6 ms·

I'm not sure what you're purporting to show other than two people disagreeing with each other, without providing evidence... Sure, an individual subway car has

I'm not sure what you're purporting to show other than two people disagreeing with each other, without providing evidence... Sure, an individual subway car has "more capacity," but does a subway system have more capacity? For example, I rarely see BART trains come within five minutes of each other. A tunnel that allows cars within 10 seconds of each other could provide more capacity than a tunnel with higher-density trains spaced further apart. Average density across the tunnel system could be higher (aka more throughout) with autonomous vehicles even if individual transport units have lower density (SUVs instead of train cars). Off the top of my head, here are issues train cars face with managing density that autonomous vehicles wouldn't face: 1. Needing to stop at stations where few people are getting on or off, meaning that trains behind you must be at least (MAX_SPEED * STOP_TIME) behind you. With cheap tunnel boring + AVs, you can quickly and cheaply build tunnel branches that skip exits, so individual cars need less space between them (you branch off to exit; no need for extra rails or switching technology, since it's built into the cars). 2. Entrance/exit from the train car: with trains, any jerk can run in as the doors are closing, preventing the train from leaving, backing up the trains behind you and leaving a huge gap with trains in front of you. With AVs, you're already in the car. 3. Trains have fairly long stopping distances, increasing the distance needed between trains for safe operation. Some quick Googling indicated that an 8-car passenger train traveling at 80mph can take over a mile to come to a complete stop — it's hard to imagine a car with such long braking distances. That puts a pretty hard lower bound on how densely you can pack an ordinary subway tunnel. Maybe I'm totally wrong, but it would be more useful to give actual data and evidence than knee-jerk shutdowns. It's not obvious to me, a non-train-engineer (like many on this site!), that the dismissive tweet is particularly meaningful.

Tiktaalik 8y ago
This is already a solved problem. BART may have huge headways, but Vancouver's fully automated Skytrain can run at 75 second headways (!!!). There is no universe where an SUV holding 4 people is more space efficient than a Skytrain car.
- reissbaker 8y ago
  75 seconds is really, really long. Cars on a highway don't leave 75 seconds in between each other; if a cutting-edge train does, that starts to imply that density issues do exist for subway systems. And how expensive was it to build the train? If you can build 3x the tunnels for AVs as you can train tunnels and trains for the same price, actual throughput dollar-for-dollar in infra spend would be much better with Boring Co tunnels. Update with data: it cost $2 billion to build 50 miles of Skytrain. Apparently this prototype Boring Co tunnel cost $10 million per mile. That's a 4x cost advantage for Boring Co mile-for-mile, and I imagine they'll try to strengthen that advantage even further; after all, that's their whole company's purpose: cheap, efficient boring.
  - endless1234 8y ago
    >75 seconds is really, really long. Cars on a highway don't leave 75 seconds in between each other ..and? A car also doesn't carry 500 people.
    - steve_musk 8y ago
      But 75 cars spaced 1 second apart...
      - dmitriid 8y ago
        Somehow you assume they will be able to get into that tunnel at these intervals
        Tiktaalik 8y ago
        haha yep. This plan is like when a city widens a highway but doesn't touch other parts of the road network? What happens? It just moves the choke point somewhere else in the network.
        reissbaker 8y ago
liotier 8y ago
The equivalent of BART in Paris (RER) has 90 seconds intervals between trains in peak time - and thats the regional train system, not the metro. It carries 50k pax per hour in each direction - that would be a lot of cars, like eleven cars a second at 1.2 person occupancy. You mention "cheap tunnel boring" - maybe that is the keystone of Musk's plan that will change the paradigm... With a little knowledge of tunnel boring I fail to see how to disrupt its crazy complexity, especially in urban environments whose density extends underground already: in my neighborhood (La Défense business district) a whole new station is being build wholly under existing construction and interconnected to existing subterranean works, without any disruption - insanely good engineering and insanely expensive.
- kwhitefoot 8y ago
  One of Elon Musk's ideas is that underground you can have a three dimensional system of tunnels and that this can help solve some of the capacity problems.
  - liotier 8y ago
    As dense and as cheap as Musk's technological vision becomes, a question remains: why carry a car at all instead of carrynig people ? Why carry mostly dead weight ? The denser the network, the more acute that question - if Musk's vision is an ubiquitous three-dimensional underground transport mesh, then why does it seem to conscientiously avoid being a public transportation system ? Does Musk believe that saturating the city with autonomous vehicles that use tunnels as network backbone can be more efficient than heavy public transportation networks ?
    - noobermin 8y ago
      Because Elon owns a car company. That's fundamentally why.
mjg59 8y ago
Peak headtimes on BART are a train every 2-3 minutes, with trains carrying up to 1400 people (let's call it 1000 for the sake of argument). So at the low end, that's around 20,000 people an hour. If you had cars with a headway of 5 seconds then that's 1200 cars an hour, so you'd need an average occupancy of over 16 people per car. That's unrealistic, so yes, a subway tunnel has more capacity.
- zimpenfish 8y ago
  cf London Underground - Central Line does ~260M journeys a year which is ~40k/hour assuming an 18 hour day. (I suspect peak times will be ~50% higher than that average.) Even the Waterloo & City Line, the most unloved of all lines, manages 3.5k/hr over 260 days (16M/year, 5 day week).
- reissbaker 8y ago
  That's assuming there's only one tunnel. As the article mentions the point is to dramatically reduce the cost to bore tunnels, so you can bore more of them. If you could bore 4x the tunnels, you'd need a capacity of 4 people per car to match peak BART. And... hey, that describes pretty much every car in existence. But using the BART comparison: recent estimates [1] put each additional mile of BART track at over $1 billion. So if it really costs $10MM/mile for the Boring Co, you could bore 32x the tunnels and beat BART capacity by a factor of two assuming even single occupancy vehicles, at about a third of the cost. [1]: https://sf.curbed.com/2018/6/18/17464616/bay-area-subway-train-rail-costs-price-bart-muni https://sf.curbed.com/2018/6/18/17464616/bay-area-subway-tra...
rsynnott 8y ago
> Sure, an individual subway car has "more capacity," but does a subway system have more capacity? For example, I rarely see BART trains come within five minutes of each other. The BART is a particularly poor example. There are _tram_ systems (ie partially-segregated rail) which do better than once every five minutes.

