Southern Rail Franchise

[phil-b259]

Can we be a little bit careful in what we post on this thread.

 

I say this because the previous thread relating to the DOO / industrial relations issues got locked. We debated to pros and cons of DOO on said thread and don't need a rehash of the arguments all over again.

 

What is helpful is news regarding the dispute - e.g. updates as regards the ASLEF talks or the calling of fresh industrial action.

 

Equally news of any fresh moves the Government may make would be welcomed as is information on the feelings of users - who have rather been ignored in all this and who have suffered considerably.

 

We know the RMTs position - we know Southern / DfTs position and we think we know ASLEFs position - none of which have changed in 9 months. The same is not true as regards the travelling public.

Edited January 28, 2017 by phil-b259

[njee20]

Trains will be driven by humans, computers will never be able to react to the infinite number of scenario's possible: they simply have no space in their system to store the programmed response for every scenario. The most common, sure, but it's not the 'most common' situations that causes an accident that a human might be able to prevent and a computer can't. No computer is fool-proof, it just hasn't been used by met the better fool yet

Like I said, we're not there yet. Cars have far more scenarios they can encounter, and we're not a million miles away from that. Automating cars makes doing trains easier too. Level crossing behaviour becomes dependable. No sleepy drivers crashing onto the track etc. Different performance characteristics are a tiny variable and most certainly not a barrier.

[Mike Storey]

Like I said, we're not there yet. Cars have far more scenarios they can encounter, and we're not a million miles away from that. Automating cars makes doing trains easier too. Level crossing behaviour becomes dependable. No sleepy drivers crashing onto the track etc. Different performance characteristics are a tiny variable and most certainly not a barrier.

 

Your faith in technology continues to inspire us all, I am sure. However, this is about DOO operation of doors at extremely crowded platforms. Even on the DLR, supposedly fully automated, humans takeover the door controls at the most crowed stations in the peaks and at other times of crush loading. A computer would simply understand that the conditions in such situations are insufficient for it to allow door closure, or at best would try to close the doors and meet constant obstacles, so the train would never move. I am sure you could tell of us of very clever algorithms that could soon make that judgement, but from what I have seen of robotic car development, these more qualitative judgements are the most problematic and as yet unproven for the on-board computers.

 

A human on the other hand, is meant to be able to make a judgement that may not correspond to absolute safety, and intervene when humans are not behaving in a logical manner. Having experienced DOO operation for many years on the GN, where the majority of platforms are straight and relatively wide, there is still significant human intervention either by platform staff or by the driver, during the peaks.

 

Nonetheless, it is very hard to distinguish between the apparent success of DOO on so many crush-loaded routes for many years now, and the case that RMT/ASLEF are making for the rest of the Southern routes (especially given that Thameslink, Metro and GatEx DOO operates well into that area already). This article, from 2015, is worth a read, although it too reaches no firm conclusion, other than that, at a time of huge passenger growth, including many who need physical assistance, it is the wrong period to be de-humanising the railway.   http://www.railmagazine.com/trains/current-trains/the-pros-and-cons-of-driver-only-operation

[great central]

As a serving member of train crew I get quite annoyed and frustrated by some of the comments from outsiders but due to company policy I'm not allowed to comment apart from in very general terms.

Passengers are humans with all their faults and idiosyncrasies as indeed are rail staff. How many commenting here dislike dealing with the automated tills in the supermarket, I know I usually finish up arguing with the damned thing!

I had a situation last night that almost certainly would have gone unknown on a DOO service and definitely on a fully automated train. Nothing over serious in itself but without responsible human intervention had the possibility of a tragic outcome. I could recall many others I have been personally involved with as I'm sure can all other rail staff.

How many times have you heard a human voice interrupt the automatic announcement because the computer can't respond fast enough to changes at stations.

In my view the only way to make a fully automatic system safe is to remove the human element altogether. That means no human passengers are allowed a time in the future when everyone is forbidden to leave their, by then, fully automatic prison cell.

[Talltim]

I discussed driverless trains many years ago with a BR signal engineer. His question "what does the driver do if he sees something on the line?" "Blows the horn". "What if the object doesn't move"? "Applies the brakes". "Then what, hits it because he can't stop in time". A train with no driver would have the same effect.

