Robo-Shinkansen rolls slowly – for now – across 5km of Japan Japan's largest train operator, the East Japan Railway Company, has successfully trialled an autonomous bullet train. The Wednesday test run was short – travelling just under five kilometres down the Joetsu line from Niigata Station. Nor did the train reach top speed of 300km/h, maxing out at 110km/h. But the train did come …
Bay Area Rapid Transit (BART) in the SF, CA, bay area, initially had plans for fully automated trains in 1972. Early on, during a test run ...
During the first months of revenue service, the Automatic Train Control (ATC) system had safety problems with its design and operation. On October 2, 1972, an ATC failure caused a train to run off the end of the elevated track at the Fremont station and crash to the ground - an incident dubbed the "Fremont Flyer".
Fortunately there were no passengers and nobody was hurt.
I have a friend now who is a retired BART driver, and he told me they have a picture of that accident up in the drivers room where he started work every day for the many decades he worked there.
I have a friend now who is a retired BART driver, and he told me they have a picture of that accident up in the drivers room where he started work every day for the many decades he worked there.
Do they also have pictures of London Moorgate on 28/2/75 and Ladbroke Grove on 5/10/99?
The rule book, and some of the automation too.
Here in London the more mechanically operated underground lines have trains with pneumatic lines that keep the brakes off the wheels. If the line breaks and pressure drops, the brakes are engaged automatically. Useful in case of a fault but it’s now also used in case a driver runs a red signal.
On these lines a red signal also raises a lever next to track that opens a valve (a “trip cock”) on the pneumatic brake line to enforce the brakes should a train run a red signal.
This system is called “Moorgate Protection” after a 1975 accident at Moorgate station where a driver, for reasons that have never really been uncovered, didn’t stop at the end of the line and ploughed the train into a wall at full speed.
That accident spurred the development of technology to take the breaking after a red signal out of the driver’s hands.
Trainstops and tripcocks have been around a lot, lot longer than that - I was a LT signal engineering apprentice in 1971 and this equipment had been around for decades, even then. What Moorgate did do was to cause the instigation of a rollout of formalising speed limits for trains entering terminal stations, and putting trainstops at the buffer end of the station to apply brakes before hitting the buffers.
>That accident spurred the development of technology to take the breaking after a red signal out of the driver’s hands.
Trouble is that such technologies aren't widely deployed on the main rail network. Both the Ladbrooke Grove (aka Paddington rail crash) in 1999 and the Salisbury crash a few weeks back, could have been prevented by track protection features intended to stop trains over running red signals...
The driver in the Salisbury crash has been shown to have acted properly and applied the brakes. In fact he beat the automatic system to it. The train crashed because of slippery rails. The classic leaves on the line. They form a thick paste that acts like a lubricant so the train braked its wheels and effectively skidded into the other train. It’s less well known that it can affect road vehicles on junctions if the carpet of leaves is thick enough or isn’t regularly cleared.
It should be redundant to mention that 1972 was 50 years ago and a lot of water has flown under the technological bridge since then. We've been prototyping self-driving cars and compared to the problem of navigating a street driving a train is trivial. Since the sensors on an automatic train can 'see' better than a human -- humans can't see where they're going much above 200Km/hr (125 mph) -- so an automatic system is likely to be better at driving than a human. Currently driving a TGV has the driver responding to in-cab signals telling him/her how fast to go with any real driving confined to legacy track and stations.
There's still plenty of real driving to do, though. Nursing a half mile long freight train through sharp curves and steep gradients is skilled work and it probably doesn't make sense to try to take those drivers out of the cab. Passenger routes, though, are getting like elevators.
As someone who has spent almost 50 years of working directly with cutting edge technology I can state that I have even less faith in the ability of 2020s technology to get me to my desired destination than 1970s technology.
The 1970s version would have been written when engineers were real engineers and small furry creatures from Alpha Centauri were real small furry creatures from Alpha Centauri.
