On high capacity systems, train dwell time becomes the limiting factor on passenger capacity. 30 seconds is generally the minimum possible dwell time a system can manage, 20 seconds might be possible during much lower demand periods. But you’re unlikely to ever achieve better than that.
The London Victoria Line, which runs with 90 second headways at peak, achieves at best 24 second dwell times in central section, but 30-40 seconds is more realistic for most stations.
Don’t forget, dwell time includes more than just passengers getting off and on. It has to include time to open the doors, close the doors (including a 2-3 second visual and audible warning!), perform needed safety checks, and eventually pull away. Those operational components the sandwich the core “people getting on and off” bit of station stops add up to a non-trivial number of seconds.
As a minor note, the NYC subway uses a system of fix automatic blocks, not a moving block system, which is what your simulation appears to use. Moving block systems are dramatically more efficient than fixed block systems. But I have no idea how you would get hold of accurate block locations for the NYC subway.
There are door opening warning in the Philadelphia subways and metros (PATCO) as well. It makes sense if the train is crowded and a rider is smushed up against the opening door.
This is awesome—a sandbox for transit nerds lol my husband will love this. The dwell time + simulation rate sliders are super fun, can do a lot with congestion indicators or bottleneck flags when trains start stacking up or something... fun to stress test bad routing decisions.
I can't figure out how to add a route from Bay Ridge to Clinton in NJ. That's my wishlist subway - got to Newark Airport from Brooklyn without going through Manhattan.
Edit: Actually Clifton is in NY, and playing with Google maps there's ZERO public transport from Staten Island to NJ, except by going through Manhattan!
So my idea wouldn't help anyway, unless they extended that subway all the way over the Goethals Bridge.
As a Brooklynite I wholeheartedly agree. Unfortunately this is more of a service designer as opposed to a track designer, and it only works with existing infrastructure :(
Looks like a 2 or 3 mile walk from the nearest public transport stops that I can find. Maybe someone local can find a closer stop.
Forest Ave/Goethals Rd North to Bayway Ave at Mckinley St.
Edit: This could be a fun game - find the two spots with the greatest ratio of driving vs public transport. I think those two are pretty good candidates.
Distance: 2.5 mile.
Drive 9AM: 5-9 minutes, 11PM: 4-6 minutes.
Public transport 9AM: 1:52, 11PM: 3:34.
Ratio (using the average drive time) 9AM: 16, 11PM: 42.8.
Another game: when is public transport slower than walking?
According to Google Maps it's not in this case - I'm getting a walking time of 14:37 (!), where the "walk" includes a ferry from Weehawken to Manhattan and the Staten Island Ferry. However Google Maps doesn't seem to know you can walk across the Goethals bridge.
It's not the bridge, the starting point seems to be on a freeway or something, if you move the starting point slightly toward the bridge it will show proper walking directions.
There are 4 bridges: Goethals, Outerbridge, Bayonne, and Verrazano. You can walk on Goethals and Bayonne. I was actually thinking a few weeks ago about walking from Elizabeth to Bayonne someday, just for the heck of it.
I want to love this but the visual language makes it kind of unusable for me. Why not match the track and train colors to their line color (red for the 123) and then use different visual indicators for train state (stopped, at station, etc)?
For example:
Selected: Black fill
Normal operation: Color fill with 100% opacity
Slowing down: 70% opacity
Stopped: striped fill, 50% opacity
At station: pulsing opacity
Sometimes when I Edit Routes and click an "<- Add" button I get the console error "Uncaught Error: coordinates must be an array of two or more positions" and the page blanks out.
I just tested it out. Since this uses OpenStreetMap data, it is very easy to adapt to other cities.
All input data and scripts are in src/data. Run the two Overpass queries, replacing New York with a city of your choice, and you will get JSON data on the infrastructure, as well as services. Replace the two existing JSON files, run the two transformation scripts, and you are done.
That is actually the idea. In this simulator, you design services that run on existing tracks. NYC subway is unique in the world in that there are a ton of interlining as well as quadruple tracks, so there are plenty of opportunity to improve system efficiency without building anything new.
This is sadly not mobile friendly in an amusing way; the map controls are nice and legible but they block basically the entire map. If there was a way to minimize them (perhaps automatically), this might work well on mobile.
Your default dwell times are wayyyyy too short.
On high capacity systems, train dwell time becomes the limiting factor on passenger capacity. 30 seconds is generally the minimum possible dwell time a system can manage, 20 seconds might be possible during much lower demand periods. But you’re unlikely to ever achieve better than that.
The London Victoria Line, which runs with 90 second headways at peak, achieves at best 24 second dwell times in central section, but 30-40 seconds is more realistic for most stations.
Don’t forget, dwell time includes more than just passengers getting off and on. It has to include time to open the doors, close the doors (including a 2-3 second visual and audible warning!), perform needed safety checks, and eventually pull away. Those operational components the sandwich the core “people getting on and off” bit of station stops add up to a non-trivial number of seconds.
https://www.researchgate.net/figure/Dwell-Time-and-Passenger...
As a minor note, the NYC subway uses a system of fix automatic blocks, not a moving block system, which is what your simulation appears to use. Moving block systems are dramatically more efficient than fixed block systems. But I have no idea how you would get hold of accurate block locations for the NYC subway.