The difference is, after the train has hit whatever, a driver can get down and attempt to sort it out. Be a long time before a self driving train can disentangle a shopping trolley from its lifeguards

 

Auto operation is ok where every train is the same length and to a degree same weight, like the Victoria Line, DLR etc. But NOT on a railway where you can have a light loco weighing say 100 tonnes following a 1200 tonne freight, being followed by a 300 tonne commuter train. All have different braking and acceleration characteristics, all will handle differently. Pure auto operation is not in my opinion an option.

I don't see why train weight and other characteristics make any difference, so long as the train knows those characteristics and is programmed correctly to deal with the variables. It's only the same as training human drivers for different types of trains and conditions

[Coryton]

As a serving member of train crew I get quite annoyed and frustrated by some of the comments from outsiders but due to company policy I'm not allowed to comment apart from in very general terms.

Passengers are humans with all their faults and idiosyncrasies as indeed are rail staff. How many commenting here dislike dealing with the automated tills in the supermarket, I know I usually finish up arguing with the damned thing!

I had a situation last night that almost certainly would have gone unknown on a DOO service and definitely on a fully automated train. Nothing over serious in itself but without responsible human intervention had the possibility of a tragic outcome. I could recall many others I have been personally involved with as I'm sure can all other rail staff.

How many times have you heard a human voice interrupt the automatic announcement because the computer can't respond fast enough to changes at stations.

In my view the only way to make a fully automatic system safe is to remove the human element altogether. That means no human passengers are allowed a time in the future when everyone is forbidden to leave their, by then, fully automatic prison cell.

 

I find this a difficult one.

 

I would like to travel on trains with a guard/conductor/whatever. And I'd rather that than pay a bit less for my ticket or have my taxes reduced slightly.

 

But I have to admit that large chunks of the railway have operated with only a driver for a considerable period of time and it seems to have worked, just as we manage on busses without conductors now.

[Zomboid]

The thing with automation is we couldn't do it on the main line right now.

But never is a big concept, and one day I expect the technology will be there. Ultimately something as regimented as the railway is just a set of rules to apply to various scenarios. So technologically speaking, the railway is ripe for automating. It's just a huge job to document all the stimulus that a driver receives and then the appropriate reactions with all the other variables. But it is just a long "if x, y & z, then do a, b & c" statement, which is the kind of thing computers eat for breakfast.

Is automating just about everything a good idea? I'm not sure, what would be the point of humans if everything was automatic? But that's a debate for somewhere else.

[Coryton]

Trains will be driven by humans, computers will never be able to react to the infinite number of scenario's possible: they simply have no space in their system to store the programmed response for every scenario. The most common, sure, but it's not the 'most common' situations that causes an accident that a human might be able to prevent and a computer can't. No computer is fool-proof, it just hasn't been used by met the better fool yet

 

But it doesn't work like that. You don't program a self-driving car with every possible situation it could encounter and what to do about it any more than you train a driver by showing them every possible combination of circumstances that could arise.

 

Like a human the computer needs to have decision making capacity and while self-driving cars aren't there yet they have racked up a huge number of miles with fairly few driver interventions.

 

Now on a train you have more variables in loading/number of vehicles, and maybe the handling of a train is more dependant on the state of the rails than a car is on the state of the road. But the train moves on a well defined path with other traffic kept out of its way. I'm not saying it's trivial, but working out when and how to brake after a yellow signal or to stop at a station must be easier than deciding when it's safe to pull out of a side-road into traffic. And computers can do that now.

 

Your faith in technology continues to inspire us all, I am sure. However, this is about DOO operation of doors at extremely crowded platforms. Even on the DLR, supposedly fully automated, humans takeover the door controls at the most crowed stations in the peaks and at other times of crush loading. A computer would simply understand that the conditions in such situations are insufficient for it to allow door closure, or at best would try to close the doors and meet constant obstacles, so the train would never move. I am sure you could tell of us of very clever algorithms that could soon make that judgement, but from what I have seen of robotic car development, these more qualitative judgements are the most problematic and as yet unproven for the on-board computers.

 

I thought on the DLR the passenger service agent or whatever they are called now always operated the doors? I've also seen them take over the driving controls and I seem to recall they are (or were) required to do so on entering stations with crowded platforms because the train can't see if someone is too close too the edge. (Which is not to say it couldn't be done. But DLR trains can't).

 

The DLR does not use the latest technology. Again if a car can drive itself I'm pretty sure a computer could look at images of doors and make sure nobody is stuck in them or leaning against the train. But maybe you're right that it does need a member of staff to chivvy people along and make sure the doors actually can be closed so the train can go.

