GPS Signal Control

[RailroadDoc]

Mark;

Thanks so much for your reply. I had no idea that trains could operate at speed as close as three miles apart. That would pretty much negate any advantage for GPS to increase capacity that much.

I don't have the current issue of Trains magazine with me. In the magazine you did mention the additional number of train crews and locomotives that were needed by Union Pacific as the average train speed decreased. From what you are saying I gather that the real capacity problem presently is at the terminals -- not enough crews/locomotives/tracks available to break down and put together the trains that are arriving and departing.

I might tell you a little bit of my background. I'm actually a surgeon with a degree in electrical engineering, besides being a pilot. About the only thing I don't really have any experience in is railroads!! Maybe that's why they interest me the way they do.

I would suggest that you take your last reply and put it somewhere in one of the Trains issues. I think its an excellent analysis of the problem.

[jchnhtfd]

It's always so nice to read your stuff, Mark! I'd just like to doubly emphasize one point you made -- everything -- repeat EVERYTHING -- we do out there is intended to be fail-safe. Which, in our context, means if something doesn't check out, you stop. Can't read the signal? Stop and find out what's up. Signal dark? Ditto. Obscure alarm on the third unit back? Stop and find out what it is. Trackside detector sounds off? Time for a walk to find out. Aircraft don't do that (they can't!) -- as a fun way to relax, I'm also a 6,000 hour plus pilot and proud owner of a Piper Arrow and I know...

And as you also pointed out, it's a lot easier and cheaper to run the railroad the way we do now; signals and control aren't the problem -- track is.

[Anonymous]

RailroadDoc:

Returning to your original question, "how much could GPS increase railroad capacity?"

Answer: zero. It's not cost-effective. It's cheaper to buy more track. Railroads have asked that question already, and come to that answer already. Certainly, there are people in railroads who are old-fashioned and stubborn, but there are many that aren't. The problem is that the technology is inappropriate to the application.

Not that railroads haven't tried. GE Harris-Harmon, Lockheed-Martin, US&S, Alstom, Ansaldo-GRS, and Class I railroads have invested over $100 million to date in various types of radio-based train control. None have been found feasible on a cost-benefit analysis, and severe technical difficulties have been resolved only at tremendous cost. Some technical problems have proven resistant to solutions. There are several prototype sections in tests now in the U.S. (Amtrak Michigan Central; Illinois high-speed corridor, NJT). Each step has required expensive modifications, a lot of software debugging, and none of these systems is considered truly operational yet, at least in a form that holds any hope of increasing capacity. An automatic train stop feature, yes. But that's all.

One month ago, I listened to a Lockheed-Martin presentation at the Kellogg School of Business at Northwestern University on the safety improvment value of PTC. Answer: zero. Train collisions and derailments caused by authority violations and speed violations that a PTC system could prevent are so rare that L-M had to multipy the data tenfold in order to subject it to standard methods of statistical analysis. Statistically speaking, wrecks associated with signal systems are indistinguishable from wrecks caused by random chance. After the meeting, the L-M engineer told me "I don't know why anyone is bothering with this. I'm happy to take their money to do useless studies, but the truth is, signaling-caused wrecks are the least of the industry's worries."

But what about a capacity basis?

The next evening, I unfortunately was not able to attend an associated presentation at Northwestern that attempted to make a case for PTC on a capacity-increase basis. The consensus of six experts in the audience I talked to the next morning (all of whom would very much like PTC to work, mind you, because it would help them compete with trucking), was that PTC cannot economically increase capacity. Maybe if the cost came down to 10% of estimates. But there are still many technical problems with PTC that haven't been satisfactorily resolved, and no one knows the cost that will be incurred.

You raise many interesting ideas in your proposal for using GPS. None of them will work, in my experience, because they all violate authority to occupy a main track, which is the fundamental concept of railroad operation. I'll address just a few salient points.

1. It's suggested that the value of this system is that trains could operate on three-mile headways instead of 10-mile as at present. Actually, trains already operate on three-mile headways, or even less. The headway depends on the braking distance of the train, and block signals are spaced accordingly: two or three miles on a typical Class I main line, maybe one mile on a heavy-rail commuter line.