The 2020s version will probably comprise of scores of layered systems all containing millions of line of code written by the lowest neo-liberal sociopath conglomerate bidder.
This is also why I won't be taken a rocket to Mars anytime soon (amongst other reasons).
If you visit the London Docklands, you will find self-driving trains that have been in operation for about 34 years now, and those trains are the most reliable in the UK.
Also in London, the Victoria Line was opened about 53 years ago, and it is almost self-driving. They do have drivers, but it would take very little effort to make it completely self-driving.
It depends on who you get to implement the system. If you take a bunch of RAD types and let them loose on a safety system then you're courting disaster. But you don't -- you use specialists who know how to design reliable systems. There are innumerable examples in daily use.
Fallacy of the new. Modern systems are much more fragile in a lot of ways too.. look at the car ignition key. They used to routinely last 40-50 years... And cost under a buck. Now several hundred bucks... And I've had to replace 3 in 7 years.
Some problems aren't meant to be solved by increasing tech. Think of the additional carnage the increased speed of this train vs the ones of the past could cause. And no one knows all the variables that the drivers actually juggle. Which is why precious attempts have failed.
Don't even get me started on car keys.
My 2006 commuter car has the keyfob (made of thin plastic, as you do) connected to the actual key. The car can detect if the fob is close to the ignition key slot. If the car turns over and the fob isn't present after a second, it kills the fuel pump and the car stops. So, no aftermarket keys, no spare keys without the fob attached, and no using your own car if the bulky fob breaks off like mine has. It's currently secured together with superglue and packing tape and seems it will not last much longer.
And yes, they no longer make the fobs anymore, and it does not appear to be user-serviceable or programmable. Not that that will stop me if my hand is forced...
I get that this is for security, but I would rather my car do the engine starty bit when I tell it to, thank you.
The previous owner doused it in enough superglue to give it a nice glazed donut look, but that hasn't stopped one half of the fob body from breaking off, which is what holds the key neck. My only option was to further douse it and tape it up for good measure.
This also means that should the battery die I'll have to try and carefully de-gunk it or I'll be SOL.
Runs on tracks. It can't steer, has well defined speed profiles and limited hazards. It doesn't even have to recognise signals, since they could be delivered directly to the controller... isn't this rather an easier problem to solve than a self-driving autonomous car or lorry?
Short haul robotic trains exist: some airports, and of course the Docklands Light Railway, but I'm surprised this hasn't been done before.
Perhaps I'm missing something obvious?
The big difference is the speed. As, for a constant mass, kinetic energy increases as the square of speed, the difficulty of stopping this one safely is vastly greater than for the DLR.
Which is why it is much better done by a computer than by a human. After all, Space X doesn't land its boosters manually, and they have even more KE to play with.
I think that anyone who has seen the sheer perfection of the Japanese train system can see that there's little need for automation from a performance point of view. Japanese train drivers routinely run their trains to within seconds of timetable.
If they ever go for full automation, the biggest challenge they face with removing train drivers from cabs in Japan is suicide detection. It's quite common for people to top themselves by jumping in front of a train. They'll need a sensor for that, somehow or other. Pretty grim, really.
"The big difference is the speed."
I'd have thought that would be the ideal reason to automate. The train already goes so fast that the line has to be secure. No people, vehicles or animals allowed to cross the tracks. Likewise, no dangerous curves, or junctions other than in known places. Also, the entire length of track is a known factor so maximum safe speed can be maintained for any specific part of it. If anything, it should be faster and safer than with a human driver. And, IIRC, the driver is already monitored to make sure they are paying attention at all times and informed by computer about what is coming up and what to do.
Perhaps I'm missing something obvious?
Being slightly more serious for a moment...
The problem is cost and complexity.
First off, all the examples quoted are practically, isolated lines/systems: They can guarantee that all the trains that run on those lines will be under the control of the signalling system. If you mix in other traffic, then that makes the system way more complex.
