Advantages and Disadvantages of driverless trains?

[6Gman]

I think mainline train drivers can relax for a few more years. Autonomous road vehicle (ARV) technology will transform the taxi and road haulage business first. Our largest cities in the next 10-20 years will become UBER++ zones. Privately owned and driven vehicles will be banned. ARV technology will go together with electric vehicles (EVs). The recently launched TESLA truck in the US has already had orders from AB In-Bev (Budweiser) and PepsiCo. They will still have drivers to begin with but you can see how this technology will expand so trucks become autonomous. When it comes to the UK I can see supermarket fleets and the likes of Eddie Stobart switching by 2025-30.

Can you imagine any political party running with that one?

Because I can't!

 

[Bessie]

Can you imagine any political party running with that one?

Because I can't!

That's a fair point. I should've said petrol/diesel vehicles will be banned from inner cities. However EV and ARV technology will develop together so things will look very different in 10-20 years time.

 

[Dave1987]

I never said that driverless trains themselves are pure science fiction, i was replying to the assertion by the person i was replying to who said that a computer should be able to recognize a scene and interpret semaphores and signals as well as a human being, is science fiction.

I actually think that self driving cars are actually harder than self driving trains on our current roads. Try driving in the streets of London. I think in certain situations such as on a motorway where ALL cars are self driving, we may be more likely to see self driving cars.

It is not the driving part that is hard. Actually a smart kid with a raspberry pi, a few motors, can create an autonomous robot that can follow "lanes" painted on a floor pretty well. Under controlled conditions it works very well. And as you said a train appears simpler, after all its on rails. That is true too.

Coding for that is actually relatively simple, after all its pure physics and mathematics. You can using formulas and equations code it. In the sixties and seventies, using computer power less than in some calculators today we automatically put satellites in orbit, and flew to the moon and mars.

But what about the out of ordinary things? This is where unless a computer is coded to do it, it cannot. And for it to be coded, someone would have to be aware of the potential problem in the first place.

Self driving trains are ALREADY here in complete controlled and isolated systems such as the DLR, and Paris Metro. These systems are specially designed infrastructure (radio sensors, and pads instead of lights for example) that is more suited for computerization. In fact it could be argued that such systems are harder for a human to drive. Also out of the ordinary issues are controlled and understood in this limited system. But to code for the National Rail network with its greater variations, equipment designed for humans to understand better than computers (semaphores, colour lights, etc) is just that much harder.

Whilst I do not agree with your ideology, I admire your pragmatism. I do not believe in “driverless” tech, in fact I find some aspects truly terrifying. But I see you are pragmatic so I certainly respect your views and you are totally correct. The real world is wild and mysterious place that is totally different to a computer screen in an office with huge numbers of variables.

I really don’t care long term. Automation is going to a very very big problem for the Government. If driverless trains ever do appear (and I am very sceptical about that) I will happily sit at home on benefits. If it ever happens jobs across the economy will be in short supply by then.

 

[3141]

Regarding a complex scene, trust me, i have seen computers screw up recognise a scene with a bit of snow (it failed to recognize the difference between sky and ground) Using human interaction devices (lights/semapores, etc) is difficult with computers.

Humans have done that too. There have been crashes in the Arctic when pilots flying over snow lost their awareness of the horizon. I think it was referred to as "white-out".

 

[Billy A]

I'm a software developer myself, although probably only a fair-to-middling one who has never worked on any kind of image processing.

However I do know that there are driverless cars in existence now. The problems with programming those must be an order of magnitude greater than a similar train system (At the most obvious level, you don't need to steer a train).

So driverless trains are impossible science fiction but driverless cars science fact? I don't quite understand that.

We're getting off topic here as the original posting was about pros and cons of driverless operation, not the philosophy of same but....truly driverless cars are much further off than was thought to be the case even a year ago and some proponents - Volvo for one - have rowed back on the idea. There's a huge difference between steering and maintaining speed in a somewhat controlled environment like a motorway with clearly marked lanes, no opposing traffic and no pedestrians and expecting a car to go from any point A to any point B over any roads encountered and in all weathers and while working with non-autonomous road users. A bit like comparing an automated underground train (a thing that was possible twenty years ago but never mind) with mainline rail.

 

[bastien]

I seem to remember self-driving cars were going to be everywhere by 2018. Now it's going to be 2021... It's a bit like a low-rent version of nuclear fusion: always on the horizon, but just near enough to put off investing today's money in the technology available now. See also, one E. Musk and his hypertubes...

 

[bahnause]

If you run driverless trains, make shure to have appropriate contingency plans ready. Didn‘t go very well yesterday in Paris:

https://www.novinite.com/articles/1...eds+Stranded+in+Sweltering+Paris+Metro+Trains

 

[Bletchleyite]

If you run driverless trains, make shure to have appropriate contingency plans ready. Didn‘t go very well yesterday in Paris:

https://www.novinite.com/articles/1...eds+Stranded+in+Sweltering+Paris+Metro+Trains

That's why I'm more of a fan of the DLR approach, i.e. guard-only operation, than completely staffless trains. In emergency the "guard" can drive the train on sight at walking pace to a suitable place to evacuate.

