Hacker News Reader

Comment by jjice | Hacker News Reader

It is a shame, but it's expected, right? How many edges would a well connected graph of the US have (or rather need if we use nodes as passthrough), how much would this cost to build and operate, and how much would these be used compared to planes? I want to say that I would love better train infrastructure as well, and spending some time in New England has shown me what a fraction of a good one could be and I love it. That said, I don't see how it can be done financially possible, with extremely rough overviews in my head.

I'm very ignorant on the topic, so please enlighten me if I'm missing some stuff, which is very possible. I'd honestly love to know.

View on HN

The viable HSR network for the US looks roughly like this:

* Existing NEC (DC to Boston via Baltimore, Philly, NYC, Hartford/Springfield/Worcester or Providence)

* NYC to Montreal and Boston to Toronto, with an interchange at Albany

* Midwest star shape, centered at Chicago, with prongs to Minneapolis, St. Louis, Detroit, Cleveland, Louisville or Cincinnati. The southern Midwest region is kind of hard for me to crayon a network.

* (Not part of the US strictly speaking) Detroit to Quebec via Windsor, Toronto, Ottawa, Montreal. Basically all of Canada's population on one rail line, with easy access to the US via interchanges at three separate points.

* Hooking up the Midwest to the NEC, although I'm somewhat dubious of the viability here.

* Texas Triangle (Houston, Dallas, Austin, San Antonio)

* California: SF, San Jose, Sacramento, LA, San Diego, Las Vegas, Phoenix.

* Maybe viable: Portland-Seattle-Vancouver

* Maybe viable: SEHSR, from Miami to Atlanta, and thence to Midwest (via Nashville) or NEC (via North Carolina's Research Triangle and Richmond).

(Citation: this is largely based on Alon Levy's crayoning, found here: https://pedestrianobservations.com/2021/03/22/high-speed-rai...)

> Existing NEC (DC to Boston via Baltimore, Philly, NYC, Hartford/Springfield/Worcester or Providence)

This is the problem. The real route is Boston - New York - D.C. The others are lower population, lower GDP and not in line with the other three. But if you build the core route and exclude Philadelphia, you'll have an uproar in the Senate.

I was on the train a week ago. Of the three of us sitting at the table who got on at Boston, one was going to Philadelphia, one was going to Wilmington, and one was going to Baltimore.

One of the chief advantages of trains is that you do not have to cater only to point-to-point routes; you can instead serve a combinatorial explosion in routes for comparatively little time: a stop adds only ~5 minutes to a HSR train, and that's the worst-case scenario.

There is no advantage to bypassing Philadelphia: the amount of money it costs to bypass the city is going to be quite large, and you will miss out on all the potential revenue from stopping at the city itself. From my practical experience riding the train, the biggest factor in existing slow trains in Connecticut isn't all the stops it makes in Connecticut, but the incompetence of Metro North's dispatching that keeps a high-speed train plodding along below highway speeds between stations.

(Also, side note: Philly's MSA is larger than Boston's, and Philly provides more Amtrak riders than Boston's 4 stations combined--and one of those isn't even on the NEC!).

> a stop adds only ~5 minutes to a HSR train, and that's the worst-case scenario.

Is that accurate? When I road HSR in Japan, I remember the speed in-city being much lower than in the countryside, so bypassing cities seemed key to maintaining the high speed part.

So the penalty I'm calculating is the stop penalty of slowing from a cruising speed to a dead stop, waiting at the station, and then speeding back up to cruising speed versus maintaining cruising speed the entire distance; you can think of this as the time penalty that the Route 128 stop generates (being on a nice straight track that allows very high speed).

What you're talking about instead is the penalties imposed by a more constrained right-of-way in urban environments. Except when we're talking about Philadelphia (or most of the Northeast in general), you can't really escape from a constrained right-of-way anyways--a quick crayoning suggests you're adding about 20 miles of railway route which, even at 200mph, adds more time than you're saving by itself.

Consider again the cost of bypassing to save time. How many millions of dollars are you willing to save a minute of travel time? It is far cheaper to put that money into straightening the worst curves of legacy inner city track than to spend it bypassing the city altogether, not even accounting for the loss of revenue by not serving what are objectively large cities.

Is there anything existing that avoids having to stop the whole train at each stop? For drop-offs, just detach the back train car at a stop. Everyone who wants to get off just moves to the back train car at each stop. For pick-ups, some relatively high-ish speed coupling to add a car at the front of the train. In this way, the train loses one car and gains one car per stop. And as the train progresses, it also slowly turns around, ready for the return trip. Sidings at each stop can allow the boarding and un-boarding process to happen without tying up the track for through-traffic. Seems like something like this must have been tried at some point. Why is this not a thing?