But the big reason is the cost of the signalling system. High Speed Train == Long Distance Journeys == Lots of signalling equipment,
If you're building a line from the ground up to be automatic train control, then the cost will be lower. But upgrading an existing long distance line will be very expensive. (Just look at the London Underground 4 Lines Modernization Program. That is, slowly, upgrading some lines to automatic train operation. But look how expensive that is. (And how much it's running over time)
Trackside signals are very difficult to see in time at high speeds. And the fixed block sizes needed would mean huge gaps between trains to be safe.
In-cab signals create moving blocks, with the signals always being just in front of the train, whatever speed it's travelling at. More trains per line, almost zero feedback between a signal changing state and the driver/train being aware of that.
Autonomous cars would be a lot easier to make fast AND safe if they ran on tracks, had some sort of network-wide signalling system that knew where they all were, and could be coupled together to form "trains" that could carry hundreds of passengers in one "vehicle".
At HST speed you can't rely on trackside signals... the TGVs, the Acella, the ICEs all have in-cab signalling, it's just that each country has it's own in-cab signalling system.
That's why the Eurostars/Thalys are not standard TGVs ( they need to deal with the UK systems ) and why the French Railways refitted a few TGV with Spanish in-cab system to send them towards Barcelone... that's also why only a few ICE are able to reach Paris.
ERTMS ( aka Eruopean Rail Traffic Management System ) is meant to be the system to tie them all together.
actually there was several reasons for the Eurostars and Thalys to be somewhat different from a normal TGV.
The main reason is not the signaling... it's how it gets its power and under which form :
- From an overhad line ( voltage/amperage/frequency varying depending on location in France[ just for France we have 25KV/50Hz, 1500V DC as main types ], but also in Belgium and Netherland )
- From a third rail ( UK, initially before HS1 was built with 25KV/50Hz [IIRC] overhead line )
The various ( French [ Several systems ], Belgian,.. ) signaling systems are just a tiny bit compared to the above.
Perhaps I'm missing something obvious?
Most railway lines are not isolated point-to-point lines like the airport shuttles. They have S&C and integrate with other lines. You can't just upgrade one line - you'd have to upgrade every line it touches so that it has full visibility of the system (or at least enough of that joining line that you can see incoming trains usefully far ahead).
It's a struggle even to get in-cab signalling working because you generally need to join up the local signalling blocks into a single omnipotent system-wide view.
They tried to do in-cab signalling during the WCML upgrade and abandoned it because it was too complicated getting all the disparate signalling systems along the length of the line to feed into a single platform which could transmit relevant data to the cab. They ended up rolling it out on the Cambrian line first.
By contrast, HS2 will have in-cab signalling (ETCS/ERTMS), partly because it's essential at 200mph (you whip past trackside boards too quickly to read them) and partly because London-Brum is your idealised, isolated line.
CrossRail and Thameslink also use in-cab signalling (CrossRail also has entirely automatic operation in the central zone) but drops down to more conventional AWS/TPWS after it passes Paddington.
There is - notionally - a national rollout of ETCS, and once you have in-cab signalling then you're halfway to full train control, because you have a direct data-link onto the rolling stock. But the rollout is slow and complicated. Perversely it's often best to start on the little branch lines, get them working and build experience in operating ETCS and then once you come to the mainlines all the stuff attached to it is ready to go.
Shinkansen are dedicated fairly straight lines that are mostly isolated from the rest of the network. This last point is key in regards to signalling needs (all in cab) and also safety from the point of view of third parties breaching the line security. You are unlikely to find a truck stuck across a level crossing on a Shinkansen because there are none. That probably means that collision detection needs are probably more achievable requiring less human like interpretation of the scene and the chances of a possible collision or derailment are significantly reduced compared to a standard rail track.
Trains are certainly the easiest transport to automate. However it is worth considering the role of a driver during an accident.
For example, when a West Coast Pendalino derailed at high speed some years ago, the driver was credited as having skillfully used the brakes to keep everything upright an in a straight line for as long as possible.