Theoretically, you could operate two trains running at 70 mph, 100 feet apart. But that gets you nowhere: at the end of the line, as everything enters a terminal, the trains are only going to go through the switch so fast. (This is the "fleeting" fallacy -- running a bunch of trains in one direction, then a bunch in the other, merely overloads terminals and makes terminals grossly inefficient because the traffic appears in spurts.) No one anticipates being able to do much about tightening train spacing with PTC except in the case of a line with trains of widely divergent braking distances, such as 40-mph drag freights and 125 mph passenger trains. But that is such a prohibitively uneconomic way to run a railroad that no one wants to do that.

The limits to track capacity are not too many trains -- it's finding big blocks of time to maintain the track. UP's triple track in Nebraska essentially consists of two active tracks, and one down for maintenance. A frog in a crossover or switch on a busy main line will require 4-6 hours of welding time a week to add back metal that's been worn off. Rail on a heavy tonnage line, on a curve, is good for maybe two years. And so forth. PTC won't help you there. Once a single-track line gets past 70 trains a day, it becomes virtually unmaintainable. Temporary speed restrictions appear, capacity crumbles, time to maintain track becomes even more scarce, and a vicious cycle ensues.

2. Fuel economy could conceivably be improved if a train was instructed to run at, say, 30 mph for the next 10 miles to time it for a meet, rather than racing up to the red light. Rather than hire an army of train dispatchers to administer this, much effort has been spent on software to automate the process. It's turned out to be beyond the ability of the software engineers to write. One Class I, which I will not embarrass by mentioning, has spent more than $50 million attempting to write that software. The software miserably failed in every test, some of which I witnessed firsthand. (Even the very simple "Automatic" software that equips most dispatching consoles, which self-clears signals and sets up meets on single track, does all sorts of dumb things. Most dispatchers will not use it.)

3. Speaking as a former train dispatcher, the elements of the system you propose would increase the workload on the dispatcher by 100% to 200%. (I'm sure my friends in the craft would all applaud you, as doubling or tripling the ranks would instantly boost them so far up the seniority roster they could bid in that cushy afternoon job they want.) Labor costs would wipe out the savings, if there were any.

4. I doubt that an ATC has more information on you and your aircraft than a train dispatcher has on his or her trains. (They're not called controllers in the U.S., though they are in Canada.) I've never visited an ATC office, nor do I know much about that craft. However, I do know train dispatching quite well, and I think you would be deeply astonished at the information at a dispatcher's fingertips. Moreover, I knew everything about my trains I needed to know -- knowing the location of a train within 20 feet is of little or no practical value, and if I really want to know that, I can call the train.

5. Current methods of rail operation are summed up as "protect, then authorize." No train, man, or machine is given authority to enter a main track until the dispatcher has protected that movement. That protection is absolute, and can be violated only through willful action or negligence (such as running through a signal indicating Stop without stopping). In other words, no matter what happens -- a dispatcher can lose all control over the signal system and all radio communication, or the whole signal system can all go dark -- and Trains Will Not Run Into Each Other. They're protected. (And yes, this happens often. It happened to me every hot, humid, rainy night in the Ozarks.) Your system includes numerous "authority take-aways," which are by definition not fail-safe, whereas current methods of railroad operation specificially exclude authority take-aways.

My suggestion is to obtain a good book on railroad operation methods, such as Elements of Train Dispatching, Vols. 1 and 2, and a rule book, learn them, and see if your ideas still fit. It would also definitely help you to spend a year working as a train dispatcher or in train service to understand the problems first-hand.

I certainly did not understand how railroads work until I worked for one. As one good friend of mine put it, "I thought I knew about railroads until I went to work for one. Then I found out I didn't know ****." After a couple of months on the console, I called him up one night and said, "I see what you mean ...."

By now, you've probably noticed that technical solutions proposed in this and other forums are regularly shot down by railroaders. It probably frustrates you to get answers that boil down to "Because." No one is trying to ignore you (in contrast, the effort to which people go to answer questions is quite large), but because a full answer to almost any of these questions can't be written in an hour. In truth, the only way you'll get a full answer is through years of education. Personally speaking, I would not even attempt to propose an alternate form of train control without first arming myself with a degree in electrical engineering, and surrounding myself with a team of experts with long experience. Railroading is a technology that worked out most of its fundamental questions over 150 years ago, and railroads are very good at accepting, adopting, and implementing new techological solutions the instant they appear. For example, when EMD's FT demonstrator appeared, most of the railroads that tried it knew within a matter of 24 hours that steam was deader than a dinosaur. Less than one day's experience with the diesel to know it was time to throw out an entire way of life, and they couldn't wait to do it!

