New Technology and The Death of an Occupation

[eastside]

cptrainman

"FTO is an advanced train management system that optimizes fuel consumption based on each train's marshalling and locomotive configuration and the challenges and constraints of the route being traveled. The system calculates the optimal speed profile for the trip using GPS data, a digital track database and advanced algorithms to automatically adjust the train’s speed throughout the trip. By evaluating train length and weight, grade, track conditions, weather and locomotive performance, the system can accurately calculate the most efficient way to run the train and maintain smooth handling. System operations are similar to the cruise control feature in today’s automobiles. "

Opinions?

 

I would look beyond management systems. I think autonomous vehicles are a lot closer than some of the skeptics here think.  Over the past several years DARPA has held contests for autonomous unmanned vehicles to navigate over a course not revealed to the entrants until the actual day of the contest.  In 2005 five vehicles actually completed the 132 miles course through the Mojave Desert.  Programming a car, truck, or motorcycle to maneuver over an unknown, undeveloped road is in many ways a far more difficult task, in my mind, than programming a train.

The 2005 contest
Current urban challenge

[Railway Man]

Txhighballer:  I and many like me agree with you 100%.  But there is a lack of good engineers, and a lack of people who are willing to be railroaders for the wages that will make railroads competitive with other transportation modes, or, keep the cost of rail transportation low enough to retain shippers who will find alternatives to any transportation mode if the price of transportation rises $0.0001 a ton-mile.  Rail rates have fallen dramatically since 1980 but rail market share has scarcely budged, which indicates how much rail pricing and costs had to be slashed just to retain the market share rail already had.  The future is hard to predict, but my best guess is that pressure on rail rates will not abate but actually get much worse.  Thus the interest in finding developing train-handling, dispatching, and train-control automation.

I have looked with equivalent interest and dread at the efforts to automate dispatching offices.  My experience from the office is that about 1 in 10 dispatchers are outstanding, another 1 in 10 are good enough most days, 6 are mediocre, 1 is bad, and another 1 is wretched. Mediocre is no longer sufficient as the need for reliability, efficiency, and cost reduction becomes more intense every day.  I seriously doubt it is possible to create a dispatching automation tool that is any better than the mediocre dispatcher, and frankly I think it will be somewhat worse than that.  But given the increasing difficulty in finding and retaining good dispatchers (the good ones soon get promoted or find a better-paying job in another industry), the quality is slowly sinking and I doubt there's any way to reverse it.

RWM

[txhighballer]

 

A good engineer can handle a train better than any known or ever to be built computer. Instincts, knowledge of the power you are handling, and knowledge of the road can never be duplicated by a bunch or circuits.

[trainfan1221]

I didn't get the impression from the OP that the idea was to eliminate an engineer's position, just to have more control over how the train is run. Either way I agree that there can be too much technology and there should always be people guiding a train, especially a freight. And I don't think there should be less than two people running it either.

[Railway Man]

Paul:

Three steps happen before a train turns a wheel:

1. Dispatching decision is made: train is given priority, direction, and time of movement.  All of these are "dispatcher decisions" and are on-going decision 24/7/365 for all the trains on the system.  The decision matrix that adjusts each train in relation to all other trains is constantly readjusted.  This step has only been successfully automated at the simplest and most basic level, e.g., simple meet/pass logic for a single-track railway, and following moves in a double-track subway-type system.  Level 1 -- The meet/pass logic simply looks at which train will reach the siding first, and it will take the siding.  Level 2 - Determine if either train cannot take the siding (e.g., weight restrictions).  Level 3 -- Are there grade crossings in the siding that cannot be blocked.  Level 4 - Determine which train has priority and it holds the main.  Level 4 -- compare the hours of service remaining for each crew (don't hold the guy who's about to hog out) versus any penalty for service commitment for each train, and compare the financial costs.  Level 5 -- look at effects on the next following train each direction.  Level 6 -- look at effects on the next trains back.  Level 7 -- look at effects on motive power and crew utilization.  A very good dispatcher reaches back to level 20 or so.  The software usually blows up when it tries to think about level 4.
2. Train-control decision is made:  Think of this as the ABS-APB system.  It asks, OK, now we have a priority, but is it safe to move the train?  APB-ABS logic is very simple, and it's effectively been automated at a naive level since the nineteen-teens.  But it only looks at its instant case in front of it; it does not look beyond the adjacent block.  In other words, each block is only informed by the block on either side of it.
3. Train-handling decision is made.  This is what the engineer does.  He has an authority (the dispatching decision); he has a signal or TWC (the train-control decision), now he accelerates, decelerates, and maintains speed accordingly.

