Driver-less trains time scale... (Metro)

[Broot]

Looking for some advice.

I have recently passed my psychometrics and just waiting for confirmation of either being given a start date or being put in the holding pool to train as a driver for tyne and wear metro service.

My question is- What is the general opinion on how far away this type service is from being driver-less? or Light rail in the UK in general?

Im in a very stable career with prospects at the moment, its not quite driver money but its getting there. Becoming a driver is a dream for me but also a risk. i dont want to find in 10 years time computer automation being retro fitted into all stock and im out of a job! - iv got 35 years before i can retire!

Any advice would be appreciated or links to some relevant articles.
Thanks

 

[gray1404]

I have wondered the same about Merseyrail. A self contained network so how long before it is driverless. We already have it on DLR.

 

[Bromley boy]

I wondered similar before taking the plunge to train as a driver recently (I'm in the same age range). However having read around the subject quite extensively full driverless is a long way off and someone in their 30s has a decent shot at seeing a full career out of it (albeit there will be quite a few changes over the next few decades).

- even the tube, a fully self contained system, is still some way from being fully driverless (at least a decade or so until the NTFL is introduced, although even this will be crewed with drivers initially).
Self contained networks such as Merseyrail are somewhere between the tube and the big railway.

- in terms of the big railway we are talking much longer time scales (30-50 years?). Have a look at network rail's modernisation plans. ERTMS (in cab signalling) is to be introduced over the next few decades and will include an ATO overlay - note this isn't the same as "driverless" it's more "autopilot for drivers". The timetable for the implementation keeps slipping due to costs.

- large areas of the network are still signalled under Victorian absolute block, and will continue to be signalled in this way for quite a few decades to come (again with reference to NR's modification plans). Driverless (or even ERTMS/ATO) won't be making an appearance until these areas have been swept away.

In short, my my concern isn't that the job will disappear in the next 30 years, it's the extent to which creeping automation may make it boring to the point where it becomes unbearable.

People often cite the DLR as a "driverless railway" which is largely is - although every train is still crewed. But you'll note this has existed since the 1980s and was a purpose built, brand new network. Here we are some 30 years after it opened and TOCs are heavily recruiting trainee and qualified mainline drivers. Draw your own conclusions from that.

 

[The Ham]

Even if driverless trains started to come into service in 10 years time they are not going to be network wide for quite a few years after that. As such someone starting out now on a network that could be one of the first would have probably about 15 years of service before needing to seek an alternative TOC to drive for (i.e 25 years from now)

Even then it's going to take a long time before the whole network is automated, and chances are natural wastage will mean that for most entering employment as a driver as metro services start to be automated would still probably be a good move.

You also have to look at things like Crossrail, HS2, Crossrail 2 and other new lines and improved Sunday services to see that demand for drivers is still very much in the way up and so shouldn't be a problem for quite some time.

 

[Need2]

Looking for some advice.
I have recently passed my psychometrics and just waiting for confirmation of either being given a start date or being put in the holding pool to train as a driver for tyne and wear metro service.

I assume you have also passed the mmi, dmi and medical?

 

[Bletchleyite]

I can't see the UK ever going for completely staffless trains - we just aren't culturally right for that. So if a network went driverless, I'd expect it would move to "guard only operation" like the DLR, so a move into that role would probably be feasible if it was a DOO route. (If not DOO the guards may get first dibs on those jobs, I suppose).

 

[daikilo]

I can't see the UK ever going for completely staffless trains - we just aren't culturally right for that. So if a network went driverless, I'd expect it would move to "guard only operation" like the DLR, so a move into that role would probably be feasible if it was a DOO route. (If not DOO the guards may get first dibs on those jobs, I suppose).

Neil, what change/s to British culture would be required to make "driverless" or "unstaffed" trains acceptable? Note that DLR is not "guard only operation" as the role is not that of a guard as the person needs to be able to drive under certain circumstances.

However, the probability is that the new segregated lines would be the first to automate driving followed by small segregated networks being "rebuilt". This would likely require period of closure of several months.

I think the concern about driving "becoming boring" is a fair concern, but in risk analysis, a bored driver is often worse than none at all, so there must be a step change (the DLR is a good example, where the train captain only intervenes in the driving role when required and is not forced to sit at the front all the time).