I've no idea what range of influence a driver has in those circumstances, but I can easily see that choice of braking strategy is going to depend on the nature of the derailment. Eg front carriages on the rails, back ones off; slam the brakes on in that circumstances and the back of the train will try and overtake the front. That's the kind of thing that a human can adapt to, and a machine cannot.
Japan's approach to rail safety is interesting. They are totally fastidious about maintenance and checking for faults. Eg the entire length of the Shinkansen track is checked every night (I think) by Dr Shinkansen (a yellow track measurement train). The train sets are very well cared for. They have an excellent signalling system and very well trained drivers and crew. All this together they have deemed adequate to prevent collisions and derailmemts, and so they don't design the carriages to take collision and derailment impacts (or so I hard)
I was working in the LT Mechanicals development dept when we were testing various incarnations of a wheelslide detector produced by an external contractor. This was one of those projects which soaked up seemingly endless amounts of resources because of the difficulty detecting when wheels stop turning due to leaves on the line, or brakes applied when the rails are wet, causing the wheels to slide, causing undesirable "flats" on the wheels. Sounds like an easy problem, but it's not when one of the stipulations of the design is it's not to interfere with any safety mechanism.
In the UK at least the trains are highly unionised. The same crews will be working on the same trains whoever wins the contract to run and services on a line, likely as not on the same terms and conditions.
Look at the trouble Southern Rail had just trying to get the driver to close the doors, apparently deemed unsafe, except the exact same tech was used on many other rail systems such as the underground with no issues at all. In fact Southern Rail used on some services.
I would imagine it will be many years before self driving trains will ever be introduced on any existing railway in the UK despite the many instances of bad driving we see every year on the rail networks.
9 times out of 10 it is not the technololgy that stops thing its the potential job losses and loss of election votes.
"But the train did come to a complete stop just 7.5cm from its intended stopping point – a result the railway deemed very promising, given the Shinkansen has a 50cm stopping zone"
Wow! it worked. This time.
I'd only be even cautiously satisfied if it did this hundreds of thousands of times consecutively from its maximum design speed with no anomalous events.
"The trains will be autonomous, but staff aboard the vehicle will have sufficient skill to intervene if needed. Those workers will not, however, be fully trained Shinkansen drivers – a reflection of the Railway Company's confidence in its automation and recognition of difficulty recruiting and training drivers." does not inspire my confidence. In reality, they'll either be skilled enough to drive the train or they won't. The only advantage of the entire idea I can see is that it'll be cheaper than employing drivers - unless it's just a dream of the autonomocrats that believe all humans (except themselves of course) are fundamentally redundant.
They'll probably be trained enough to drive it to a station, but only under remote supervision and only at reduced speed.
Being talked in by Lloyd Bridges San, who keeps saying things like "it looks like I picked the wrong week to give up drinking Sake"
You are not aware of the fact that on the London Underground's Central Line drivers are not even allowed to voluntarily run under manual operation ("Coded Manual"), unless it's a Sunday and they are outside the central zone?
It's simply a result of human drivers not being able to safely run the trains at the short spaces and stopping distances required in central London. I remember a couple of years back when there was a computer failure on the Central Line and every train had to fall back on manual operation. Instead of a train every 90 seconds it became a train every five minutes (wholly inappropriate for central London, causing dangerous levels of overcrowding) if you were lucky. A train every ten minutes was more common.
There are a couple of problems with human drivers. For one, they can only see the train ahead of them, if they are lucky. A computer can see all the fifteen trains ahead of them and pre-empt a "bunch-up" because six trains ahead there is a slower train. It also turns out to be quite difficult to train drivers on stopping on time. On the Central Line a train is expected to come into the platform at full speed and only initiate braking roughly halfway along the platform. Human drivers tend to find that unintuitive and feel they need to brake when the station is approaching. That makes running railways with over approx. 25 trains per hour (tph) practically impossible. Some metro systems run up to 36 tph, thanks to computer operated trains. Finally, computers are never hungover, had a poor night's sleep or in the middle of a messy divorce. That doesn't mean the Central Line doesn't have people referred to as "drivers" in the front of the train. They are vital in deciding when to close the doors and when to tell the computer it's time to depart, and vital in evacuating the train in emergencies. It's only the train movement that's handed over to computers.
