Slashdot Mirror
French Train Engineering Giant Alstom Testing Automated Freight Train (bbc.com)
French train engineering giant Alstom is to test automated freight trains in the Netherlands this year. From a report: The automated train prototype can travel for about 100km (60 miles) without driver intervention. Automation will free the train driver to focus on supervising the train's progress. The test's purpose is to provide a live demonstration that the train and the signal system can communicate effectively to drive the train. Alstom signed an agreement with the the Dutch infrastructure operator ProRail and Rotterdam Rail Feeding (RRF) to carry out the tests along the Betuweroute -- a 150km double track freight railway line connecting Rotterdam to Germany.
n/t
Very cool but we already have AI-enabled self driving cars all over the place, and trains are on rails on a controlled environment. Surely this is a solved problem?
drivers are going to feel like horses to the next generation.
"did they really have *people* operating transportation?"
I'm finding it hard to believe that this technology has not been available for a long time. Of course the system requites installation of the signaling systems. Here in the United States, major freight carriers have chaffed at even installing modern safety systems (The more we learn about Amtrak derailment the stranger it gets). But then again, those are "non-revenue generating" while this certainly has the potential. But also remember that unions have a say in manning. Again here in the US, unions fought tooth and nail when rail companies got rid of the caboose.
If you want news from today, you have to come back tomorrow.
lot's of manual switches on rail lines still
Automation will free the train driver to focus on supervising the train's progress.
I've already done that job. Decades ago, I took a ride on a TGV. I was sitting there sipping refreshments and thinking "Are we really going as fast as they say? It sure doesn't feel like it at all". So I looked out the window at the km markers and timed it with my wristwatch. Sure enough, we really were going that fast. I was also monitoring our progress on my paper map.
The thing is, that wasn't a paying position. In fact, I had to pay them a pretty penny for the privilege to do that.
And will biofusion be involved.
We've been through this three years ago (and earlier, but I can't find the links).
As everybody knows, trains' primary purpose is not to haul cargo or transport passengers, but to provide jobs . Not just the drivers (excuse me, engineers), but even the announcers (excuse me, conductors) can not be eliminated.
Automating them will causes them to fail in that primary purpose and therefor can not done. From the Socialist scum to the seemingly respectable Slashdotters, everyone is against that... #ResistOrSomething
This is not a technical problem.
In Soviet Washington the swamp drains you.
Have a look at Wikipedia’s article on rail transport; particularly, in the “history” chapter, the parts about electric and diesel locomotives. It might sway your opinion a bit. There’s been plenty of innovation and research in the railway industry, and I find it rather interesting that its early adoption of the electric motor (beginning with the 1890 underground line in London) was largely spurred by environmental concerns.
are being screwed this way. We need to outlaw automation to protect union jobs.
Leave station.
Is this my turn? No. Is this my turn? No. Is this my turn? No. Is this my turn? No. Is this my turn? No. Is this my turn? ...
Arrive at station.
Done.
Rio Tinto has been working on this on their railroad serving their mines in Australia. They've already run some very long and heavy trains automatically.
Trains are way much more self-driving friendly than cars and even though this is no piece of cake. After (if ever at all) autonomous trains have been tested and deployed at a large enough scale, we could assume that these approaches are reasonably mature at least under certain conditions. That would be the moment to start considering more difficult scenarios like cars. In fact, this is pretty much how the original evolution of these vehicles occurred: firstly, the highly-restricted trains and, only after having that technology working reasonably well, the fully unconstrained version (cars).
Custom Solvers 2.0 = Alvaro Carballo Garcia = varocarbas.
Fully automated trains were first used during the Manhattan Project, to ferry excessively radioactive material from early reactors to enrichment plant. Those trains used 3rd/4/5/6th rails to transmit electricity and commands, furthermore the special track was confined to within Hanford site.
Next, fully automated metropolitan underground railways happened in France from the late 1980s, where stations with fire supression and evacuation facilities are relatively closely placed and they are crewed. If shite hits the fan urban fire services are close by.
The problem with generic revenue service surface railways is that the locomotive driver is also an onboard mechanic. Be it diesel or electrified, locos are really large and very powerful machinery, where thousands of horsepowers are at work and things can /do break. When things break that usually happens in the middle of nowhere, thanks to Murphy.
For example the fuel injectors of 150-300 liter displacement diesel blocks are a nightmare and if a pressurized oil feed pipeline breaks, there is an immediate risk of fire but sometimes you have to press on for traffic safety or not to totally collapse the timetable for a day. Such a decision requires an experienced human mind. On electric locos the porcelain insulators can literally explode if there is a large voltage surge or the pantograph can break off and wreck the expensive gizmos of the "roof garden" if there is a fallen tree branch caught on the catenary.
Thus, command and control automation is one thing to solve for railway,, but one still cannot replace the loco driver guy as the on-board mechanic with an R2D2 style repair robot. If there is a human still needed on-board but predictably bored most of the time, why not include him in the train driving loop, alongside positive control based safety systems?