 

But this is what I was saying at the start. It may not be a pleasant thought but I am pretty sure that the time will come when all trains could be made driverless. In that case people are probably going to still want a member of staff aboard (as on the DLR) and it may be necessary for reasons such as the ones above.

 

So while at the moment guards' jobs are under threat but not drivers, I suspect the time will come when the tables are turned.

Edited January 28, 2017 by Coryton

[royaloak]

I find this a difficult one.

 

I would like to travel on trains with a guard/conductor/whatever. And I'd rather that than pay a bit less for my ticket or have my taxes reduced slightly.

 

But I have to admit that large chunks of the railway have operated with only a driver for a considerable period of time and it seems to have worked, just as we manage on busses without conductors now.

Your fares wont go down and neither would the subsidy, what would happen is that the franchisee would see a bit more profit, not as much as the supposed 'savings' from DOO because of the increased risk of an incident leading to a Court case but by the time you factor in the costs of the Court case and compensation payout for somebody losing their legs etc there is still a saving in real terms.

 

Can you show me where we have run 12 coach DOO trains with unstaffed rural stations for years please?

 

Can anyone explain to me why the McNulty report had to be deferred and rewritten when there was a change of Government?

Surely the report would have come to the same conclusion regardless of what colour Government we have!

Edited January 28, 2017 by royaloak

[great central]

I find this a difficult one.

 

I would like to travel on trains with a guard/conductor/whatever. And I'd rather that than pay a bit less for my ticket or have my taxes reduced slightly.

 

But I have to admit that large chunks of the railway have operated with only a driver for a considerable period of time and it seems to have worked, just as we manage on busses without conductors now.

 

Granted DOO does seem to have worked up to now, but we now live, rightly or wrongly, where everything is always someone else's fault. There are cases ongoing where responsible people find themselves in court or in prison, not talking just railways here, for either shirking those responsibilities or not carrying them out properly.

I have no wish to see anyone prosecuted over another persons blatant stupidity which I or any other responsible person has the misfortune to see happen.

As I understand it this is why the drivers are now taking a stand, where you had 4 car trains with maybe 300-400 passengers we're now talikng up to 12 cars with over 1,000 passengers and rail staff are responsible for every one of them and their actions.

A bus is somewhat different, not to belittle the responsibility in any way, it's a single unit vehicle with the driver pretty able to be aware of whatever was happening, I can certainly remember bus drivers stopping the bus on the way home from school and coming upstairs to tell off some unruly kids, not this one I might add even though I was no angel I always had a certain repect for people in public facing roles, also my brother in law was a bus driver, eventually had to quit the cab due to stress.

[Coryton]

Your fares wont go down and neither would the subsidy, what would happen is that the franchisee would see a bit more profit, not as much as the supposed 'savings' from DOO because of the increased risk of an incident leading to a Court case but by the time you factor in the costs of the Court case and compensation payout for somebody losing their legs etc there is still a saving in real terms.

 

 

If - as has been reported many times - the DfT is behind all this, inspired by the McNulty report, I would be a little surprised if their aim was to let the TOCs make more profit rather than reducing the overall subsidy to the railways.

 

 

Can you show me where we have run 12 coach DOO trains with unstaffed rural stations for years please?

 

No. I have no idea if this is the case or not.

 

But in terms of someone losing their legs as you refer to above, I'm not sure what difference it makes if it's an unstaffed rural station (which I regularly used to use 6 coach DOO trains at) or a staffed urban station with no regular platform staff.

 

Likewise I don't know if a 12 coach train is much more dangerous than a 6 coach train for a driver to dispatch.

 

Or if a bank of monitors in the cab is more dangerous than platform screens/monitors.

 

In any case I am not trying to argue the case for more DOO. I would rather not see it happen. But I'm not sure that there are good grounds to object to it on safety grounds.

 

However, putting that aside, as I've said above from a customer service point of view I think a guard closing the doors is overall better for passengers than one who has effectively become another passenger who goes down the train checking tickets and giving travel advice.

[great central]

But it doesn't work like that. You don't program a self-driving car with every possible situation it could encounter and what to do about it any more than you train a driver by showing them every possible combination of circumstances that could arise.

 

Like a human the computer needs to have decision making capacity and while self-driving cars aren't there yet they have racked up a huge number of miles with fairly few driver interventions.

 

Indeed you can't train a driver with every possible situation, neither can anyone in any kind of work be trained to deal with everything they might encounter. What you do is give them the ability to respond in the MOST APPROPRIATE way to ANY given situation. You say yourself that self driving cars have had human intervention, maybe because the computer didn't recognise the problem or maybe a minor error in the computer?