When you get your July issue in a couple of weeks, read the article on B&O train control experiments in 1965. It goes right to the heart of your proposal, and describes why it won't work.

On the other hand, railroads have severe cultural and ideological problems that are obvious to the outside, and outside solutions would work (assuming you have the power and imperviousness to matters of consience of Josef Stalin). But that's because cultural problems are human in form, and it takes no technological knowledge to understand them, only knowledge of human nature. Technical problems require techical expertise.

[RailroadDoc]

dharmon: thanks for the referrral.

The initial question I asked was how much the PTC/GPS system would increase railroad capacity by allowing more trains to be run closer together. I doubt (although I may be wrong) if there are any busy mainlines anywhere that are dark.

Now as to some of the points. . .

1) As it now stands the crew doesn't know whether they are taking a siding or not. I don't see the problem with a dispatcher telling the train crew which siding they should take and then having them confirm that they have done so.

2) What happens when trains lose contact? There two parts to this question. If GPS contact is loss, the computer will still have a very good idea where the train is from the history of its position and speed. If the GPS signal doesn't return within x number of seconds, then radio contact will be made with the crew, and information about the train location and speed can be forwarded over the radio. If radio contact is lost, than the train will go into an automatic shut down mode -- depending where the known other traffic in the system is. If the train is on a dark section of track with no other trains around for 100 miles it wouldn't be quite as urgent to shut things down, as opposed to a busy mainline, where you would probably have CTC backup.

3) Every piece of equipment would not need a GPS transponder. Every individual operating a piece of equipment on the track would need one. Therefore, the GPS/transponder units would be portable, and go with a crew -- the same way an EOT device is.

4) I might mention that air traffic control is done by the use of transponders. This is a piece of equipment located on almost every aircraft that continuinously transmits its identity and altitude to air traffic control. Before a aircraft takes off it is assigned a transponder code, which the pilot programs into the transponder. Central control takes that information and combines it with the radar sweep so that they see your location and correlate it with your altitude and identity. If your transponder quits transponding, then air traffic control knows something is wrong and will immediately try to contact you.

4) I don't think anyone has really made an argument to take CTC down -- although I did eventually mention that possibility. Some type of control would need to respond to be able to remotely line the switches, unless you wanted the train crew to stop and do it. The value of this system is to allow trains to operate much closer together -- ie, three miles apart instead of ten miles apart (or whatever the standard is now).

5) As far as "your little handheld" not being interactive -- actually, some of them are. The last time I checked you could buy a combination portable aircraft transceiver/GPS map unit with airport information for about $1000. Well worth the cost!!!

6) Operating conditions? If we're talking about conditions that a portable radio can operate in, than a GPS PTC device will be able to operate in it.

7) Reliability? Well, probably not quite 99.9999% reliable, but nothing on a locomotive is either. Many small airplanes will have two GPS moving maps systems mounted to provide backup in case one goes down. And the price is only about $5000 apiece.

8) "You won't run this off a PC, either"? You might be interested to know that the generic PC computer that you can buy for $600 at Walmart is considerably more powerful than supercomputers were 15 years ago. The speed of computation (now 3 gigahertz ++!!), the RAM memory (one gigabyte is no problem), and the hard memory (200 gigabyte is something like $150) is considerably beyong what even a multimillion dollar supercomputer was able to do just a few years ago.

Now I'm not suggesting that the system be based on a PC -- but hardware will not be the problem with implementing such a system. It will be the software that will have to be developed.

And even that should be quite doable. The Air Traffic Control System is built on ridiculously obsolete computer hardware, and they are able to handle way more aircraft than there are trains going about ten times as fast as trains -- and besides that, they have to keep track of their altitudes too.

And as "far as 10 or 20 thousand elements moving across a grid of several hundred million points" . . . why in the world would you do it that way? With air traffic control aircraft are handed off from one district to another, and each district just keeps track of planes in their district. As you leave one district you are instructed to contact the next district on their appropriate frequency. Meanwhile, the district you are departing from will forward all the computer information on you to the district you are entering.