PTC as it is now envisioned is simply a safety overlay on Step 2 (train-control) and Step 3 (train-handling).  It doesn't know diddly about priority or direction or juggling competing economic needs, which are accomplished in Step 1.  In the future, there will be stand-alone PTC which will simply envelop the existing ABS-APB system of Step 2, but fundamentally all it does is replace the field hardware with other hardware that does the same exact thing (but at lower cost for buying and installing the hardware).  No additional automation has occurred.

The simplest place to automate is Step 3, because automation at this step doesn't have to know anything about Step 2 or Step 1, it just has to obey its rules. In other words, it consists of blocks of software.  Block 1 is a recapitulation of GCOR or NORAC or whatever rule book applies.  Block 2 is a recapitulation of the railway's train-handing rules, using real-world inputs about train TPOB, braking horsepower, the vertical and horizontal profile en route, weather and temperature, and some expert-level decision-making matrices.

It's possible to automate Step 1, 2, and 3, but only in a very simple, very repititious system, where train operations are about as complicated as dropping identical marbles down a tube into a bucket, one after another.  Basically, that's a description of the Pilbara iron-ore lines.  Trying to do this on a complex corridor with variable operations, where all the marbles are different sizes, weights, densities, smoothness, and shape, and they all want to end up in different buckets, and there's some trying to climb back up the tube ... that's a little harder.

RWM 

[Paul_D_North_Jr]

Railway Man

(PTC is not an automated train-control system . . .  FYI).

RWM

What is the difference, then, aside from the semantics ?  I infer that PTC is primarily restrictive only - it will tell or make sure the train slows down or stops as much as is needed to assure safey, but doesn't do anything to tell it or make it 'Go'.  In contrast, an automated train control system would have that latter 'Go' function equally important as the 'Slow' and 'Stop'.  Is this correct ?  Or, is there some other nuance that I'm mmssing here ?

- PDN. 

[Railway Man]

BMLP was a much bigger reach than what we're talking about here.  It was not only a an automated train-handling system but also an automated dispatching system AND an automated train-control system, all in one package.  Radio and processor technology was not sufficiently sophisticated to manage all the data flows, or even know if it had a valid data flow, in the 1960s when this was designed.  (PTC is not an automated train-control system or an automated dispatching system, FYI).

RWM

[Paul_D_North_Jr]

Stated another way, it provides guidance to the locomotive engineer to operate the train much like your grandmother drives her car - no 'jack-rabbit starts' but instead accelerating slowly and gradually from stops.  At signals, easing off on the throttle long before the red light and coast up to it, avoiding using the brakes - and maybe it'll turn green in the meantime, so a full stop won't be needed.  In between, cruise at a speed to take advantage of the next opening in the traffic pattern, without having to go too fast or too slow or delaying the following traffic, etc.

I believe the BMLP automation had to be discarded, shortly after it started - not sure why.  The only fully automated such operation of which I'm aware was the now-dismantled Muskingum Electric coal line in eastern Ohio, which was a much shorter run and in a more-closed environment, if I'm not mistaken.

- Paul North.

[Railway Man]

Ulrich

Technology is great...but the human capacity to think and to solve problems is far greater, and we're a long way yet from replicating that in technological form. I don't see it happening for a number of reasons that have already been stated...however one person crews are a distinct and likely possibility.

 

Automation is possible in some simple operations...the Black Mesa operation in AZ for example was automated 30+ years ago...but these are exceptional operations that involve simple routing much like your HO trainset.

Just to make sure we're not talking past each other, the software I discuss above is a speed-control system only.  It picks the best speed for network fluidity, train handling, and fuel economy for the given train, the specific location, and the traffic ahead and behind.  It's already in place and working on a number of railway systems.