When I owned a video shop we used the 'industry standard' software in our computerised till system. Every so often it threw up a glitch which instead of charging for an overdue it actually took money off the initial rental fee. Everything else on the screen was correct date out, date in, daily charge, etc. It was just that the total owed was wrong. Speaking at some length to the software company at a trade show, they were well aware of the problem but all they could offer by way of explanation was ' we know it does that but it can't do it!' In other words there was something somewhere that made it happen but they couldn't identify the issue.

Also a friend has a DCC controlled layout, every so often one or other of the chips will go 'wild' for no apparent reason and the train becomes unstoppable apart from the hand of god, a pair of O gauge Heljan class 20s takes quite a bit of holding back I can tell you

[Coryton]

Indeed you can't train a driver with every possible situation, neither can anyone in any kind of work be trained to deal with everything they might encounter. What you do is give them the ability to respond in the MOST APPROPRIATE way to ANY given situation. You say yourself that self driving cars have had human intervention, maybe because the computer didn't recognise the problem or maybe a minor error in the computer?

 

Self driving cars are still under development. Maybe there is some fundamental reason that they will never be able to cope with the last fraction of a percent of situations they can't currently handle but I suspect not. In any case it is far too early to say that they will never become good enough.

 

Of course human drivers also make a lot of mistakes and in principle a computer only needs to become as good as a human. But I suspect for public acceptance they will have to do quite a bit better than that.

 

If you're concerned that computer programs have bugs and sometimes behave randomly then you need to be concerned not just about self-driven cars but also about human-driven modern cars where a computer sits between the controls and what the car actually does. 

[great central]

Self driving cars are still under development. Maybe there is some fundamental reason that they will never be able to cope with the last fraction of a percent of situations they can't currently handle but I suspect not. In any case it is far too early to say that they will never become good enough.

 

Of course human drivers also make a lot of mistakes and in principle a computer only needs to become as good as a human. But I suspect for public acceptance they will have to do quite a bit better than that.

 

If you're concerned that computer programs have bugs and sometimes behave randomly then you need to be concerned not just about self-driven cars but also about human-driven modern cars where a computer sits between the controls and what the car actually does. 

 

I'm well aware of that, you might hope though that, again, human intervention could prevent or at least reduce the severity of any such incident. The computer will just carry on as if there's nothing amiss.

Sorry but I don't buy into this idea that a machine is the answer to everything.

Anyway I must get ready to do some human intervention in unexpected situations now, I'm getting ready for work. Hopefully an incident free day but, hey, that's what makes the job what it is and I thoroughly enjoy it and the challenges it provides

[The Stationmaster]

Good points raised in your last paragraph Roy, and as you're no doubt aware a 1,200 ton freight train is by no means the heaviest on today's network (ours at FLHH vary anywhere between 300 and 3,300). As for braking characteristics, you can have the same loco and same set of wagons on the same route five days in a row and the braking 'feel' will be different each time.

 

Interestingly some trials were conducted on the WR in the early days of ATP installation to see exactly what would happen if a Driver ignored single yellow and red signal aspects (in a 3 aspect signalled area) and the results were very carefully recorded on instrumentation.

 

Every time an HST was tried, irrespective of the speed it was running at - right up to linespeed, it came to a stand either in rear of the red signal or within the overlap with both dry and wet rails.  Several combinations of freight train load were tested in all cases running at maximum permitted speed for the particular loco/wagon combination and with nothing done to check the wagons before the test was carried out.  Over several tests, again on both dry and wet rail, only one stopped within (just) the overlap and most overran it by a considerable margin, the worst was a loaded train of MGR wagons which simply sailed past the signal at danger .  After the running tests the trains' brake adjustment was checked and all were within tolerances.  A second set of freight train tests returned results which were no better despite various changes to the equipment.  Simple fact was that it was considered extremely difficult to get consistent brake perfomance on freight vehicles (or putting it another way - as Nidge said - 'the braking 'feel' will be different each time).

[The Stationmaster]

Your fares wont go down and neither would the subsidy, what would happen is that the franchisee would see a bit more profit, not as much as the supposed 'savings' from DOO because of the increased risk of an incident leading to a Court case but by the time you factor in the costs of the Court case and compensation payout for somebody losing their legs etc there is still a saving in real terms.

 

Can you show me where we have run 12 coach DOO trains with unstaffed rural stations for years please?