When handheld GPS units first came in (a little over ten years ago) the biggest purchasers were airline pilots. Most airliners were not equipped with GPS until several years after that. At that time it was not unheard of for private pilots who flew Cessna 182's to kid pilots who flew 757's about the antequated nature of their navigation equipment -- because they didn't have GPS units.

I find it a little ironic now that if I am flying around in a Cessna 182 now -- an aircraft worth maybe $100,000 -- that my air traffic controller will have much more information about me and every other aircraft around -- than a central traffic controller will know about his train on the ground that's moving much slower and is worth much more -- maybe $20 million dollars.

[Anonymous]

First, one of the goals of PTC is to be able to extend signal-type protection to non-signaled track. If you have to install a signal system (circuits in the track) then you eliminate part of the benefit.
Second, the crew doesn't know when they will be taking a siding or necessarily why. That is controlled by the dispatcher.
Third, the train length is provided by the lists and is fairly accurate, depending on wether the train has picked up or set out cars enroute.
Fourth, yes but what happens to the trains behind the train when it loses contact? If the train ahead disappears does the system let the following train speed up? If an opposing train disappears, does the system let the train in the siding out on the main?
You have to account for loss of contact because it will happen.
Fifth, yes but you would have to install it in every engine and every EOT, and it would have to be interactive, transmitting a location as well as recieving it. That's a couple hundred thousand things to keep track of. Throw in on track vehicles and maintenance equipment, there's another hundred thousand things to track. That's a lot of satellite air time.
Sixth, that info is nice to know, but really the only thing they need to know is what signal indication or speed they need to move on or how long they will be stopped (to cut crossings, etc).
Seventh, you would probably need CTC to remain in effect at all times as a back up. Remember your first point, those signals in the track? Well that is a more complicated CTC system, so you can say you are "retiring" it but in reality you aren't, you still have all and more of the cost with your coded signals in the track.
Finally because a little map isn't interactive. You don't have to be two way with a handheld mapper, you don't have to have it stand up in -50 deg with a 90 mph wind and 2 ft of snow and then 120 deg heat a month later. Your littel GPS indicator doesn't have to be 99.99999% reliable. Why do you think you can buy a am-fm radio for $10 but it costs thousands, if not hundreds of thousands for a commercial grade airline communications system. You won't run this off of a PC either. The amount of data you will have to process on a real time basis to track 10 or 20 thousand elements moving across a grid of several hundred million points is staggering. This would have to be a multi main frame based system.

It becomes a big deal real fast.

Smith

[dharmon]

Yo Doc....more food for thought.....

http://www.trains.com/community/forum/topic.asp?TOPIC_ID=9234

[:)]

[RailroadDoc]

Interesting concepts. I'm not a railroad man, although I'm a private pilot, and of course love the GPS systems.

A few thoughts;

First, I don't think the GPS would need to be accurate enough to differentiate between parallel tracks. This information could be achieved by other means and transmitted back to the central computer the same way the GPS coordinates are transmitted. One thing that comes to mind would be a simple low frequency tone that is transmitted through the rails, different for each set of tracks. It would be simple for a computer to keep them straight. Even aircraft don't rely solely on GPS units; a plane's altitimeter gives a much more accurate report on a plane's altitude than a GPS unit could ever hope to give.

Second, you would think the train crew would know which siding they are supposed to be taking. If they wind up on the wrong one they can certainly get on the horn and check out what's going on with central command.

Third, you really don't have to GPS determination of the length of the train. You already have that information when the train leaves the switching yard. For calculation purposes the train will always be equal to the calculated length from the number and type of cars plus a 2-3% fudge factor.

Fourth -- As far as the GPS losing a signal? Not a problem. The central computer will know not to expect a signal from a train while it's in Moffat tunnel, for instance. It can easily be programmed so that ten or twenty second interruptions are ignored. Anything longer than that will trigger a call to the crew -- sort of a "Train 54, where are you?"

Fifth, as far as equipping every piece of equipment with a GPS and transmitter -- Why? Wouldn't it make more sense to regard the GPS sort of like a EOT device? In other words, the GPS goes with the crew and the unit is mounted in the locomotive they are using. It certainly can be made small enough to do this without difficulty.