It would remove all flexibility in capacity between different towns, and require most people to move to a different carriage at least once after boarding — which is especially inconvenient with bags. Even if you couple the carriages at high speed, how do you connect the doors between them sufficiently safely to allow people to move along the train?

The coupling of modern high-speed trains is also a significant part of their crash safety. Keeping the train connected if it derails means there's less chance of a loose vehicle punching through another. (See how e.g. [1] has most of the vehicles pretty much intact.)

I think in most cases, running an express train followed by a stopping train is good enough.

[1] https://assets.publishing.service.gov.uk/media/547c9037ed915...

Old HSR, yes. Maglev? No.

The issue is that steel wheels on steel tracks doesn't provide enough grip to accelerate, so the train takes a lot of the time to get up to speed.

Maglev, meanwhile, uses a linear motor, so you can accelerate very fast if you want to.

I think the bigger constraint would be noise. Wikipedia talks about a 70dba noise limit in residential areas after protests and such, but doesn't say what speed exactly they're running at to meet that.

70dba is rather a lot, though, I imagine you'd end up with a lower limit in the US since I don't think most people are going to be easily convinced that another inter-city transit method is going to be worth the noise to them.

> no advantage to bypassing Philadelphia

Trains spend a lot of time speeding up and slowing down. A single stop adds much more than five minutes to total transit time. [EDIT: Apparently this doesn't apply to maglev. This could be a game changer for HSR in the American political landscape.]

> A single stop adds much more than five minutes to total transit time. [EDIT: Apparently this doesn't apply to maglev. This could be a game changer for HSR in the American political landscape.]

That is false even for non mag-lev. Do the math as I did: at .7 m/s/s (which is what shinkansen does) it takes less than 90 seconds to get to 300 km/h. That leaves far more time than is needed at the station (1 minute is plenty if the train is designed right). You don't count stop time at all because it averages out with the acceleration.

Maglev trains can start and stop much, much faster than regular trains.

Conventional trains' acceleration profiles are already limited by passenger comfort rather than engineering difficulties. Being able to start and stop much faster isn't helpful if it's already outside maximum operating guidelines.

I was under the impression this wasn't the case. Passenger planes accelerate at around 3.5m/s/s. That's 100km/h every six seconds.

And when planes take off, all the passengers and crew are seated in seats with fastened belts, all luggage is stowed, and tray tables are closed.

By contrast, when a train departs a station, there are frequently passengers who are still roaming the aisle with their luggage. And I might have a cup of coffee or something sitting on the table in front of me (or behind me, from the perspective of the train's direction of acceleration).

Of course, you shouldn't have people roam in the aisle if you want to accelerate faster. If you're seriously considering rail as an alternative to airtravel, and at 600km/h you are, then having people stow their luggage and sit down is not a serious issue.

For the sake of comparison - a Toyota Corolla will do 100km/h/6s. You don't need to fully stow your cargo, you just need not to have drinks on the table. Even standing you would be fine, though there is some danger you'd trip.

Beyond that, buses see accelerations on the order of ~0.2G often, and that's with people standing up, in an unpredictable fashion.

The root context of this discussion is basically whether or not it makes sense to add in intermediary stops on a rail line. If you're insisting that people have to be seated and have no drinks out for every intermediate stop, that rather sharply raises the discomfort level of stops. Another unpredictable element of train acceleration is the need to slow down to negotiate a curve, which isn't necessarily the case only in the immediate vicinity of a station.

FWIW, the numbers I generally see bandied about for acceleration on (decent) trains is about 1 m/s^2.

The "intermediate stop" being mentioned was Philadelphia. I don't think asking people to be seated for 45 seconds before and after each stop adds that much discomfort.

0.1g is the acceptable lateral acceleration when standing up without bracing. (https://link.springer.com/article/10.1007/s40864-015-0012-y)

If you get caught standing up, even 3.5 m/s/s is not dangerous. It is uncomfortable, but in a high speed train you are expected to be sitting down.

Also you wouldn't need to put drinks away. As I said, that's the acceleration of a Toyota Corolla. You would need the drinks to be covered though.

The limiting factor in longitudinal acceleration in trains is that we expect people to be standing up for the whole trip. This isn't the case in a high speed train.

As for curves - the solution is not to go through any sharp curves unless you are already close to a stop and would have to slow down anyways.

There's not much point accelerating faster if it means waiting longer so everyone can find a seat.