 

Can anyone explain to me why the McNulty report had to be deferred and rewritten when there was a change of Government?

Surely the report would have come to the same conclusion regardless of what colour Government we have!

 

No, but I can show you where 8 coach trains have operated and many instances where 5 and 6 coach trains have operated/operate and have done for more than 20 years where they are 'full to the gunwhales'/'rampacked'/ jampacked or any other description you might want to apply to full and standing 'and then some'.

[Zomboid]

The tube operates DOO, though with platform dispatch staff at particularly busy stations. I think those trains could be considered to be "packed".

[Derekl]

The Stationmaster: "DOO(P) does not involve any new technology at all, certainly from what I have heard on Southern where it is all established technology previously used in both Britain and elsewhere.  DOO(P) has never, to my knowledge, been responsible for any sort of operational safety incident since it was first used in Britain 30 odd years ago."

 

These are links to 4 RAIB reports on passenger injuries when boarding/alighting from DOO trains (I am not sure of the relevance of the "(P)"):

 

https://www.gov.uk/raib-reports/huntingdon-train-door-incident

https://www.gov.uk/raib-reports/serious-accident-at-west-wickham-station

https://www.gov.uk/raib-reports/passenger-accident-at-brentwood-station

https://www.gov.uk/raib-reports/passenger-accident-at-hayes-harlington-station

 

It possibly cannot be said that the presence of a guard would have prevented or made these incidents less likely, but it seems that the driver missed the presence of the passengers for one reason or another and a guard possibly with the assistance of platform staff seems to me more likely to have done so.

 

I am aware that some incidents involve passengers possibly being foolish, but that is what occurs.

 

The Hayes and Harlington incident is the one where it was appreciated that the doors closed signal may not be reliable.

 

It should also be said that there are some reports of similar incidents on guard operated trains, although there seem to be remarkably few compared to the number of guard operated trains in use.

[Coryton]

PS: Continental cars are longer then UK ones, by a considerable margin. As for passenger numbers, many railways here operate double-decker trains, their capacity is generally 70% more over a single decker train of the same length. Having driven them myself, I can assure you that 2xVIRM-6 (about 1100ft) can hold approx 2000 people in dense rush hour traffic with relative ease. In super-dense crash mode 2300 would be possible, but for safety reasons only for short stretches.

 

I hope that was supposed to be "crush" mode!

 

What happens if 2300 people try to use the train for more than a short stretch?

[Coryton]

 

 

regulation and will wait for the next one if the train is full. But that's because our timetable is fairly predictable, as it's quite rigid: every 10, 15 or 30 minutes a train in each direction, often augmented with other trains (regionals, locals, etc) in a similar pattern*.

 

 

Not sure the frequency of trains is the full story - it can be hard to get people on the underground in London to wait for a less full train 2 minutes later.

 

I think to some extent it depends on your belief that the next train will actually turn up as expected...

[The Stationmaster]

The Stationmaster: "DOO(P) does not involve any new technology at all, certainly from what I have heard on Southern where it is all established technology previously used in both Britain and elsewhere.  DOO(P) has never, to my knowledge, been responsible for any sort of operational safety incident since it was first used in Britain 30 odd years ago."

 

These are links to 4 RAIB reports on passenger injuries when boarding/alighting from DOO trains (I am not sure of the relevance of the "(P)"):

 

https://www.gov.uk/raib-reports/huntingdon-train-door-incident

https://www.gov.uk/raib-reports/serious-accident-at-west-wickham-station

https://www.gov.uk/raib-reports/passenger-accident-at-brentwood-station

https://www.gov.uk/raib-reports/passenger-accident-at-hayes-harlington-station

 

It possibly cannot be said that the presence of a guard would have prevented or made these incidents less likely, but it seems that the driver missed the presence of the passengers for one reason or another and a guard possibly with the assistance of platform staff seems to me more likely to have done so.

 

I am aware that some incidents involve passengers possibly being foolish, but that is what occurs.

 

The Hayes and Harlington incident is the one where it was appreciated that the doors closed signal may not be reliable.

 

It should also be said that there are some reports of similar incidents on guard operated trains, although there seem to be remarkably few compared to the number of guard operated trains in use.

 

I did say 'operational safety incidents' - i.e. incidents affecting the safety of the line/train, and I know of none; operationally DOO(P) has a good safety record.