Furthermore, suicide prevalence is a problem in some places like Japan and Hungary. Really determined people sit down on the rail as the train is incoming, they know it cannot stop in less than 1000 paces. On one hand that's a really tasking burden for the loco drivers constantly expecting when their next victim is going to walk onto the tracks but on the other hand, how do you handle the legalities with a fully automated train? Robot driver may not even notice if people or cattle were hit.
(Criminals sometimes try to cover up murder by placing the victim on the tracks at night. There was such a case in Hungary about 3 months ago. Police are experienced and easily spot that trick, provided the train stopped promptly and the corpse wasn't overrun a dozen times until dawn, thereby mangling any evidence to oblivion.)
The same thing that makes you think this is rife for automation is also the reason why it has nothing at all to do with cars in the slightest (actually it more to do with batch chemical manufacturing).
The stellar record without autopilot is also why there's little reward for implementing such systems and makes it pointless to prioritise trains over cars instead of working on both side by side or rather actually focusing on the car problem first.
The same thing that makes you think this is rife for automation is also the reason why it has nothing at all to do with cars in the slightest (actually it more to do with batch chemical manufacturing).
Trains and cars have many of the same problems, like braking for obstacles and managing speed for turns. They lack one obvious similarity, of course.
The stellar record without autopilot is also why there's little reward for implementing such systems and makes it pointless to prioritise trains over cars instead of working on both side by side or rather actually focusing on the car problem first.
The reason why you don't automate the trains isn't the safety record, it's the ratio of crew to passengers. Since there are many passengers for each crew member, they already make sense. When you get down around buses or cars, there are massive economic gains to be made by eliminating the driver.
What's being lost in this argument is that cars are stupid and their freedom is largely illusory. Unless you have a 4x4, you can only go where the roads go anyway. Even a 4x4 only has limited ability to traverse most types of terrain without a road. Grasslands, OK. Plains, sure. Deserts start to get tricky. Swamps are mostly a no. Forests are totally a no. Can't cross open water unless you've got a thousand horsepower tube frame truggy with paddle tires, and the water must be flat at the time. We should be putting "all" the vehicles on rails (Except in rural areas) which solves the self-driving problem just about completely.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
has nothing at all to do with cars
Evidently, trains and cars are very different, but automating any long-stretch transportation systems does have quite a few aspects in common. In that regard, trains might be considered over-constrained cars and, as such, the "training wheels" for any serious self-driving attempt. At least, in an ideal world where long-term engineering concerns are the most important thing when dealing with these issues. It would also be impractical for other reasons like car manufacturers having little to do with train ones. But my post was mostly meant to highlight the huge complexity associated with these implementations (still problems to account for the much easier scenario) and the way in which proper engineering usually evolves as opposed to extrapolation-, marketing-based promises.
Custom Solvers 2.0 = Alvaro Carballo Garcia = varocarbas.
Trains and cars have many of the same problems, like braking for obstacles and managing speed for turns. They lack one obvious similarity, of course.
No they really couldn't be more different. You're either way overcomplicating train automation or way simplifying car automation. There's a reason why traditional train automation is handled by basic industrial PLCs while car automation is a box full of computers processing the whazoo out of vision and radar sensors.
but automating any long-stretch transportation systems does have quite a few aspects in common
Actually it has very few. The only thing they have in common is going from a to be. You can consider trains over constrained versions of cars, but when you put enough constraints on any problem it gets broken down to a simple yes / no type logic. It's this massive simplification that allows automated trains and metros to run on simple industrial PLCs whereas a car has a boot full of GPUs doing real time vision analysis.
Incidentally this is also the reason why train drivers are at constant risk of falling asleep, boredom combined with even the slightest bit of fatigue is a bad combination.
This kind of replacing-people-is-easy plans usually underestimate the tremendous complexity associated with emulating most of human skills, mainly when dealing with dynamic, complex enough scenarios. Extremely simple problems intuitively solved by anyone in a split second might provoke a catastrophe when a not-trained-on-that-specific-event machine is in charge. Trains are certainly a huge over-simplification over cars, but this is precisely my point: under what might be considered extremely simple conditions self-driving (controlling a big object moving at high speed, sporadically having problems and/or interacting with other moving objects) isn't straightforward. And I am not just talking about strictly technical complexity, but also about cost (to earn what?!) and, much more importantly, risk. Fully replacing your asleepy train driver is far from easy.
Custom Solvers 2.0 = Alvaro Carballo Garcia = varocarbas.
Amazing. The more automation, the better. We won't have to worry about workers cracking under work stress or even fatigue. Automation doesn't have this problem. The issues it does suffer from will improve as technology improves. With the way things are going, this will be the norm. Especially with Uber moving into freights (https://gofreightfox.com/en/uber-freight-impacting-shipping/). Can't wait to see how this goes!
Unless all the track is elevated and there are no at grade crossings where accidents with other vehicles cannot happen, THIS WILL NOT WORK. You cannot trust drivers' lives to your train's computer!