic shockwaves, how slow cars drive in the fast lane, the accidents caused by human error. Really curious how they came to that 75% number.

      • vocornflakes@lemmy.world
        link
        fedilink
        English
        arrow-up
        7
        ·
        edit-2
        1 year ago

        I was slightly wrong. From page 237 of Algorithms to Live By, The Computer Science of Human Decisions by Brian Christian and Tom Griffiths, further referencing the paper How Bad is Selfish Routing? by Roughgarden and Tardos, it says that

        “…the “selfish routing” approach [of cars] has a price of anarchy that’s a mere 4/3. That is, a free-for-all is only 33% worse than perfect top-down coordination.”

        Anyways, the way they got to that number is mathematical game theory. In this case people will choose the fastest route which happens to not be so bad.

        It’s also very possible that what they’re concluding is significantly abstracted, but I haven’t read the source reference to know for sure.

        • frezik@midwest.social
          link
          fedilink
          arrow-up
          5
          ·
          1 year ago

          Just the number of people being moved on a bus or light rail for a given amount of space tosses that efficiently number away.

          • vocornflakes@lemmy.world
            link
            fedilink
            English
            arrow-up
            3
            ·
            1 year ago

            Exactly. The point it was making is that perfect top-down coordination takes a ton of resources for a whole lotta nothing.

    • platypus_plumba@lemmy.world
      link
      fedilink
      arrow-up
      4
      ·
      1 year ago

      I don’t think cars are ever going away, even if public is the main transportation method. Which obviously sucks, but it’s the way it is.

      I’ve always imagined a protocol that lets cars communicate their planned speed. I’m pretty sure this is how cars will work in the future. A decentralized mesh of coordinated vehicles. This means that cars can:

      • Maximize constant speed time, improving energy consumption and traffic flow.
      • Minimize distance between vehicles based on speed and acxeleration while complying with safety standards.
      • Connect to devices such as semaphores in order to tell if the vehicle will pass or not, to make a better decision.
      • Connect to other mesh devices such as AI cameras that feed events to the vehicle mesh.

      Public is obviously the best option though. Imagine a city with no streets, only subterranean public transportation. You wouldn’t even need such a large public transportation system, cities would be a fraction of the current size. I wonder what percentage of the area of a city is wasted on streets.

      • duffman@lemmy.world
        link
        fedilink
        arrow-up
        2
        ·
        1 year ago

        In the 90s in school, I did a report and imagined computers would be too expensive to have in every car, so the road itself would have wireless infrastructure to control the cars.

      • phx@lemmy.ca
        link
        fedilink
        arrow-up
        1
        ·
        1 year ago

        A hybrid system would be cool. I could see a future where electric vehicles could link up to a pod like train cars for long trips along standard routes, and schedule automated disembarkation for their “stops” to continue the rest of the way to their destination. Full autonomous driving is a difficult problem be a lane pods of this nature could be quite efficient and easier to automate