 

As far as incidents involving passenger safety are concerned the only valid comparison is between those where the decision to close the doors and depart is solely in the hands of the driver (and the on-train equipment has no design faults) and those where such decisions and actions are in the hands other members of staff be they a Guard or platform staff.  And - again - I'm not aware of any comparative statistics being published but they might lurk somewhere within an ORR Inspectorate Report because such incidents have very definitely occurred.

 

Equally - again as far as I'm aware - no summary is available of injuries or fatalities sustained by Guards when despatching their own train but I do know that such things have occurred although they seem now to be less common than they once were (usually such incidents occur when the Guard has been on the platform checking the train or watching for the tip from platform staff and has tried to join the train as/after it started to move).

[Coryton]

such incidents occur when the Guard has been on the platform checking the train or watching for the tip from platform staff and has tried to join the train as/after it started to move).

 

There can't be many trains around nowadays that can move before the guard's door is closed.

[Glorious NSE]

I know that currently driverless trains only work in nice contained systems with considerable support from signalling equipment.

Apropos not much - by 'not having staff on the train or on the station' I wasn't including the driver in that as they are for the most part not customer facing and busy concentrating on driving the train, so apologies for kicking off a divergence into entirely unstaffed trains! 

 

Can you show me where we have run 12 coach DOO trains with unstaffed rural stations for years please?

 

East Anglia's network currently sees 12 x 321 (and 12 x 350?) formations operated DOO - i'd be very surprised if every station those call at are staffed from first to last train?

Edited January 28, 2017 by Glorious NSE

[jjb1970]

On the automation question I think there are two separate and distinct questions, could you automate heavy rail, and would it make sense. Those two questions may have very different answers.

If considering the technical question of whether it could be done then I think that also needs to be split into two separate questions, a new build railway and retrofittment to a legacy railway. If you were building a new railway (for example, HS2) then I think it would be a long way from being the most challenging process automation program and it would be entirely possible. For a legacy railway then I think ultimately it could be done but the integration with existing infrastructure and assets, a lot of which is very old, would make it orders of magnitude more demanding. And that is before considering how you'd manage the work and transition whilst continuing normal operation of the lines.

Does it make sense? That depends on cost-benefit. You automate processes for a reason, that could be greater quality in manufacture, heightened safety by reducing the probability of human action leading to an incident, greater efficieny and asset utilisation (in a railway sense probably translates to tighter train spacing) and lets be honest - reducing head count to save money. To make sense, the benefits from those improvements need to make it worth the investment in automation. For many processes its a no brainer, the more pertinent question would be to ask why you wouldn't automate the process. However for a large legacy railway network I really question whether the advantages would make it worth the effort. For certain intensive corridors then yes, but in a general sense no. For a new build railway then I think it probably would be worth going automated.

One of the traditional problems with automating public facing, safety critical activities has been public acceptance (I'm sure many remember the scare stories about the Airbus "scarebus" fly-by-wire system), but we now live in a different world where people are much more comfortable with allowing automated systems manage things. As self driving cars start to be a normal part of life (and I believe they will, sooner rather than later) then people are unlikely to be particularly worried about automated trains.

And for all of that an automated train is not the same as an unmanned train. Most automated processes retain humans in the chain somewhere. On a train you'd still want some sort of customer facing staff to help passengers and to act in an emergency.

[Coryton]

For a legacy railway then I think ultimately it could be done but the integration with existing infrastructure and assets, a lot of which is very old, would make it orders of magnitude more demanding. And that is before considering how you'd manage the work and transition whilst continuing normal operation of the lines.

 

I'm not so sure. The train would interact with the signalling system in the same way (track circuits/axle counters). And if a car can stop itself at traffic lights then in principle an automated train could stop at lineside signals.

 

I suspect the chance of failure to stop a signal would be considered too great so you'd want direct communication with the train, but we're moving that way anyway because humans miss signals too.

 

By the time we've gone over to ERTMS, computing technology may well be at a point at which automating the driving is fairly straighforward. 

 

And I don't see a problem with mixing automatic and manual trains over a transition period - much easier than with cars because trains don't directly interact with each other.

 

As for manning, I imagine in general there will be a desire to retain at least one staff member on a train - but I think the time will come when that person isn't in the cab.

 

If the move to DOO isn't turning out to be smooth, imagine what the move to COO (conductor only operation) will be like!

 

And even then, on a frequent urban service with manned stations (London Overground) could the time come when the trains have nobody staff member on board?

 

The Lille Metro is automated and the trains are completely unmanned and it's been running for quite a while...