8 ms·
Can I ask a stupid(?) question: So this Positive train control system seems super complicated. Dedicated radio frequencies, track signal loops, and so on. And a
by Someone1234 11y ago
Can I ask a stupid(?) question: So this Positive train control system seems super complicated. Dedicated radio frequencies, track signal loops, and so on. And all this complexity has been holding it up for years...
So my question is this: Car navigation as a problem set is similar but massively harder than train navigation. My car can tell me speed limits, which lane to be in, and knows information about all major roads in the area.
Why don't trains have a Google Maps-like system? You have 1/100th fewer "roads," you have directional assurances, you have speed limits set by a single authority (no city, state, federal distinction, etc), and so on.
Heck if you've played Train Simulator you KNOW that they have all of this route information digitised somewhere. So why not just throw it in an app, hook up GPS, and get it deployed tomorrow instead of ten years from now?
Is this a classic example of over-engineering the problem?
- dustingetz 11y agoamtrak operates at a loss, any investments need be approved by congress http://qz.com/409824/yes-amtrak-was-sabotaged-by-congress/ http://qz.com/409824/yes-amtrak-was-sabotaged-by-congress/
- mcpherrinm 11y agoThe simple answer: Trains can't stop quickly, and can't switch tracks easily. A car stays in lanes that go one way. Trains don't: they often share a segment of track for both directions, and you better be darn sure you're not going to run head-on into an oncoming train, or into a stopped one. Thus, trains need a lot more communications and coordination to ensure safety, unlike cars which use the rules of the road and quick maneuvering and stopping to adjust
- jonknee 11y agoThat still doesn't explain why a simple solution can't control speed limits (which would have prevented this accident).
- tyingq 11y agoThe barrier is likely the process through which PCT was mandated. Starting with this: https://www.fra.dot.gov/Elib/Document/2189 https://www.fra.dot.gov/Elib/Document/2189 "(3) POSITIVE TRAIN CONTROL SYSTEM.—The term ‘positive train control system’ means a system designed to prevent train-to-train collisions, over-speed derailments, incursions into established work zone limits, and the movement of a train through a switch left in the wrong position." The deadlines, definitions, etc, don't line up well with a "What can we do now, with existing tech, that would add value at a low cost?" approach.
- melling 11y agoStill, someone in junior high can do the math to figure out the solution given speed and location of each train. My guess is that no one knows the location or the speed of the train because we don't want to spend the money. Of course, it'll be extremely expensive to solve this because of how Amtrak is managed.
- rayiner 11y agoYeah, someone programmed the Uber app with junior high math to show you where your driver is. But I regularly see the driver pull up when the Uber app shows him still a couple of blocks away. I wouldn't Uber's app controlling my fricking train.
- Anechoic 11y agoSome transit agencies and railroads are experimenting with GPS, but IIRC the biggest impediment are tunnels,
- mikeash 11y agoSeems like that would be trivially solved by using dead reckoning in tunnels. You know the exact path of the tunnel, you know when you enter it, and you can count wheel revolutions to see exactly how far you've gone within it. But perhaps I'm missing something that makes this harder.
- iSnow 11y agoIn principle you are right, but precision is an issue: the train control system needs to know the position in the centimeter range. Counting wheel revolutions is notoriously unprecise because as the wheels wear down, they rotate faster. Public transportation is furthermore not an area of "hey I know, I just ask on Stackoverflow for an algorithm that counteracts the wear on the wheels". They do extremely conservative and defensive programming there (I once upon a time read the source code of one of the leading ETCS systems) in extremely restricted and conservative hardware (embedded 8080's in the 2000's) where the production process is audited and nailed down for everything (CF cards costing hundreds of $ for 128MB). Absolutely nothing is trivial there - the usual answer for an improvement idea would be a resounding "no way!". They are mortally afraid of the law of unintended consequences there.
- mikeash 11y agoWhy do trains need centimeter-level precision in tunnels? Especially if we're just talking about figuring out speed limits.
- Someone1234 11y agoI don't see why tunnels would be an issue. If the app has been tracking the object for a while it can calculate projected speed, and as it loses the GPS signal it can use an internal clock to continue to estimate the train's current location. Navigation systems already do this for signal losses (and to save battery).
- OliverJones 11y agoThe GPS owners (US DOD) have the prerogative to turn it off, or at least impair its resolution significantly. without a lot of warning. So a backup system is necessary. But, it sounds like the engineer's knowledge of the route is that backup.
- jonknee 11y agoIf GPS gets shut off there will be such disruption that we can just stop all the trains.
- AnimalMuppet 11y agoWhen do they shut off (actually go to lower resolution) the GPS? When there's a war (invasion of Iraq, for example). Or when there's a bottleneck for two long in the process of launching replacement GPS satellites. You want to stop all the trains if one of those occurs? Until the military action is over or there's a successful satellite launch? Get outta here.