Perhaps this would be interesting to you, "Passenger Stability Within Moving Railway Vehicles: Limits on Maximum Longitudinal Acceleration".

https://link.springer.com/article/10.1007/s40864-015-0012-y

I have actually read the study and linked to it in another comment.

You don't need to wait for everyone to be seated. You're not going to fall over at 0.2G or 0.3G, it will just be uncomfortable. But you should plan for everyone to be able to find a seat.

> Maglev trains can start and stop much, much faster than regular trains

Ah, wasn't aware of that. Thank you.

Speaking only from my own experience, HSR stops take a lot longer than that. It takes the TGV five minutes to slow down and approach the station! The TGVs I've ridden also do not do fast turnaround at stations. They sit there, like airplanes, while people board for the (obviously longer) trip. Yes, it's faster than an airplane to turnaround (less TSA nonsense, etc.), but not 5 minutes and not like a local or regional train, in my experience.

The whole reason the Houston to Dallas route is workable is that it does not stop at all in between.

I've had this idea of feeder trains that rendezvous with the primary train by pulling along side, matching speed. Then a brief and orderly movement of passengers between feeder & primary train occurs. When complete, the feeder train disengages, slows back down — the primary train continues to its next destination.

Feeder trains bring new passengers on board the primary train. They leave the station a little ahead of the incoming primary train in order to accelerate up to its speed just at the rendezvous point.

Another empty feeder had already rendezvoused with the primary before the station and passengers disembarking were offloaded onto it.

In short, the feeders can be small. The primary train need not stop at all the points in between.

I would love to model this graphically.

TVG is a badly designed system and so it uses far more time to both at the stop and in acceleration. Try looking at shinkansen for an example of what is should be.

One thing to consider is that these routes, if they get proper use, do tend to bring the population density and GDP along with them.

The US has world-class train service for freight. The lower population density of the US than Europe and East Asia makes planes a better investment in the US. Less rail to lay per person server makes a big difference.

The midwest of the US has similar population density to France. The western half of the US with few people (not to mention Alaska) really skew the population density.

By way of Alon Levy's blog (although original credit goes to a Twitter user), here's the TGV overlaid on the Midwest: https://twitter.com/Artifact_Scott/status/137770286794400973...

The major Midwestern cities are remarkably close in distance from Chicago as major French cities are from Paris, and in terms of population counts, the US has consistently larger cities.

Sure, but the problem is what do you do once you get there?

In France, the cities are dense with decent to great transit.

In the US, you have low, flat, sprawling cities with no real transit to speak of. You have to drive.

> "...what do you do once you get there?"

car rental, ridehailing, metro bus/train, taxi, friends/family pickup, etc., just as with air travel.

This is why looking at the current number of aviation passengers on these routes that's often the most valuable thing. It shows where people think the time savings justify hiring a car/taking taxis.

Clearly not having overheads of airport security will probably reduce the minimum distance required for this, but as a first pass it's good data source.

I have long said that we need to work on local transit before we work on HSR, but in truth the two do compliment each other. All cities of any size at all have at least a small amount of useful local transit. Expand the local system and make it useful would be a great investment and HSR will make it more useful.

Most local transit systems are run by people who drive though, and it shows in how bad the majority of the system is.

Same thing with planes except you start way outside the city.

It's interesting that Amazon (who spends a fortune on trucking) has created their own cargo airline and starting buying metal because it works out better for them than paying a trucking company. They make money on fast turnaround and big distribution warehouses close to population centers, then they truck the last mile with their own vans.

Historically, the cheapest ways to ship anything are by sea, by rail, and by truck, in that order. Airplanes are still inordinately more expensive but folks are willing to pay for that quick delivery that trucks and rail simply can't do.

Our only hope is that, at some point, the US can reap the advantages of tech advances elsewhere and go straight to really great trains, rather than having a bunch of meh passenger rail we can't justify replacing.

Kind of like how, supposedly, 4G networks were more widespread in pockets of Africa you'd never expect before they were widespread in the US, because those parts of Africa didn't have 2G/3G infrastructure to depreciate and customers to migrate off before 4G could be deployed.

The problem is everyone is looking to the next advance and so nothing gets built. Eventually you have to commit to something and build it.

Right not politicians are looking at things like hyperloop which maybe can be useful in the future, but right now are power point slides and other hype. Instead we could go with something that works - there is plenty of things much better than we have.

Your comment reminded me of this video illustrating the travelling salesman algorithm on a US map (I love the music):

https://youtu.be/SC5CX8drAtU

Reply on news.ycombinator.com