 

[Bromley boy]

Neil, what change/s to British culture would be required to make "driverless" or "unstaffed" trains acceptable? Note that DLR is not "guard only operation" as the role is not that of a guard as the person needs to be able to drive under certain circumstances.

However, the probability is that the new segregated lines would be the first to automate driving followed by small segregated networks being "rebuilt". This would likely require period of closure of several months.

I think the concern about driving "becoming boring" is a fair concern, but in risk analysis, a bored driver is often worse than none at all, so there must be a step change (the DLR is a good example, where the train captain only intervenes in the driving role when required and is not forced to sit at the front all the time).

I guess the DLR is closer to guard only operation than it is to driver only operation - the train captain only drives in very limited circumstances and can do so on line of sight (essentially to get a stalled train out of trouble if the signalling goes down). The driving aspect of their role, although a requirement, is not part of their day to day duties.

Fully unattended operation may well come in eventually on the tube (and has already on some tube style metros around the world). The idea of having no staff members on board at all becomes a lot more problematic when rolled out to mainline trains carrying people at higher speeds over greater distances and through remote areas.

 

[Mintona]

Having seen how proficient Network Rail are at electrifying a main line, I wouldn't be too concerned about their ability to automate everything.

 

[Bletchleyite]

Neil, what change/s to British culture would be required to make "driverless" or "unstaffed" trains acceptable? Note that DLR is not "guard only operation" as the role is not that of a guard as the person needs to be able to drive under certain circumstances.

Better behaviour, and a better response to emergencies.

As for "GOO" - a DLR Train Captain is more like a guard with basic driving skills than a driver.

 

[Bletchleyite]

However, the probability is that the new segregated lines would be the first to automate driving followed by small segregated networks being "rebuilt". This would likely require period of closure of several months.

I know there's still a driver at the front, normally just operating the doors and pressing "go"[1], but a conversion to ATO hasn't required "a period of closure of several months" on any of the LU lines so converted.

[1] The deep Tube lines are much more overcrowded than the DLR, and on-board revenue protection is rare, so moving him out into the train would have limited benefit.

 

[rebmcr]

I was dreading this thread from the title — it appeared that the Metro newspaper might have published a nonsense pop-science article!

 

[Broot]

I assume you have also passed the mmi, dmi and medical?

MMI yes - at the same time as Psychometric.
DMI yes - before the testing.
medical - not yet im not concerned with this.

 

[NotATrainspott]

I see that people are falling into the normal trap of assuming that automation will require novel signalling systems to be installed deliberately. When autonomous vehicles will be able to handle travelling in two dimensions in a far wider range of circumstances, it's laughable to suggest that they won't be able to handle driving in one.

However, it's something I don't see arriving overnight. For one, the degree of efficiency saving to be had is much, much smaller. The smaller the vehicle, the larger the relative cost of employing a driver. For an individual taxi driver, the cost of the driver is the largest component cost of the ride. For a 12 carriage commuter train, the driver won't be anywhere near as significant. Also, when staff are deliberately employed to drive already, there isn't really much of a safety or time-efficiency justification either. Commuters on trains can already spend their journey reading books or watching movies, so they won't gain much individually from the train driving itself.

Where you're likely to see autonomy first is in non-passenger service. The Elizabeth line will have this, as drivers will be able to get out of their cab and walk along the train to the other cab while the train drives itself into and out of the Westbourne Park sidings west of Paddington. That's not going to be needed when things go right, but if there are delays it'll allow trains to save a few minutes off their turnaround time after terminating at Paddington from the east. Automating ECS turns like that will have an easier safety and economic case (if a member of staff will be needed on board anyway in passenger service, the saving from getting rid of a driver job is minimal) while quite possibly making more driving jobs available. With ECS made cheaper and easier, you can run more passenger services which do require human drivers, while human drivers won't then need to spend their time doing donkey work.

Of course, that's not going to save all jobs forever. But, it means that train driving won't be as badly affected as trucking or taxi driving.

 

[bramling]

I see that people are falling into the normal trap of assuming that automation will require novel signalling systems to be installed deliberately. When autonomous vehicles will be able to handle travelling in two dimensions in a far wider range of circumstances, it's laughable to suggest that they won't be able to handle driving in one.