Tiktaalik 8y ago

This is already a solved problem. BART may have huge headways, but Vancouver's fully automated Skytrain can run at 75 second headways (!!!). There is no universe where an SUV holding 4 people is more space efficient than a Skytrain car.

reissbaker 8y ago
75 seconds is really, really long. Cars on a highway don't leave 75 seconds in between each other; if a cutting-edge train does, that starts to imply that density issues do exist for subway systems. And how expensive was it to build the train? If you can build 3x the tunnels for AVs as you can train tunnels and trains for the same price, actual throughput dollar-for-dollar in infra spend would be much better with Boring Co tunnels. Update with data: it cost $2 billion to build 50 miles of Skytrain. Apparently this prototype Boring Co tunnel cost $10 million per mile. That's a 4x cost advantage for Boring Co mile-for-mile, and I imagine they'll try to strengthen that advantage even further; after all, that's their whole company's purpose: cheap, efficient boring.
- endless1234 8y ago
  >75 seconds is really, really long. Cars on a highway don't leave 75 seconds in between each other ..and? A car also doesn't carry 500 people.
  - steve_musk 8y ago
    But 75 cars spaced 1 second apart...
    - dmitriid 8y ago
      Somehow you assume they will be able to get into that tunnel at these intervals
      - Tiktaalik 8y ago
        haha yep. This plan is like when a city widens a highway but doesn't touch other parts of the road network? What happens? It just moves the choke point somewhere else in the network.
        reissbaker 8y ago

liotier 8y ago

The equivalent of BART in Paris (RER) has 90 seconds intervals between trains in peak time - and thats the regional train system, not the metro. It carries 50k pax per hour in each direction - that would be a lot of cars, like eleven cars a second at 1.2 person occupancy. You mention "cheap tunnel boring" - maybe that is the keystone of Musk's plan that will change the paradigm... With a little knowledge of tunnel boring I fail to see how to disrupt its crazy complexity, especially in urban environments whose density extends underground already: in my neighborhood (La Défense business district) a whole new station is being build wholly under existing construction and interconnected to existing subterranean works, without any disruption - insanely good engineering and insanely expensive.

kwhitefoot 8y ago
One of Elon Musk's ideas is that underground you can have a three dimensional system of tunnels and that this can help solve some of the capacity problems.
- liotier 8y ago
  As dense and as cheap as Musk's technological vision becomes, a question remains: why carry a car at all instead of carrynig people ? Why carry mostly dead weight ? The denser the network, the more acute that question - if Musk's vision is an ubiquitous three-dimensional underground transport mesh, then why does it seem to conscientiously avoid being a public transportation system ? Does Musk believe that saturating the city with autonomous vehicles that use tunnels as network backbone can be more efficient than heavy public transportation networks ?
  - noobermin 8y ago
    Because Elon owns a car company. That's fundamentally why.

mjg59 8y ago

Peak headtimes on BART are a train every 2-3 minutes, with trains carrying up to 1400 people (let's call it 1000 for the sake of argument). So at the low end, that's around 20,000 people an hour. If you had cars with a headway of 5 seconds then that's 1200 cars an hour, so you'd need an average occupancy of over 16 people per car. That's unrealistic, so yes, a subway tunnel has more capacity.

zimpenfish 8y ago
cf London Underground - Central Line does ~260M journeys a year which is ~40k/hour assuming an 18 hour day. (I suspect peak times will be ~50% higher than that average.) Even the Waterloo & City Line, the most unloved of all lines, manages 3.5k/hr over 260 days (16M/year, 5 day week).
reissbaker 8y ago
That's assuming there's only one tunnel. As the article mentions the point is to dramatically reduce the cost to bore tunnels, so you can bore more of them. If you could bore 4x the tunnels, you'd need a capacity of 4 people per car to match peak BART. And... hey, that describes pretty much every car in existence. But using the BART comparison: recent estimates [1] put each additional mile of BART track at over $1 billion. So if it really costs $10MM/mile for the Boring Co, you could bore 32x the tunnels and beat BART capacity by a factor of two assuming even single occupancy vehicles, at about a third of the cost. [1]: https://sf.curbed.com/2018/6/18/17464616/bay-area-subway-train-rail-costs-price-bart-muni https://sf.curbed.com/2018/6/18/17464616/bay-area-subway-tra...

rsynnott 8y ago

> Sure, an individual subway car has "more capacity," but does a subway system have more capacity? For example, I rarely see BART trains come within five minutes of each other. The BART is a particularly poor example. There are _tram_ systems (ie partially-segregated rail) which do better than once every five minutes.