 

[delt1c]

there seems to be this crazy rush to do away with staff and make things cheaper to operate, but lets look at the other side of the equation, no jobs means no money , means no customers.

 

[Dave1987]

there seems to be this crazy rush to do away with staff and make things cheaper to operate, but lets look at the other side of the equation, no jobs means no money , means no customers.

It’s an extremely interesting debate. Opinions seem to be that as tech gets increasingly more sophisticated and takes over more jobs that it creates new jobs in similar numbers to what it replaces. This argument seems somewhat of a wing and a prey one. A kind of ‘we hope it creates as many jobs as it’s replacing’. If they are wrong then the economy will face a huge challenge and could cause a huge depression. Train driving is one of the hardest jobs to automate because of the number of factors involved so the crisis will probably have already hit before we are replaced by automation. I really fear for others though unless Government stops burying its head in the sand over the issue.

 

[mushroomchow]

I seem to remember self-driving cars were going to be everywhere by 2018. Now it's going to be 2021... It's a bit like a low-rent version of nuclear fusion: always on the horizon, but just near enough to put off investing today's money in the technology available now. See also, one E. Musk and his hypertubes...

They'll still be around before the 769s at this rate.

Joking aside, it's not like driverless tech isn't already applicable in some scenarios for railways, particularly urban settings. The DLR already exists and operates near flawlessly with relatively little input, and there's real talk of at least one tube line going the same way within a generation. Digital and on-board signalling already exists, removing the complexity of having to recognise such prompts within the environment.

The obvious upside of a closed, urban system going driverless is that it reduces the risk of closure in the case of strikes. I was in London the last time there was a DLR strike and it was still practically running a full timetable. That changed my mind on the issue - I had previously been apprehensive about the concept but it clearly works if implemented right.

As for the Hypertube concept, it's only very recently that somebody (read - Elon Musk) has actually started R&D to turn it from science fiction into reality. It's very early days as a technology, so I wouldn't write it off just yet given that they've already managed to get a short Railplane-esque test track built and operational for something that has only actively been pursued since 2012. That's a very promising turnaround, provided that people take up the mantle and continue to develop what Musk himself declared an "Open Source" technology.

For comparison, the first "proper" driverless car was tested in Japan in 1977, so you're talking about an R&D lifetime of 41 years to get to the point we're at now.

 

[James James]

It’s an extremely interesting debate. Opinions seem to be that as tech gets increasingly more sophisticated and takes over more jobs that it creates new jobs in similar numbers to what it replaces. This argument seems somewhat of a wing and a prey one. A kind of ‘we hope it creates as many jobs as it’s replacing’. If they are wrong then the economy will face a huge challenge and could cause a huge depression. Train driving is one of the hardest jobs to automate because of the number of factors involved so the crisis will probably have already hit before we are replaced by automation. I really fear for others though unless Government stops burying its head in the sand over the issue.

This "all jobs will be lost" argument has existed for at least a thousand years. Humanity has to adapt, and will adapt. A country that chooses to legislate away short-term job issues will long-term end up in economic hell. Then again the UK has already chosen that for itself so no difference if they do that again.

 

[Dave1987]

This "all jobs will be lost" argument has existed for at least a thousand years. Humanity has to adapt, and will adapt. A country that chooses to legislate away short-term job issues will long-term end up in economic hell. Then again the UK has already chosen that for itself so no difference if they do that again.

As long as everyone benefits from the tech revolution and not just those at the top..... For people to embrace tech they have to feel better off from it not worse off.

 

[Bletchleyite]

From https://www.railforums.co.uk/threads/arriva-rail-north-doo.137291/page-255#post-3749406

A train already drivers itself in the Thameslink core right? The driver accepts automatic train operation and the train drives drives between stations afaik?

Correct, and on several LU lines too, as well as the DLR.

 

[ComUtoR]

Correct

Which services run with ATO through the core ?

 

[Bletchleyite]

Which services run with ATO through the core ?

Apparently not rolled out yet (correction posted on the other thread).

 

[edwin_m]

From what is says in December's Modern Railways I believe the drivers can use ATO if they've been trained on it, but it isn't necessary until the service has stepped up to 20TPH and with all the timetable issues they've been prioritising training of route knowledge over ATO.

 

[Meerkat]

Surely the trains will have to go ATO soon?
Not only are computers getting cleverer and safer but the drivers are surely getting more dangerous as more technology comes in - ie it starts getting too easy and therefore too hard to concentrate.
Interested to hear drivers’ opinions - by the time you get C-DAS telling you how fast to go will the workload be high enough for drivers to be able to maintain concentration to see the mythical cow on the line (or get much job satisfaction)? Would this be more of an issue on short metro type lines like Moorgate, or long distance with fewer stops?

 

[jagardner1984]

Surely a middle ground is the ideal.

Let the computer do the bulk of the driving. Statistically in cars it will reduce lots of accidents.

Insist on a person to be at the controls.

In the railway world, entirely driverless operation is surely rather terrifying. What happens in the event of a suicide ? What happens in the event of poor wheel adhesion and overshooting a platform ? What happens in the event of a medical emergency on board ?