However, it's something I don't see arriving overnight. For one, the degree of efficiency saving to be had is much, much smaller. The smaller the vehicle, the larger the relative cost of employing a driver. For an individual taxi driver, the cost of the driver is the largest component cost of the ride. For a 12 carriage commuter train, the driver won't be anywhere near as significant. Also, when staff are deliberately employed to drive already, there isn't really much of a safety or time-efficiency justification either. Commuters on trains can already spend their journey reading books or watching movies, so they won't gain much individually from the train driving itself.

Where you're likely to see autonomy first is in non-passenger service. The Elizabeth line will have this, as drivers will be able to get out of their cab and walk along the train to the other cab while the train drives itself into and out of the Westbourne Park sidings west of Paddington. That's not going to be needed when things go right, but if there are delays it'll allow trains to save a few minutes off their turnaround time after terminating at Paddington from the east. Automating ECS turns like that will have an easier safety and economic case (if a member of staff will be needed on board anyway in passenger service, the saving from getting rid of a driver job is minimal) while quite possibly making more driving jobs available. With ECS made cheaper and easier, you can run more passenger services which do require human drivers, while human drivers won't then need to spend their time doing donkey work.

Of course, that's not going to save all jobs forever. But, it means that train driving won't be as badly affected as trucking or taxi driving.

I think this post is spot on. The issue is not so much whether a driverless train is technically feasible, but whether the benefits are sufficiently worthwhile. In many cases I suspect the answer to that is that they may not be.

 

[Broot]

Thanks for all replys.
So the general consensus is that the roll isn't going anywhere in the next decade and beyond.
I like the comment about the relative cost of the driver compared to the size of the vehicle - not something I'd considered. Cheers

 

[Chester1]

The DLR works because it is both segregated and mostly elevated. It would be extremely expensive to alter exisiting lines to its type of system. The technology needs to become much more advanced before its viable for full automation on normal lines. I agree with other posters that it will be a generation or two before technology does more than help a driver on the NR system. There are several thousand level crossings that would need to be closed first before driverless trains could even be considered.

 

[gallafent]

There are several thousand level crossings that would need to be closed first before driverless trains could even be considered.

Why? A decent train-driving AI should be able to handle quite a bit of input data — for example taking feeds from all the CCTV on and around the crossing, in order to be able to assess imminent (rather than only current) dangers, as well as having a view from the front, sides, and rear of the train (which it would of course also have, using fusion of optical, infra-red, ultrasound, LIDAR, ………). In that way it would have a significant advantage over a human driver, who is not capable of handling all those streams of data at once, and only has access to a small subset of them.

 

[daikilo]

Why? A decent train-driving AI should be able to handle quite a bit of input data — for example taking feeds from all the CCTV on and around the crossing, in order to be able to assess imminent (rather than only current) dangers, as well as having a view from the front, sides, and rear of the train (which it would of course also have, using fusion of optical, infra-red, ultrasound, LIDAR, ………). In that way it would have a significant advantage over a human driver, who is not capable of handling all those streams of data at once, and only has access to a small subset of them.

Assuming that your equation is "how to stop the train if the crossing is obstructed", even a human driver can do it if s/he receives a warning in due time. The issue is that most automatic crossings are not designed for that, thus the AI "driver" will be no/little better off. In fact, your explanation is around how to know whether a crossing is clear and not around how to drive a train.

 

[talltim]

There are several thousand level crossings that would need to be closed first before driverless trains could even be considered.

I too don;t really understand this argument. A human driver can't do much to prevent accidents caused by misuse or failure of a crossing except in very slow speed cases. An autonomous train with the same inputs as a human would be in much the same position but with faster reactions. An autonomous train with access the sensors on the crossing would be in a better position.
Public image is a different matter. If a train hits someone misusing the crossing then all the blame is on the mis-user and the driver (and by extension the railway) gets the sympathy. If an autonomous train hits a mis-user then I'm sure some ire would be directed towards the railway for not producing trains that can defy physics!
Edit: I think the problem is that you are thinking of an automated system, whereas we are thinking autonomous individual trains

 

[Dave1987]

Why? A decent train-driving AI should be able to handle quite a bit of input data — for example taking feeds from all the CCTV on and around the crossing, in order to be able to assess imminent (rather than only current) dangers, as well as having a view from the front, sides, and rear of the train (which it would of course also have, using fusion of optical, infra-red, ultrasound, LIDAR, ………). In that way it would have a significant advantage over a human driver, who is not capable of handling all those streams of data at once, and only has access to a small subset of them.