On main line railways there are other things to bear in mind than in metro systems. Level crossings for instance. It would come as no surprise that all main line railways experimenting with ATO (such as the Shinkansen or the Dutch dedicated freight route "Betuweroute") are fully grade separated.
That said, even main line trains have been relying on machines for decades. A stopping distance of 1500 metres is not unusual for some trains, far beyond the visual range of a human driver, so they rely on signals to warn of problems far ahead. It's also why train drivers are "route-trained", they are typically not certified to drive trains on routes they haven't been trained on. Because they need to know minutes in advance that a sharp bend is coming, or to expect a signal two kilometres ahead.
In short, it will be quite some time before many main line trains will be running without people in the front, but it has also been quite some time since trains were run without relying on computers to make everything work safely.
The high frequency of some London Underground services has another complexity: At the end of the line where the driver has to change ends, there isn't enough time for the driver to walk to the other end of the train. Instead, another driver is waiting at the platform ready to step into the rear of the train to take it back out.
Amsterdam’s newest line uses automation to solve this issue. Instead of the “stepping back” that you describe where drivers move to the train behind them and end their shift a couple of back of the one they started on, Amsterdam uses a system where the driver takes the same train.
At the terminus island platform the driver gets out at the far end and crosses the platform diagonally to the other end. Meanwhile the computer has driven the train out of the station into a siding, reverses it there and then drives the train to the opposite end platform where the original driver has arrived at the front to take over his original train.
Theoretically this mean you might use one driver fewer in your rota than in the London situation.
>Meanwhile the computer has driven the train out of the station into a siding, reverses it there and then drives the train to the opposite end platform where the original driver has arrived at the front to take over his original train.
Pointless time-wasting train movement.
When I wrote a utility for "Stepping Back" (as it is known internally) for the computer system at Cobourg Street, the minimum layover period was 7 minutes, which allowed the driver to change ends and have a PNR (Physical Needs Relief - fill up his teapot and have a pee, hopefully as distinct operations). Stepping Back naturally halved that period. ISTR this was only needed at Morden and Brixton for the lines controlled by Cobourg Street.
A good example is Elephant & Castle (Bakerloo line) which seems to operate Stepping Back even outside rush hours. This is because there's only two platforms to effect reversal there, whereas at the other end there are two sidings at Queens Park, I think two at Stonebridge Park, and one at Harrow & Wealdstone. A shutdown at E&C is the worst nightmare for that line, though trains can terminate short at Waterloo.
Apologies if I've already related this story...
When the Jubilee Line first opened it ran between Stanmore and Charing Cross. I was working for the Mechanicals Dept at the time. One day we ran a test train of 1973 tube stock up to Amersham and back into London via the Jubilee Line. All went well until the train inexplicably applied emergency brake running into Neasden. The Operating Dept took a very dim view of what they thought was us "playing trains". We were accused of causing a Major Delay which continued through the evening peak hour. Some time later, days, months, we found out we did the Operating Dept a big favour.
It seems that a large percentage of Jubilee trains reversing at Charing Cross had to have their north-end tripcock reset in order for the train to depart (the driver has to do this manually, which is time-consuming). The rolling stock running on this line did not detect "Back-Tripping", whereas 1973 stock both tripcocks are pneumatically in-circuit when the driver's key is activated at either end of the train.
It should be noted that trainstops not relevant to the signalled direction of travel should be lowered prior to the passage of a train, such that they do not engage with any tripcock on a train. It seems that, at Neasden, there was a trainstop which was taking a long time to lower, and our test train fouled it, causing our emergency brake to operate.