RWM

[Ulrich]

Technology is great...but the human capacity to think and to solve problems is far greater, and we're a long way yet from replicating that in technological form. I don't see it happening for a number of reasons that have already been stated...however one person crews are a distinct and likely possibility.

 

Automation is possible in some simple operations...the Black Mesa operation in AZ for example was automated 30+ years ago...but these are exceptional operations that involve simple routing much like your HO trainset.

[Railway Man]

 Here's the summary:

1. The train-handling software already exists
2. It's not perfect, but it doesn't need to be.  It just needs to be good enough to achieve a positive cost-benefit ratio and productivity improvement.
3. 90% of the cost-benefit economics can be summarized as loss of main track capacity (some) against increase in fuel economy (quite a bit), and ability to have more predictable operations.  Loss of main-track capacity includes: (1) there will no longer be any "hot" engineers (2) the software will be more cautious because it will assume more worst-case scenarios (3) offset against this is there will no longer be any slowpoke or incompetent engineers, either; (4) and fewer bad-handling break-in-twos.
4. It's reasonable to expect to see this rolled out immediately after PTC implementation
5. A significant cost of doing this is installing more wayside detectors -- a lot more of them.  That's figured in to #3 above.
6. Dispatching automation will be implemented if not simultaneously, immediately afterward.

No-man is probably not economically feasible without a secured corridor (no at-grade crossings, fenced).  That's way off because grade-separations are very expensive.

The goal is increasing productivity. 

RWM

 

[Paul_D_North_Jr]

System lost my last reply here, so I'll try again . . .

henry6, it's called 'SOP' - for 'Seat-Of-Pants'.    I'm sure our locomotive engineer members here can tell us more about that skill set.

I wonder, though, if it will allow sometimes useful but nevertheless frowned-upon techniques such as 'power-braking' or 'stretch braking' to continue to be used ?  To what extent can the system be 'over-ridden' to do that ?

NS thinks it will gain from saving 5 million gals. of diesel fuel a year - at 3.00 per gallon, that's saving around 15 million dollars a year, or about 42,000 a day, for each year going forward.  I don't know how many 'train starts' NS has on a typical day, buit with each modern unit consuming 150 to 200 gals. per hr. in Run 8, the savings per trip are several hundred to maybe a thousand dollars, 'depending'.  And that's even with still keeping the engineer onboard and on the payroll

-  Paul North.

[henry6]

Technology is great but removing the "human factor" from many procedures takes a lot away, often things not understood by technogeeks (for lack of a better term).  Digital music is a good example.  Although the sound reporduction is perfect most musician (and quite a few listeners) will agree that something is missing; and that is "air" or "atmosphere" or "ambiance" or something to that effect which only the human ear can percieve but something that makes the sound "whole".  The same goes for automating moveing objects like trains: there has to be a human element: eyes, ears, nose, and emotion or intitution (again for lack of better terms) to allow the technology to function at its best yet be safe for the equipment itself, property, and people.  So we propose a human aboard to baby sit and override the system.  But this human cannot be just a baby sitter; he or she has to have as much skill and acumen and knowledge as the technology and thensome.  So have we actually gained anything for the additional expense of technology, besides witholding taxes, social security payments, and a health plan? 

[tree68]

It's going to take time and money to develop the specifics for each route.  The return on investment will have to be there.  Thus, on a heavily travelled line, you might see something like this implemented.  On that twice-a-week line, not so much.

The technology has existed for years to automate our highways - I saw it demonstrated in the 60's and things have certainly progressed since then.  But I have yet to see it implemented.

[Paul_D_North_Jr]

As wabash1 said, I believe NS has something similar - I'll see if I can find a post a description or link.