- jhayward 11y agoYour larger point about backup being required is very salient, however it is no longer the case that GPS can be turned off or impaired by DOD. When SA was turned off in 2000 it was irrevocable. GPS satellites designed since that date do not carry the SA feature capability.
- gsnedders 11y agoThe requirements of PTC go beyond anything a car's navigation system has to, most notably maintaining separation between trains: also, unlike a car's navigation system, because it's a system there to ensure safety rather than provide assistance to the driver, the system has far higher requirements for failing safe. As far as I'm aware, proceeding without movement authority is a far more common cause of rail accidents than overspeed, hence why systems have traditionally been mostly concerned with separation and movement authority rather than speed. Really the biggest flaw of the US's push for PTC is reinventing the wheel: ETCS[0] has been standardising this all over Europe (with parts, such as GSM-R, being adopted in large parts of the world!), and the goals for PTC largely match up with ETCS Level 2, as far as I'm aware. There are a number of off-the-shelf systems available, leaving the largest issue as GSM-R frequency acquisition. And, I suppose, the political issue of using a non-American system. ETCS Level 3 essentially allows pure-radio positioning, but there's currently no planned deployments because nobody yet has constructed a system that meets the required "safety integrity level". Note that one of the issues here is you need devices at both the lead and rear vehicles to ensure train integrity, whereas currently typically one only has such equipment in the driving vehicle(s). There's hope to use the EGNOS satellite augmentation system to get better accuracy guarantees than GPS/GLONASS/Galileo provide, but it's all still experimental, and the only experimental project there failed. [0]: https://en.wikipedia.org/wiki/European_Train_Control_System https://en.wikipedia.org/wiki/European_Train_Control_System (an article which I will openly say defines far too little jargon it uses!)
- rayiner 11y ago> Car navigation as a problem set is similar but massively harder than train navigation. My car can tell me speed limits, which lane to be in, and knows information about all major roads in the area. Your car navigation isn't "positive control." It can't slow your car down for you, and when it fucks up (like Uber's navigation does 50% of the time on the way to Union Station), it can't fuck up control of the car. It's a fundamentally advisory system that is not designed to be trusted (and very often is not trustworthy). You could hack together an app to tell an Amtrak driver the speed limit in any given location, but that wouldn't have helped Amtrak 188. To prevent that crash you needed a system that can override the train operator and slow the train. Once you give the computer authority to override the driver, a huge amount of complexity sets in. First, a substantial portion of the Northeast Regional track does not have cellular coverage. The trains have no way to communicate information in those areas. If you're PTC is going to slow down a train in a dead zone, it's not going to be acceptable to not be able to relay that information off the train. That requirement drove the system design. A big part of the delay in deploying PTC was setting up the necessary wireless communications infrastructure. Then there is the issue of integrating with the existing signaling system. Your car navigation doesn't interact with the signal lights and tell you when to stop. The PTC needs to be able to do that. It'd be a recipe for disaster to have a computer system that can slow down or stop a train but had no idea what the signaling system was doing.
- mortehu 11y ago> It can't slow your car down for you AEB systems can, though: http://www.euroncap.com/en/vehicle-safety/the-rewards-explained/autonomous-emergency-braking/ http://www.euroncap.com/en/vehicle-safety/the-rewards-explai...
- dopamean 11y ago> It's a fundamentally advisory system Couldn't the train's computer simply advise the engineer (that's the guy controlling the train correct?) to slow the hell down when traveling too fast in a particular zone? Would that not be a step in the right direction? Maybe the system could even require that the engineer dismiss the warning by actually applying the brakes. Maybe after that doesn't happen the train can apply the brakes for you. That doesn't seem that complicated. I'm not trying to be argumentative. I'd love to hear some specifics about what complexity that adds.
- blibble 11y agothe UK's railways have a cheap and simple solution: two transmitters at different frequencies inside the track, the distance between them controls the maximum speed. if the train detects the second too quickly after it detects first it applies the brakes. they look like this: https://upload.wikimedia.org/wikipedia/commons/thumb/2/2f/Tpws_loop.jpg/250px-Tpws_loop.jpg https://upload.wikimedia.org/wikipedia/commons/thumb/2/2f/Tp... the wikipedia page has more details: https://en.wikipedia.org/wiki/Train_Protection_%26_Warning_System https://en.wikipedia.org/wiki/Train_Protection_%26_Warning_S... there's a more expensive system called ATP that is a lot more complicated (and had a limited rollout), and ECTS is coming...