Sixth, the system could easily be adapted to give the crew a computer update on the traffic around them -- i.e., "Train 67 is headed toward you but should be taking the Birmingham siding in 13 minutes. At your present speed of 68 mph you will reach that siding in 27 minutes. Train 13, which is five miles behind you, will follow you through that siding."

"Remember at milepost 136.4 there is a 20 mph restriction because of recent flooding. This will be in effect until milepost 144.3."

Finally -- The system can be installed gradually. First the GPS units become available, with a rudimentary central control computer. After some experience is obtained with them, a certain section of track could be used with CTC backup. Only after the system was fully functional and debugged would the CTC (or whatever system) be taken down.

And as to the costs. . . . .You can get an incredible sophisticated moving map GPS unit for a small airplane for around five grand. And the small plane market isn't all that great -- not all that many people can put out $250,000 for a fancy fair weather toy. I would be interested in just how the one million dollars a mile figure is derived.

Thanks for everyone's replies!!!

[tree68]

Mark and Dave - I agree fully. That's why GPS has to be a part of the solution, not the whole thing. Even though GPS may be fuzzy, it still has a certain level of accuracy. Using the concept of factoring in speed, geography, train length and weight, and necessary buffers (to account for the fuzzy GPS), optimal distance could be maintained. For instance, if for some reason track speed was reduced, trains would be able to follow more closely, rather than be constrained by the fixed signalling system.

By mixing GIS (computer mapping), track circuits, and GPS, it should be possible to manage things very nicely. The fact that a railroad is a fixed plant can be used to advantage, as there is near absolute (never say never) control over what occupies the space. Very few alleys or side streets to worry about.

Re: Curves and Loops - Even an inexpensive GPS will mark waypoints. There is no reason why a railroad GPS tracking system should be any different. Thus with GIS and plotted waypoints, a computer system could track the length of a train very accurately, even with the fuzzy GPS finagle factor figured in. As for loops, while the "as the crow flies" distance between the lead loco and the EOT may be measurable in single digit yards, the tracking would take the loop into account and present a fairly accurate number. Combining that with track circuits and other sensors would reinforce the information.

I'm certainly not saying this is the be all and end all technology for controlling trains, but it's probably going to happen some day. The technology exists today to make it happen. It still comes back to bucks. As already mentioned, all components must be suitably equipped, and the computers must be properly programmed to process the information and provide it to those who need it. When you add run-throughs, etc, it gets more complicated - witness the gyrations necessary with existing railroads that have special signalling systems.

Of course, this begs the question of whether the train crews can be replaced. Not likely. The dynamics of running a train are many and varied, and have been discussed at length on the forum. A solution such as this would simply be another tool with which to control traffic. Its usefulness would be on the busy lines. On a one-train-a-day line, it would be overkill.

[dehusman]

Regarding using GPS to measure train length, that only works on tangent track. GPS measures the position horizontally between two points and on a railroad those two points can vary significantly in the distance between them.
Consider a train approaching a loop track, the train is exactly as long as the loop is around. As it goes around the loop it gradually becomes "shorter" because the relative distance between the engine and EOT decrease until the Engine passes over the rear of the train and the length of the train effectively drops to zero. While this is an extreme example, the same distotion will occur on any curve, only to a lesser degree.

Dave H.

[Anonymous]

Larry: Interesting comparison you make to emergency services. It allows me to illustrate the difficulty of railroad operation. My wife was a paramedic and later a paramedic dispatcher for the city of Fort Worth. They dispatched using GPS and an electronic map of the city, just like you see in Phoenix. She thought railroad dispatching would be about the same, right? Wrong.

In her words, "railroad operation is more than 100 times harder than ambulance and fire apparatus dispatching." Part of the problem is that railroads have one degree of freedom (forward and reverse) and used a fixed guideway, whereas rubber-tired apparatus have two degrees of freedom and a self-steering guideway. Thus, railroad operation is all on absolutes, all of the time. A train absolutely owns a section of track, and has absolute boundaries. GPS has fuzzy boundaries.

[tree68]

I agree that GPS needs to be a component, not the whole shebang. Putting a GPS in both the FRED and the lead loco helps demonstrate train length. Adding some unit ID (such as the trucking industry now uses) with that helps the dispatcher keep track of who is where and how they are doing (all engines in notch 8, trainline normal, etc). On a straight section of track with no diversions, GPS with unit ID, together with some computer oversight (and a reasonable degree of confidence in the system), it would be possible to stuff quite a few trains in. Speed, weight, track profile, and the corresponding stopping distance could all be computed continuously, allowing train seperation to be minimal (factor in a cushion, there). Cab indications would replace lineside signals, and could give distance to next train, as well as a recommended speed.