Ultimately, if a driverless car wraps round a lamp post, thats very sad, the appropriate authorities are called and the cleanup and diversions for a few dozen cars are organised. If someone throws themselves under a driverless train at Finsbury Park, the potential delays to thousands of people almost immediately whilst appropriate people take control of the train, organise resources and establish exactly what's happened, are very extreme.

Really, when metal boxes with 1000 people in them are being thrust around the country at 125mph, I want someone to be in the front.

 

[InOban]

Someone in the front who can do nothing to prevent the incident, and is likely to develop PTSD symptoms as a result?

Of course we still need on board staff to deal with a variety of passenger-related incidents. Probably more of them.

 

[Bletchleyite]

Someone in the front who can do nothing to prevent the incident, and is likely to develop PTSD symptoms as a result?

Of course we still need on board staff to deal with a variety of passenger-related incidents. Probably more of them.

I agree - hence my suggestion that a properly trained guard is that person. Possibly two in some places.

 

[Bletchleyite]

In the railway world, entirely driverless operation is surely rather terrifying. What happens in the event of a suicide ? What happens in the event of poor wheel adhesion and overshooting a platform ?

What can a human do in those cases that a computer can't with far more precision and accuracy and no emotions?

What happens in the event of a medical emergency on board ?

The guard deals with it, who is in a far better position to do so than a driver would be.

Really, when metal boxes with 1000 people in them are being thrust around the country at 125mph, I want someone to be in the front.

In my view this is irrational. I want a computer driving it, because fundamentally it will not make a mistake. A computer already does pretty much drive most proper high speed trains anyway - ATP, ETCS and the likes seriously restrict what the human driver can do.

Humans are good at dealing with humans but make terrible pseudo-automatons. Automate the work of the automatons, and get humans dealing with human problems.

 

[Eccles1983]

Computers don't make mistakes?

And yet somehow other people are being irrational?

I'll remember that next time a software malfunction ruins my computer, or causes websites to crash, or just decides to give up because the problem is outside of its operating parameters.

 

[Dave1987]

What can a human do in those cases that a computer can't with far more precision and accuracy and no emotions?



The guard deals with it, who is in a far better position to do so than a driver would be.



In my view this is irrational. I want a computer driving it, because fundamentally it will not make a mistake. A computer already does pretty much drive most proper high speed trains anyway - ATP, ETCS and the likes seriously restrict what the human driver can do.

Humans are good at dealing with humans but make terrible pseudo-automatons. Automate the work of the automatons, and get humans dealing with human problems.

And like I say continually to you let me know when it happens and I will quite happily toddle up the dole queue .

 

[InTheEastMids]

From what is says in December's Modern Railways I believe the drivers can use ATO if they've been trained on it,

I think we're losing the distinction between automatic and autonomous operation. If a driver needs training to use it, they're still in supervisory control so by definition this isn't genuine autonomous operation, it's more like the rail version of Adaptive Cruise Control with Stop/Go.

I want a computer driving it, because fundamentally it will not make a mistake

Until, of course, said computer fails to accurately perceive its situation, or is confronted by a situation that it's not programmed to deal with.

 

[Strat-tastic]

In my view this is irrational. I want a computer driving it, because fundamentally it will not make a mistake.

You've never had a computer crash on you? Built by humans, they will go wrong on occasion.

 

[Indigo Soup]

You've never had a computer crash on you? Built by humans, they will go wrong on occasion.

Which will result in the train sitting in a platform going nowhere (or worst case sitting on a running line going nowhere) exactly as if any other technical failure had occurred. Any autonomous system will be designed to fail safely, possibly with a recovery mode that allows (say) running at reduced speed to a safe place to terminate if only two of three independent processors are working.

 

[InTheEastMids]

Which will result in the train sitting in a platform going nowhere (or worst case sitting on a running line going nowhere) exactly as if any other technical failure had occurred. Any autonomous system will be designed to fail safely, possibly with a recovery mode that allows (say) running at reduced speed to a safe place to terminate if only two of three independent processors are working.

Potentially much easier said than done. You're not that far from describing AF449. There was a sensor issue, the 3 automated computers disagreed, and tried to fail safe by handing back control to a flight crew in a situation they weren't properly trained to deal with, who then lost control of the plane with catastrophic consequences. The wrong response at this point is to say "well, trains don't fly over water". The point that needs to be taken away is that complex systems fail in complex, unpredictable ways through chains of events. So designing a system to fail safely without failing all the time because of false positives could prove very hard to do.

Google/waymo found their sensor system struggled to differentiate between a child and s plastic bag. Is it OK for the Thameslink core to stop because of a plastic bag on the line at Blackfriars?

 

[Kneedown]

The bottom line is, a computer can only make decisions on the data it has, ie what is happening. It cannot assess the build up to a situation wheras the Mk1 human eyeball can.
Drivers jobs are safe for many, many.... many years yet.

 

[Bletchleyite]

You've never had a computer crash on you? Built by humans, they will go wrong on occasion.

You aren't exactly going to run something safety-critical on Windows. Do some more research into how things like the Airbus computer systems work.