- jameshart 11y agoThat's a very elegant solution; has the disadvantage that its failure mode is not safe - if the train fails to detect a signal at all, then it will just continue at whatever speed it likes. And of course, like a lot of physical engineering solutions, it's not built to be secure against malicious actors. Someone who broadcast fake signals could slow a train down - or confuse it into ignoring the real signals. This is the kind of thing which autonomous-driving skeptics need to keep in mind when they worry about 'how will you stop people from continually stepping out in front of self driving cars to stop them', or 'how will you stop people from jamming your lidar sensors?' How do trains stop people from interfering with these safety systems? They don't - the fact that interfering with the safe operation of a train is dangerous and subject to huge legal consequences is what stops people from doing it. Same applies to your worries about car technology. What stops people interfering with lidar? The fact that most people are not psychopaths, and they would probably go to prison if they did.
- jcrawfordor 11y agoOne of the big problems with rolling out infrastructural upgrades to the US train network is, in general, its size. Consider that there are still substantial sections of freight rail in the country that are what is referred to as "dark territory," meaning that there is no signally installed. Additionally, there is usually no radio or cellular coverage in these areas. Train motion in these areas is typically governed via track warrants, meaning that an engineer must obtain written permission from a dispatcher before entering the area (and only one train is given this permission at a time). In the NE corridor, everything is of course fully signalled, but installing additional wayside equipment such as for PTC is a huge project over many miles of rail, even in relatively small areas. Solar-powered radio equipment has driven down the cost of installation, but not by as much as you might think. A big part of the high cost is the engineering requirements of the federal railway administration, which are (for good reasons) extremely strict and require proven high reliability. This reliability must apply everywhere. GPS availability can be iffy under a number of circumstances, most obviously tunnels. There needs to be some non-GPS method of positioning trains in most systems. An example which the HN community will be relatively familiar with is the BART, which features automated train control. BART trains position themselves (and receive speedcodes for central control) via a continuous antenna strip mounted to the safety cover of the third rail. If you inspect the third rail across from a platform you will see it, as well as occasional cables connecting to the trackside control equipment. Installing this sort of thing over miles gets very costly! The closest analogy to what you are describing is a product called GE Trip Optimizer that uses information about the route and train consist to automatically control throttle for fuel saving purposes. Trip Optimizer does not meet requirements for a safety system, though, and so the engineer is always permitted to override it (and must if any unusual circumstances are encountered), and the engineer must manually account for signals. To be a full PTC system it would still need quite a bit more, most notably reliable communication with signals (better than current cab signal systems). There are sub-PTC safety systems available, such as train stop devices (which can be rigged for overspeed protection). These are still expensive to install though and at this point most everyone wants to go directly to PTC, which provides much greater protection.
- pm24601 11y agoNo. PTC is more than GPS. Much more than that: 1) which track is the train on? GPS off by a few meters 1 way or the other means a different track. 2) What are the switch settings ahead? 3) Is there a train stopped ahead? 4) Is there any obstruction ( i.e. rock fall )? 5) Are the crossing gates ahead properly functioning? 6) Are there vehicles across the crossing ahead? 7) What are the normal speed restrictions in place? 8) Are there maintenance workers ahead? 9) Are there temporary speed restrictions in place because of tracks needing repair? 10) What is the mass of the train? Heavier trains = greater stopping distance. 11) What is the track surface - leaves in the fall can create a biofilm that increases stopping distance. 12) What is the cargo - some cargo requires more gradual deceleration - bottles of wine anyone? 13) What about tunnels? 14) What about mountainous terrain with poor line of sight to GPS satellites? 15) What about signals that may interfere with a GPS fix? 16) What about going through cities with buildings and in trenches? Expanding on a few additional points: "Where is the train in front of this train"? a) Its not enough to know where the locomotive is: some freight trains can be 2 miles long. So if the train in front is 2 miles ahead but is a passenger train - no problem. A freight train? Big problem. b) Is the train ahead on a siding that the current train will not hit because the switch is thrown to transfer to another track? "What is the allowed speed?" a) Is the train going to take a speed restricted curve? b) Or continuing straight at a higher speed? c) How fast can the train decelerate - freight is slower than passenger trains. In a car its annoying -but having a train get confused and go into an emergency stop - that is something else. This is not a simple problem. Then throw in all the systems that need to interact and function correctly, have good failsafes, etc. There is a reason why trains need humans driving them!
- iSnow 11y agoTrain control is not super complicated, it has been deployed in various forms in Europe for decades. Rail infrastructure providers offer turn-key solutions based on common standards like ETCS. It is expensive due to a lot of hardware that needs to be deployed, but that's it. Even countries like China with their CTCS use it on modern rail lines. One of the problems is that you can't really use GPS for location awareness because it can be down without notice and is not available in tunnels. Maybe that restriction will be lifted one day. But there are proven solutions, the US is just too stingy to buy them.