On the other hand, for close quarters (yards, passing sidings) where exact location is crucial, there's nothing like a track circuit.

But, all that tech stuff costs money. It'll be a while.

As an example of the application, though, the Phoenix, AZ, Fire Department has equipped all their apparatus with a system that includes GPS. They now dispatch by the equipment closest to the scene, as depicted on a map of the city that shows where all apparatus are.

[rrnut282]

Europe is going to announce that their new Gallaleo system will be compatable with GPS. They also won't degrade the M code like the US does. Having more satellites and frequencies to read should help lessen lost or reflected signal problems.
As for having to equip every piece of equipment, why not just equip locos and ETs to pinpoint each end of a train?

[mudchicken]

Meter grade GPS is still insufficient for trains passing in sidings. This bug has been known since Rockwell's LARS, NTRAC & ANSAC days in the mid- 1980's.... To get things down to survey grade RTK GPS you need beaucoup base stations and a ton of software to sort stuff out on the fly. (Thus the expense...plus if the receiver loses "lock" under a bridge, tree or building you're had......then there is "multipath"...)

The little Garmin handheld GPS units are anywhere from +/- 15 feet to hundreds of feet off in precision (Precision & accuracy are NOT the same thing....and accuracy is subjective)....when trains can pass each other with 1 or 2 feet two spare, the uncertainty is too much.....

In time the computer processing and hardware costs will drop, in the meantime we'll just have to wait......Everybody is complaining about RCO's, this makes RCO's look like a minor bump in the road.......

Travelin' Feathers

ps.....(GPS & GIS are not the same either, far too many do not understand the difference.... Technology is a good thing - Failure to understand the technology creates unwanted expectations and/or blunders, as in brain failures that kill.....)

[Anonymous]

What you are really talking about is Postive Train Control, which may use GPS as a component, or may not. It depends on the system. The capacity increasing aspect of PTC is it incorporates a system of "floating blocks," that is, the blocks travel with the trains rather than are permanently fixed to geographic locations.

The short answer to your question is PTC might increase capacity, but at a cost that is very unattractive, so far. The electronic equipment is more expensive than traditional Centralized Traffic Control, which itself costs about $1 million per mile, because it requires equipping every locomotive and mobile track machine assigned to the PTC-equipped territory. The software is very complex and expensive -- think fly-by-wire systems on an aircraft. If a railroad doesn't equip its entire fleet with PTC, it takes upon itself a severe loss of flexibility in fleet management.

The fundamental difference between GPS and railroad methods of operation is that GPS is an approximation (albeit a pretty good one) and railroad operation is yes/no. That is, a train is either approaching a control point, or it has passed a control point. All GPS does is tell you where a GPS transceiver is -- more or less -- which is not at all the same thing as a signaling system. If an aircraft is plus or minus 20 feet while flying, who cares? A train plus or minus 20 feet is on another track or beyond a control point.

I've seen some bold claims of capacity increases with PTC, but there are many skeptics who disbelieve them. No one is rushing to buy it except people using other people's money (that is, taxpayers' money). We might see PTC experiments on high-density freight railroad routes within 10 years.

You asked how much capacity could be increased: no one knows the answer to that question yet. Manufacturers claim 20-30%, but no empirical tests to validate those claims have been performed. And without a huge committment of cash on someone's part, a real-world installation (which would give you some real numbers) isn't going to happen.

[RailroadDoc]

Everybody is well aware of Union Pacific's problems now in Southern California and elsewhere -- too much business, not enough infrastructure to run as many trains as they need to run.

I know that GPS signal control is being investigated -- on the Alaska Railroad, among others.

So here's my question -- GPS train control will allow trains to operate much closer together. Obviously, if you are in a train and another train is on your track say five miles ahead it's probably not a problem if its moving away from you at 60 mph. It's a real problem if its stopped, or even worse, headed towards you at 60 mph.

If you could wave a magic wand and have total GPS train control over a system -- how much would that increase railroad capacity?