In the first place, I severely doubt that human drivers are currently stopping Shinkansen trains manually. The trains stop so that each door in the train is aligned with a gate on the station platform. If they are not aligned, people cannot get in or out of the train. Anybody who has driven a train will tell you that this precision cannot be obtained reliably by a human.
Only some shinkansen stations have gates and barriers between the platform and the track. I don't think any of the minor shinkansen stations do, and not all major stations (one(ish) per prefecture) do either. Hell, the Tokyo Metro probably has more.
But yes, they do always stop at the designated spots.
I severely don't doubt it. ATO was introduced on the Jubilee Line a decade *after* platform edge doors were introduced on the extension to the line - provided the train was stopped within +/-50cm of the desired point, the alignment between carriage and platform doors was good enough to allow them to operate, and typically drivers would get the alignment much closer than that, typically by one main application of the brakes to slow right down to a crawl (or occasionally to a full stop followed by a brief reapplication of power), and then a second smaller application to hit the stopping point.
Given the article also mentions a 50cm zone (though not whether it means a zone 50cm long centred on the stopping point, or one zone with 50cm tolerance either side of the point) being currently applicable, I suspect the Shankansen drivers are just as capable of stopping their trains sufficiently accurately as their TfL counterparts.
Elsewhere on the LU network (and I'd suspect on other networks as well) trains which remain fully under the control of human drivers are also sometimes required to align pretty accurately with a fixed stopping point in order to ensure all the carriage doors that are supposed to open at that station are actually aligned with the platform, even though the platform itself doesn't feature doors/gates...
" In reality, they'll either be skilled enough to drive the train or they won't."
The ideal aircraft cockpit crew is composed of a man and a dog. The Man's job is to take the controls if something goes wrong with the automation and the Dog's job is to bite the Man if he attempts to take the controls
The reality of human supervision of near-perfect automation is that attention wanders, people get bored and start doing other things than what they're supposed to be doing
It's probably better to layer extra supervisory monitoring systems in (dumb ones with simple tasks) that will stop the train if things go amiss (including constant online montoring going offline) and have sufficiently trained staff onboard to be able to drive in "limp home mode" to the nearest safety point from there if it can't be done remotely (we're at or nearly at the point where "waldoing when required" is feasible around most of the world and definitely so in places like Japan)
At some point, actually HAVING humans continuously in the loop _is_ the most dangerous part of the whole operation and your safety systems have to be designed against bored&redundant people fucking around or falling asleep
A friend of mine is a psychologist specialised in psychometric testing for recruiting for highly-specialised roles. Among other things he has been involved in developing testing for train driver recruitment programmes.
His insights into train driver requirements are quite interesting. For instance, you need someone who is able to stay concentrated staring out of a window for hours on end and still be able to respond within a second to a light signal along the track. No day dreaming or wandering minds. They also need to be comfortable with barely interacting with other humans for hours a day. Just those two requirements rule out huge swaths of the population.
The ideal aircraft cockpit crew is composed of a man and a dog. The Man's job is to take the controls if something goes wrong with the automation and the Dog's job is to bite the Man if he attempts to take the controls
You got the joke wrong. The man is there to feed the dog.
Low cost carrier Thromby Air tries Canine First Officers.
(It's just a single-panel comic strip, folks, not a real airline.)
And then you have situations and captains like US Airways flight 1549 and Captain Chesley 'Sully' Sullenberger who lost propulsion when a flock of birds entered the engines and landed the plane in the Hudson River - safely.
And don't forget that automated systems are designed by humans, for example Boeing's outsourced engineers implementing the infamous MCAS control system and the finance division overseeing that work.
Agreed, it's usually not the driver's salary that makes Automatic Train Operation attractive (although it's said that being a driver on the London Underground is among the best paying jobs available for people without a university degree because of the required skill sets and responsibility put on them).
