Wednesday, September 25, 2013

Green Line "Next Train" Signs: How They Work

Inside the MBTA Operations Control Center  (OCC)            Courtesy: MBTA

With last weeks unveiling of the MBTA's new "Next Train" signs for the Green Line, some of you might be wondering why does the Green Line still not have real time arrival data?

Map of AVI locations       Courtesy: MIT
Although these new "Next Train" signs are new to the Green Line, the technology that they run on has been around for the past two decades. AVI or automated vehicle indicators are how the Green Line is currently tracked by MBTA employees. 

Around the system there are a series of AVI readers embedded in the tracks. When a train passes over one of these readers it reports the trains location to the T's Operations Control Center (OCC) which is located at 45 High St in Boston. When a train passes a specific point it's number appears next to that AVI point on a large map at OCC. This means the T knows that a specific train is at or beyond that specific point. When it reaches the next reader it will disappear from the first point and appear on the next one.

Seems like the T should know where most Green Line trains are right? Unfortunately, it's not that easy. In the central subway, which runs between North Station and Copley Station, the AVI points are somewhat close to one another. When the trains go above ground however, the points are very far apart. All above ground routes are monitored by an actual human being. An MBTA inspector will post themselves at a certain stop along the line and monitor when trains are passing by.
Still a little confused? Here's an example: Lets say Green Line train 3651, which has a destination of Boston College, just passed the AVI point at Copley station. The T knows that the train is at or beyond Copley but it has not yet passed the next point which is at Kenmore Station. Once the that train passes Kenmore the next AVI point isn't until Chestnut Hill Ave. which is 16 stops from Kenmore. On average, that's about a 25 min ride over 3.5 miles. During that time central control has no visual on that train unless the inspector on the line physically reports its position. 

Overall, the T knows what train will be next (B,C,D,E) but they do not know the spacing between trains. Therefore no real time train data can be determined for the Green Line. These new "Next Train" signs are a big step for the Green Line but bigger and better things are coming. Back in January the MBTA announced that they would have real time tracking data for the Green Line by 2015. Since it's still over a year away, these "next train" signs will have to do for now. 


  1. This is amazing - I do think that the arrival times have made a big difference for the lines which have them. When you are a passenger, the worst thing is the feeling of total helplessness not knowing what's going on.

  2. Sounds like something an iPhone tucked away in the in the motorman's compartment could fix, certainly for the surface portion of the line in conjunction with the subway AVI system, maybe in the subway also also now that there's cellular coverage. The good people at Nextbus told me a couple of years ago they had the technology and offered it to the T, but apparently the T prefers to implement these things themselves. And so, we wait.

    1. Yes, in theory, the location of the trains can be easily tracked by using triangulation using cellular signal timing between two cellular transceivers and assisted GPS above ground, but the system also needs to be robust enough to stop the train in the event of an impending collision. Triangulation would also prove to be difficult where signals will be bouncing off walls in curves in the tunnels.

      I understand why the T wouldn't want to pay NextBus for train tracking. Remember, their ultimate goal is to improve throughput AND safety; next train prediction is just a feature that can be derived from the data used to run a computer-managed system. This can be achieved with PTC or CBTC and there are a lot of products out there, but the MBTA is also locked in by its budget. Last cost estimates of computer-enabled train tracking safety systems came in at over $600 mil, a non-trivial number that would also generate well over that much in economic activity where the Green Line to suddenly become a beacon of Boston/Massachusetts/American transit modernisation.

      The other issue is security - the very nature of proprietary PTC/CBTC systems that makes them expensive also makes them secure from malicious interference. You'd still need to build in-tunnel infrastructure on top of the existing AVI system for the train to talk back to the OCC without exposing the data using unsecured standard cellular data or WiFi and report its location at a high enough frequency for proper data resolution.

    2. Interesting info on the security problem! But safety and customer experience are two different problems which do not require the same solution. Improving the customer experience, with basically the same technology used for MBTA's buses (also from NextBus?) would go a long way and also help the MBTA manage its fleet and react better to service delays. Such a system doesn't even need 100% coverage, though it certainly helps. This could be done for pennies on the dollar compared to the robust, remote safety system you describe.

    3. NextBus uses information from the built-in TransitMaster automatic vehicle location (AVL) system that has been in use by bus operators at the MBTA's bus OCC. The company that makes TransitMaster also makes rail vehicle location systems, but again, these are proprietary technologies that the MBTA would need to bid out for. Even if you did install TransitMaster on all the trains, they would be useless underground where the majority of delays happen. Service delays today are also disproportionately from faulty rail switches or signal failures, nothing that can be helped by knowing where the trains are. These vehicle location systems really benefit customers by permitting the MBTA to run trains closer to each other during rush hour while simultaneously ensuring a safe braking buffer between trains.

      I totally agree that the MBTA should pursue any cost savings it can and implementing a vehicle location system for the Green Line using smartphones would undoubtedly be cheap, but there are a number of technical limitations due to the tunnel infrastructure alone. Remember GPS doesn't do so well underground and tower-based assisted GPS would have limited efficacy because of tunnel geometry and non-overlapping cells.

      On top of that, the MBTA barely has the man power to get its core duties done at this point because of all the cost cutting and workforce cutting we, the public, have demanded after years of corruption due in no part by the people in power today who are hemmed in by these decisions. I doubt there'd be enough man power to develop even a basic train location system, assuming you can overcome the technical issues.

      Things are certainly changing at the T, but good luck convincing the engineers at the T to let (grad) students at MIT or Northeastern to develop this system. The biggest reason management has been hesitant to open up the CharlieCard system is because of the security expose MIT students did at DEFCON years ago. For better or for worse, agree with their decisions or not, their top concerns are safety and security. A number of security loopholes were plugged shortly after the expose and a healthy DHS grant has had innumberable cameras and authentication systems installed across the entire system, but the last thing T management needs/wants is another weak, unsecured cog just to improve the customer experience at platform level beyond what they've hacked together today.

  3. Again, all excellent points and lots of obstacles, but management should know the (big) difference between purely informational systems and those with real security concerns like payment and control systems. The new cellular infrastructure ought to offer information on call routing in the tunnels, but even if it didn't, many users would benefit from predictions which could be made for the entire inbound service and much of the outbound service as well. Heuristics and integration of existing AVL data could provide fairly complete coverage. There are plenty of delays on the surface lines, due to on board payment, equipment failure, downed trees, etc.

  4. I've never really gotten a good answer from the T why they cannot simply install more AVI readers along the tracks. Yes, the prediction service won't work exactly the same way that the GPS transponders do, but it is possible to write software that looks at AVI reader data and churns out reasonable predictions. And they won't need to install a GPS unit in every train.


Related Posts Plugin for WordPress, Blogger...