EDIT - ADD;  It's NS' 'LEADER' system, as follows from the March/April 2009 issue of BIzNS, Vol. 1, No. 2, at pp. 19, right col. [21 of 28], and 22, right col. [24 of 28], at - http://www.nscorp.com/nscorphtml/bizns/bizNS1-2.pdf 

''Keeping an eye on fuel consumption also is essential to NS’ success.  One way the company is doing that is by installing a computer system on its locomotives called Locomotive Engineer Assist Display and Event Recorder, or LEADER.  Equipped with a computerized track profile that includes such things as grade, curvature, and track speed, and using satellite-based GPS coordinates to pinpoint location, the system is designed to provide engineers with train-handling instructions for maximum fuel efficiency.  The system generates instructions, such as optimal speed, that engineers can use in handling the train to achieve maximum fuel savings. 

“LEADER is all about fuel conservation,” Lawrence said. “We burn nearly 500 million gallons of fuel annually, so every percent we can reduce consumption is about 5 million gallons.  At today’s cost of fuel, that adds up to a lot of value, plus it makes Norfolk Southern even better from a sustainability perspective.” 

“Using new technologies to better manage our assets to operate more safely and more efficiently will position us to continue to be the leader in the transportation industry,” Lawrence said. “We will continue working on newer and better ways to benefit our customers and our shareholders.”  BizNS

Locomotive Engineer Assist Display and Event Recorder

''LEADER helps our engineers make better train handling decisions by providing real-time coaching on performance against an optimal “golden run” for a route. LEADER monitors the train’s location, track topology, speed, acceleration, and in-train forces and recommends optimal operations for throttling, braking, and minimum and maximum speeds.''

It's more like an auto-pilot for aircraft, to take over most of the routine in-flight stuff.  You still need the pilots for the takeoffs, landings, bad weather, system breakdowns, terminal maneuvers, and extraordinary events - Capt. Chesney 'Sully' Sullenberger [sp Q] landing the US Airways plane that hit the birds and lost all engines in the Hudson River, and the Continental Airlines pilot who died in mid-flight over the Atlantic last month are two easy examples.

Anyone who has dealt with a GPS navigation system that tells you to "Turn here !" will also understand why.

The 'worst case' for that system would be something like a poorly made-up train with heavy cars on the rear, on a hog-back profile, with lots of high-profile cars in a heavy quartering wind, maybe a rainstorm or snowstorm to add to the fun, and then an undesired emergency brake application - a 'dynamiter'.  You'll still need an engineer to put that train back together after the knuckles are broken.

- Paul North.

[wabash1]

that program has been out there for years not to worry it wont replace engineers

[Awesome!]

Explain to me how its going to work in 3% grade and curves at the same speed? What about crossing thru a town?

[john_edwards]

 Just like remotely piloted aircraft, its possible but it will never happen in our lifetime.  Well at least not in this country with the ratio we have of lawyers to people (sorry gabe).

[cacole]

 There's no way freight trains will ever become remote controlled as long as there are grade crossings and the potential of trespassers on the tracks.   A human being will still be required to watch for hazards and take corrective action as necessary.

[BaltACD]

How will FTO respond to the uncoupled air hose that hit a road crossing 100 cars from the engines and inspect the train after it has activated a wayside defect detector?

[cptrainman]

We all know of passenger operations that are completely automated. Two that come to mind are the shuttle between terminal 1 and 3 at Beijing's International airport and the LRT in Vancouver, B.C. I had always thought that freight trains would be safe from this at least in the forseeable future since track profiles and train consists are so variable. Well, we are one step closer to saying good bye to the occupation of locomotive engineer.

A news release came out on CP's employee website describing a new technology named FTO (fuel trip optimizer). Basically, it is cruise control for freight trains and the purpose of this technology is to operate a train with as little fuel as possible. The software is an upgrade to the current software on GE's Evolution series locomotives.

Of course we all know about PTC. That technology stops the train. FTO drives the train.

 Here is a snip from CP's news release that better describes the technology.

"FTO is an advanced train management system that optimizes fuel consumption based on each train's marshalling and locomotive configuration and the challenges and constraints of the route being traveled. The system calculates the optimal speed profile for the trip using GPS data, a digital track database and advanced algorithms to automatically adjust the train’s speed throughout the trip. By evaluating train length and weight, grade, track conditions, weather and locomotive performance, the system can accurately calculate the most efficient way to run the train and maintain smooth handling. System operations are similar to the cruise control feature in today’s automobiles. "

Opinions?