Scrappy Real-Time Train Tracking Systems

29Oct10

“If it has 80 percent of the features and it costs 10 percent of the money, I’ll take it”- MTA Chairman Jay Walder

In June, the MTA hired a bunch of interns (grad students) to explore cheap ways to bring cellphone service in the subway system. The entire project only cost $30,000.  This shift is part of Jay Walder’s new mindset that “The best is the enemy of the good.”  Similarly, the MTA had discussions with OpenPlans about a cheap method for providing real-time bus location data to riders. The MTA is turning scrappy!

But the holy grail is real time subway times and this one hasn’t gotten the refresh it needs — the MTA is spending $384 million to implement countdown clocks in each station which tell riders when the next train arrives. The work is already underway, but won’t be complete until 2016.

This seems crazy.

I haven’t seen any impressive scrappy third party solutions yet other than Roadify which focuses on crowdsourcing.  But it’s hard to imagine this getting the mainstream usage it needs. Furthermore, a crowdsourced system would break when you need it most: late at night when trains are infrequent.

Here are a few solutions I’ve come up with (All of these assume that the MTA hasn’t given you permission to install any equipment on their property.)

1) Consider that trains are actually visible at times: they cross over the bridges as they leave and enter Manhattan.  Why not install a detection device focused on these bridges?  Find a DUMBO startup with a nice view of the Manhattan Bridge and mount a camera or sensor in their window to detect when trains cross.  Then broadcast that information to the world.

2) Similarly, the ground shakes when a subway goes by.  So identify a bunch of different buildings that rumble when trains goes under them.  Leave an iPod Touch (connected to wifi) in the basement of these buildings. The device’s accelerometer could detect the rumble of each passing trains (which I’m betting would be a slightly different signal depending on the direction of the train). With a few of these detection points around the city sending back information to a central database, you could infer train locations for the entire system.  And this system would require a minimal capital investment:  buy fifty used ipod touches for $100 –even ones with broken screens are fine.  With only $5k equipment costs you’ve got yourself an 80% solution.

3) There are other ideas too that might work but don’t seem as elegant: find apartments near the subway exits and install a motion sensor/camera above the subway staircases — when floods of people come out, you know the train just came.  But this really works only at crowded times like rush hour when train schedules don’t matter much.

4) Use directional microphones pointed at subway grates to hear when trains go by.

Have other ideas on how to do real-time train tracking on the cheap?  I’d love to hear ‘em in the comments!

Possibly Related Posts:


Filed under: Lifehack, local, MTA, NYC   |  13 Comments
  • http://www.bradleylautenbach.com/ Bradley Lautenbach

    someday our generation is going to make it so the only thing in the world that takes 7 years and $384M dollars is curing cancer. right?

  • Tom Powell

    This is long overdue and those are some awesome ideas for hacking together an unsanctioned system. Do you happen to know why it's costing that much money to outfit the stations with countdown timers? Seems extraordinarily high for such a seemingly simple function. And what did the DC metro do to have timers installed there? They have had them for years. I wonder how much that cost to implement.

  • http://www.samianrosen.com/ Samuel Ian Rosen

    What about if we pay the attendant's who sit in the platform shops per tweet every time the train passes?? If no service, we could get them those motorolla radios that police / firefighters use and work every where. And with a $348MM budget, I think we'd have flying cars by the time the city ran out of money

  • PJH

    Type your comment here.

    WHAT attendants? And those Motorola radios worked SO WELL on Sept. 11, 2001, after all.

  • http://blog.jwegener.com Jonathan Wegener

    I sure hope so..

  • http://blog.jwegener.com Jonathan Wegener

    I actually ran the numbers on something similar. hiring folks at $10 an hour to stand in the subway stations. Assume you need 50 people spread out as “sensors” across the city and you want them there 24/7. That's 50 people x $10 an hour * 24 hours * 365 days a year = 4 million a year in costs.

    So the $384M budget effectively could pay to run this people-on-the-ground system for 100 years.

  • http://blog.jwegener.com Jonathan Wegener

    I'm not really sure. The article says: “The system, which cost $213 million to implement, works by rerouting signals to a central control center in Midtown that can generate an approximate waiting time using a computer algorithm. That information is then sent back to a communications room at each local station, where it is transmitted to an electronic sign. The signs and a revamped public address system cost an additional $171 million to install.”

  • http://harlan.myopenid.com/ Harlan H.

    Good post. A few ideas. For detecting trains, glue an RFID tag on every subway car. That'll set you back a $20. Put RFID detectors in every station. More expensive, but not crazily so. Just have those detectors send whatever data they have to a central database. A lot more info than the sound-based ideas.

    As for providing data to people, screw those expensive mil-spec overhead signs. At least in stations with cell phone signals, paste QR-codes on the walls that say “scan me to get realtime train arrival data!”. Easy.

  • http://twitter.com/davidmarcus1 David Marcus

    But of course that wouldn't get you the screens. I've been trying to develop a solution too, and it's really the screens where the cost adds up.

  • http://www.facebook.com/profile.php?id=9002508 Eric Boucher

    I think there is some confusion about where the gaps in data are. The MTA already know the realtime location of every train in the system. That is how they direct train traffic. The complicated part is how to disseminate that information to the masses. You don't get service in most subway stations so they are installing those countdown clocks.

    If we just want like a smartphone app or website, it would just take the MTA coming up with a way they can broadcast the information they already have in way that they are comfortable with from a security standpoint.

  • Somudro Gupta

    According to the MTA website, there are 468 subway stations in New York. That means the cost is over $820k per station. Absolutely absurd.

  • ALD

    I'd like to bring some clarification where I can here. I used to work for the MTA on budget issues and saw from the inside the unbelievable cost projects had. The scale sometimes seemed beyond comprehension – whether it be $5million for a boiler or $300 million for countdown clocks. Some of that expense comes from expensive union contracts. Other parts of the expense comes from the need for proprietary parts. Other factors include the notion that the system needed to be the “100%” solution noted above.
    Components of the $384 million dollar countdown clocks actually are much larger investments, such as system-wide communication infrastructure, high quality public address systems (notice how garbled messages are disappearing?), improved track signaling, and developing several algorithms to predict arrival times based on variances in signal systems, trackage, and schedules. While it's hard to visualize how this comes to $384 million, it gives some idea of the scale of the project. A “back of the envelope” solution will have to transmit a clear, decipherable message to thousands of people at hundreds of locations quickly and in a (mostly) subterranean environment cut off from communication infrastructure that we have become used to on the surface. To the extent that it relies on the MTA's own data it will have to do so with high degrees of differences in the type, scale, and fineness of the data.

  • http://twitter.com/mhookey Mark Hookey

    I wonder what the cost is of providing cell coverage in the underground (like in some other cities such as madrid)? If there was coverage then there are plenty of 95% options around smartphone location services (assuming at least a few people had their phones on), especially with multi-tasking. Google maps could do the same as they do with traffic.

    Of course the downside is you'd need to listen to new yorkers on their phones (one small respite on the subway).