Anecdotally: last year I was riding in a train that hit someone who fell onto the tracks because the platform was so crowded. 30 minutes later, one station away, a woman lost both her legs when a train ran them over.

When I lived in Tokyo, they put platform guard walls in at many stations. They were about chest high and had sets of doors that would open before the train doors opened, and if we had them in NYC, both of those folks would have made it home without an incident.

In London, they simply close the station entrances (at street level) before it becomes so overcrowded. It can be a daily occurrence, e.g. at rush hour during a month of upgrade work on a particular line.

There is no record of anyone being killed or severely injured through overcrowding.

Jubilee also has station doors at new stations.

There are numerous examples on the Tokyo Metro/Toei. The cheaper ones go up to about chest height and the more expensive ones completely block off the tracks by going up to the ceiling.

They also have the added benefit of showing people where to line up and make it harder for someone to cause the train to be late by rushing/holding the door

The MTA is currently working on a solution that doesn't use barriers. [1] The short of it is that it will use thermal cameras and other sensors to determine if a human (and not a rat, for instance) has entered the tracks and is in danger.

It won't kill the trains but slow them down considerably and of course if it's a last second jumper or someone who falls right as the train is entering the station it won't work. But I imagine it will help solve a majority of the issues.

If you're nervous for your safety then I would recommend to stand on the far end of the platform where the front of the train stops at the station.

[1] http://www.nydailynews.com/new-york/mta-testing-systems-save...

have you ever seen the tube system in London? it's a fully blocking barrier with a set of sliding doors that lines up precisely with the doors of the train cars when they pull in to the station.

that's probably the ideal solution, but not practical for New York because the system was never designed with precision stopping points on the platform in mind. there's a margin of error that is at least a meter or so in either direction right now.

Only on new lines with straight platforms. Older ones are almost all on curves, a guy in the sliding doors business told me they looked into but the cost was too high as each station is on a different circumference bend so they would all be custom jobs.

Only on the newest part of the newest line, the Jubilee Line Extension. (1999? something like that.)

There are no "platform edge curtain" doors on other lines, even though many other lines also have computer controlled trains. (The "driver" is still there to press a button to start the train, but many lines almost drive themselves.) Although, even on the fully manual lines, the driver is expected to stop the train precisely -- the platforms aren't long enough for inaccuracy.

They seemed common in Japan when I visited:

https://alicegordenker.wordpress.com/2012/01/17/subway-safet...

(not my article)

Something like this maybe? http://philly.curbed.com/archives/2014/10/23/korean-company-...

They investigated installing them here (Toronto) but the cost per station is pretty high, on the order of >$1M per platform. When you multiply that across a system as large as NYC that's easily half a billion dollars.

I'm not sure how many ads you have to sell to recoup the cost, but it's probably a lot.

edit - thought we were still talking about a simple low wall or gate, not the floor to ceiling glass in the GP post. leaving below for posterity..

> on the order of >$1M per platform

That's insanity! Over $1M for something every amusement park has for every ride? Lemme guess, these one have to be engineered to the point of 'impossible to fail' and installed by the employees of municipal cronies at 3x fair wages?

Talk about perfect being the enemy of good...

Insanity? That's dirt cheap. Think about how much mechanical complexity goes into each door, how it has multiple motors, servos, linear actuators and such. It has to be safe when closed, and it has to be safe when opened, it can't spontaneously shut in someone's face or pin them in the middle.

It needs multiple sensors that can handle being exposed to a lot of wear and tear. It needs to operate for years on end with, ideally, very little maintenance. It has to endure being buffeted by the forces of the train entering and leaving the station which exerts significant strain on any hinges or track mechanisms.

So you need a pair of these for each set of doors on the train, and if each car has three doors, and you have eight cars, that's 48 doors.

That leaves you with only $20,800 per door. You still need to build the walls themselves and you need to have this installed by the right technicians.

Do the math, don't bitch just because it sounds expensive.

It occurs to one that elevator doors are every bit as complex as what you ecan scribe and have been doing their job for over half a century at a cost of far less than that.

But really, you don't even need a door - a simple gate like on a ride at Disneyland, or heck even the gates at the station entrance would do nicely.

Not many elevators have 48 doors, and those that do cost a shit-ton more than a million dollars.

Sorry, but $20K per door is still ridiculous. This is not new technology, automatically operated doors are used in many different settings. It's pure gov't waste to pay that much. It would be wasteful if it was 1/4 that much.

As an example that most automatic doors don't have to deal with, but is essential for railways:

The door must detect a 1-2cm obstruction (e.g. fingers, child's wrist) and be sufficiently weak that a child can pull their hand/wrist free. The door must not prevent a train from departing if only a hem of a skirt/trousers/coat is trapped. They must be strong enough to prevent anyone falling from the train while it is in motion, and must not be able to be forced open between stations.

Otherwise, approximately one person per year will die.

How much would you charge, to make doors to this specification?

And, some time in the next five years, an accident will happen, somewhere in the world. Your next door will need to prevent that situation, too.

[This is the situation in "wealthier" Europe, where it's pretty difficult to die on a train. Other countries may vary.]

The first part just means they need to have rubber padding. Still expensive since the design it's a one time job and can be reused for trains, trams, buses etc.

It's not that simple. Something different is required depending how the doors close (pneumatic, electric) and the expected use (high / low speed, crowded or not).

If the padding is too squashy, a trapped hand won't be detected and the person could be dragged under the train. Too sensitive, and someone leaning a suitcase on the door prevents the train from leaving.

It's a lot, but it's objectively not ridiculous.

You have no idea how much an elevator costs, do you?

...maybe kick Melania and Barron out of NYC?

I think it's crazy that the MTA hasn't already done this. Not only would it save a lot of lives, but it would also allow the platform to be climate controlled. You have no idea how many times I've sat there underground for 30+ minutes in sub-freezing or 100+ degree heat.

So we also have to pay for new HVAC systems on almost five hundred commercial buildings (that's what they are, even if underground) with no space for them? Cool.

And the advertising potential on BOTH sides of the wall!

Platform doors are generally installed when you install modern signalling (CBTC) onto a line, which much of the NYC subway does not have. CBTC equipped trains can operate under Automatic Train Operation (ATO) meaning that trains can be made to always stop at the same location on the platform.

Furthermore, platform doors require you to have trains that have the doors in the same positions on the car body, which much of the newer subway fleet (R142,R188 & R143,R160,R179) has but the legacy fleet does not. This issue get compounded even more where multiple lines intersect (which is often due to the NYC subway's unique connectedness) requiring that certain stations would need to support 3 or 4 different car classes.

Crazy is too strong a word because the costs involved would be astronomical / extraordinary / insane.

There are gaps or doors that line up with the train doors when the subway stops

So this solution also requires operators able to stop the train at a particular point. Training is expensive!