Computers have other benefits compared to human drivers, though. They are more reliable and predictable, can work 24 hour shifts and don't require complex staffing schedules, are never on strike and never ill. That has cost implications too, though a bit harder to calculate.
are never on strike
Because the London DLR never stops working when staff go on strike.
www.londonreconnections.com/2021/the-political-myth-of-the-driverless-tube-train/
Like the statement:
"What this means is that the more complex the transport system, the more learning required and thus the less suited it is to automatic operation. "
Given how everyone goes on about how much simpler the environment trains operate in compared to cars, it implies cars are even less suited to driverless operation...
Unfortunately the computers tend to be programmed by software developers who are far enough removed from the real world to consider safety in the same way as a real human sitting in a cab, controlling a vehicle travelling at speed. Although maybe not as glamorous as air travel, there is n argument to be made for wanting similar levels of redundancy, multiple instances of independently-developed software, etc. as is the norm in aviation.
It will literally tear itself to pieces if that software isn't working.
This doesn't happen very often.
Not to mention that all high-speed rail has in-cab signalling and automated signalling systems.
All that is software-driven. Software interlocks so the signallers can't send two trains at each other etc.
Don't tar safety-critical software with the same brush as web development. They aren't the same kind of people.
Do, however make sure you toss any manager who thinks they might be similar out of the nearest 10th floor window. Even if it means carrying them up some stairs.
It’s interesting that they are reacting to the perceived threat of autonomous cars.
No idea on the cost of rail transport in Japan but it must be cheaper than in the UK.
In the UK ever since congestion charging was introduced the cost of a train from the burbs into London has kept pace with the ever increasing costs of driving into London.
As a family it’s far cheaper for me to drive to and park in Westminster (handy underground car park by Parliament) and then use the cheap inner London transport than it is to do the whole lot by train.
It could be cheaper by train if I could just pay the train from my town to the outer London zone then cover the rest with a travel card but they don’t let you do that unless you leave the station at the outer London zone and renter using the travel card.
At least Japan have the right incentives here for ensuring train is competitive.
If you are a tourist you will qualify for the Japan Rail pass, which is an absolute bargain, imagine a two week whole of UK travelcard for circa £240. Living there makes it a lot more expensive, in 2019 my round trip Tokyo, Kyoto, Osaka, Tokyo cost me more than the JR Pass.
Driving in Japan is pretty straightforward, my only gripe being the traffic lights hoist in the sky rather than being on the street. Overnight on-street parking was not permitted in Aoba-Ku, where I lived, which made car ownership really expensive, I was advised that it would cost circa £200 a month to rent a space. On the bright side it reduced the amount of traffic when I had to hire a car and streets are just much nicer when not lined with cars.
Given that driving at very high speeds are all done via signalling; those elements are pretty easy to implement. Like electronics, the behaviour of a capacitor or a resistor is well known "within" the sandbox of standard parts. But if you put a Gherkin into your circuit you break the walls of the sandbox.
A controlled environment like the Paris Metro where access to the remotely driven subway trains by anything OTHER than the train is nigh-impossible is one thing. Open country a hell of a lot more difficult to account for gherkins getting into your system.
I wonder what the investment driver actually is for this, because unless you are addressing questions of safety, reliability, financial gain etc. it's just doing it for the sake of it being a "cool" thing.
I'm not an enormous fan of driver-assistance models like some of the newer underground lines - where the driver's job is basically to tell the computer "OK to drive"; if only because I can think of few more difficult things to do than hold your attention on a relatively trivial task for hours at a time. Something where you're more involved is easier to pay attention to.
Based on the comments above re metro-like systems, it means the ability to run trains closer together, thus increasing the line capacity. But as you and others have pointed out, if it's not a closed system, it becomes very hard to do. Probably still much easier by at least one order of magnitude than autonomous cars on the public road though.
Running trains closer together is one of those problems that get progressively harder then are suddenly trivial once you reach maximum capacity. With enough trains it becomes one train.
It's like how traffic accidents increase in congestion until everything is perfectly gridlocked - when they cease
This post has been deleted by its author
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