> NYC subway uses a system of fix automatic blocks, not a moving block system
Lines with CBTC use moving blocks. That’s the L, 7, and parts of the E, F, M, and R.
Several of those dwell components can be done in parallel.
You don’t need a warning for opening the doors. The warning for closing the doors will overlap with people boarding.
That being said 10s is too low as a default but it can easily be changed to a more reasonable 30s or more.
> You don’t need a warning for opening the doors.
Seems like the Swiss railway didn't get your message because there is acoustic warning on open nd on close.
There are door opening warning in the Philadelphia subways and metros (PATCO) as well. It makes sense if the train is crowded and a rider is smushed up against the opening door.
How do you know all this?
I contract out work to MTA, specifically their AI/ innovation teams.
I'm 100% showing this to them today just for fun. They'll get a kick out of it.
I'm super curious to learn more about what AI/Innovation looks like for the MTA.
Is any of your/their work published?
This is awesome—a sandbox for transit nerds lol my husband will love this. The dwell time + simulation rate sliders are super fun, can do a lot with congestion indicators or bottleneck flags when trains start stacking up or something... fun to stress test bad routing decisions.
I can't figure out how to add a route from Bay Ridge to Clinton in NJ. That's my wishlist subway - got to Newark Airport from Brooklyn without going through Manhattan.
Edit: Actually Clifton is in NY, and playing with Google maps there's ZERO public transport from Staten Island to NJ, except by going through Manhattan!
So my idea wouldn't help anyway, unless they extended that subway all the way over the Goethals Bridge.
As a Brooklynite I wholeheartedly agree. Unfortunately this is more of a service designer as opposed to a track designer, and it only works with existing infrastructure :(
Looks like a 2 or 3 mile walk from the nearest public transport stops that I can find. Maybe someone local can find a closer stop.
Forest Ave/Goethals Rd North to Bayway Ave at Mckinley St.
Edit: This could be a fun game - find the two spots with the greatest ratio of driving vs public transport. I think those two are pretty good candidates.
Distance: 2.5 mile.
Drive 9AM: 5-9 minutes, 11PM: 4-6 minutes.
Public transport 9AM: 1:52, 11PM: 3:34.
Ratio (using the average drive time) 9AM: 16, 11PM: 42.8.
https://maps.app.goo.gl/WU1jzjYh8tkTCuh87
Another game: when is public transport slower than walking?
According to Google Maps it's not in this case - I'm getting a walking time of 14:37 (!), where the "walk" includes a ferry from Weehawken to Manhattan and the Staten Island Ferry. However Google Maps doesn't seem to know you can walk across the Goethals bridge.
It's not the bridge, the starting point seems to be on a freeway or something, if you move the starting point slightly toward the bridge it will show proper walking directions.
There are 4 bridges: Goethals, Outerbridge, Bayonne, and Verrazano. You can walk on Goethals and Bayonne. I was actually thinking a few weeks ago about walking from Elizabeth to Bayonne someday, just for the heck of it.
I forgot about Verrazano! I was only thinking about routes to NJ.
I want to love this but the visual language makes it kind of unusable for me. Why not match the track and train colors to their line color (red for the 123) and then use different visual indicators for train state (stopped, at station, etc)?
For example: Selected: Black fill Normal operation: Color fill with 100% opacity Slowing down: 70% opacity Stopped: striped fill, 50% opacity At station: pulsing opacity
That is a good idea. I just added color for the tracks. I'm in the process of redesigning how trains look like.
Very cool.
Sometimes when I Edit Routes and click an "<- Add" button I get the console error "Uncaught Error: coordinates must be an array of two or more positions" and the page blanks out.
Thanks - this is likely due to adding two tracks facing each other, i.e.
-->-->-- --<--<--
I'll add some logic to prevent such options from showing up.
Finally, I will make a subway that crosses the park!
92nd and Broadway, CPW, 5th Ave, Lex, 2nd Avenue, follow Astoria Blvd to the Grand Central to LGA is a no-brainer.
Very cool! It would be nice to have a bit more information in the readme about the project structure and e.g. how to adapt it for other cities :)
I just tested it out. Since this uses OpenStreetMap data, it is very easy to adapt to other cities.
All input data and scripts are in src/data. Run the two Overpass queries, replacing New York with a city of your choice, and you will get JSON data on the infrastructure, as well as services. Replace the two existing JSON files, run the two transformation scripts, and you are done.
What's the point of designing custom routes, except I guess for fun? NYC is never going to build any new routes.
That is actually the idea. In this simulator, you design services that run on existing tracks. NYC subway is unique in the world in that there are a ton of interlining as well as quadruple tracks, so there are plenty of opportunity to improve system efficiency without building anything new.
I love it! Amazing work.
A slider to do a bit of time-travelling if possible would be also a nice to have
Very nice! I thought of doing the same thing in the past!
you need to model passenger behavior at rush hour in the face of train cars with a truly foul smells and puddles on the floor!
This is a very cool project!
This is sadly not mobile friendly in an amusing way; the map controls are nice and legible but they block basically the entire map. If there was a way to minimize them (perhaps automatically), this might work well on mobile.