Your comments remind me of people who design systems in their cosy little office and have never been out and about in the real world who are then bewildered as to why the system they have designed in their perfect world on their computer doesn't work in the real world. Happens over and over again. Many many millions of pounds have wasted over engineering things that don't work or don't work reliably under real world conditions. I can think of problems that of happened very recently with stuff that over engineered to nth degree but stumbles at the first hurdle when not in the perfect world of the designers computer.

 

[daikilo]

I think that Galafent is right that the technology exists to ensure that crossings are free before trains are allowed through. My concern is that this implies that until "free" is demonstrated then trains must plan to stop before the crossing. It doesn't significantly matter whether it is wo/man or machine who achieves this. That said, are we ready to invest to make it possible, including the probable delay to either rail or road traffic?

 

[Chris M]

In terms of the present technology, the only area where an automated driver scores over a human driver is consistency, but a human driver scores over an automated one in flexibility because it takes a lot more time and effort to teach an automated driver a new route than it does a human. This means its only worth spending that time and effort when the consistency is crucial, and that is really only the case when you need to get the absolute maximum throughput out of your infrastructure which just doesn't happen outside of very high frequency metro operation, which is why even Thameslink is not going to ATO outside of the core section.

Now this may change, when the self-driving technology being developed for road vehicles develops, but driving on line of sight is significantly different to driving based on a signalling system like most railways. One of the biggest differences is the need to anticipate and extrapolate - things that humans are historically much better at that than computers. On a road a self-driving vehicle never needs to anticipate things further ahead than it can see, but on a railway you need to do anticipate much further ahead and apply information about your current surroundings to your future surroundings (e.g. there are lots of leaves on the embankment here, which means that there are likely to be many leaves on and around the track in the cutting round this next corner and I know that adhesion is often poor there when there are leaves on the line so I need to adjust my driving style appropriately before I get there.) This is not insoluble by any means, but it will take time.

 

[jon0844]

I was dreading this thread from the title — it appeared that the Metro newspaper might have published a nonsense pop-science article!

You weren't the only one.

I saw a thread on another forum recently that said 'Daily Mail Watch', which turned out to be a thread looking at what the paper said - not a watch made by the newspaper that probably reminded you to be outraged about different things every hour of the day.

 

[philthetube]

In terms of the present technology, the only area where an automated driver scores over a human driver is consistency, but a human driver scores over an automated one in flexibility because it takes a lot more time and effort to teach an automated driver a new route than it does a human. This means its only worth spending that time and effort when the consistency is crucial, and that is really only the case when you need to get the absolute maximum throughput out of your infrastructure which just doesn't happen outside of very high frequency metro operation, which is why even Thameslink is not going to ATO outside of the core section.

Agreed, the big advantage is capacity, automating the last few miles into London on most routes would have huge benefits, the further out you go and the smaller the gains.

 

[NotATrainspott]

Agreed, the big advantage is capacity, automating the last few miles into London on most routes would have huge benefits, the further out you go and the smaller the gains.

The Thameslink automation systems get progressively more applied as trains get closer to the core. Out at the far reaches of the network, it'll be off completely, but as trains approach the core it'll give the driver more and more guidance. This gives a smooth transition from full driver control to ATO and is absolutely vital for the core timetable to work. Without it, trains may well be able to get themselves through the core automatically but they won't arrive on time for the ATO systems to maximise capacity. You don't want trains turning up 10 seconds too early or 10 seconds too late due to minor variances in driver behaviour.

 

[Sunset route]

Why? A decent train-driving AI should be able to handle quite a bit of input data — for example taking feeds from all the CCTV on and around the crossing, in order to be able to assess imminent (rather than only current) dangers, as well as having a view from the front, sides, and rear of the train (which it would of course also have, using fusion of optical, infra-red, ultrasound, LIDAR, ………). In that way it would have a significant advantage over a human driver, who is not capable of handling all those streams of data at once, and only has access to a small subset of them.

What including every accommodation/occupation, public footpath, bridleway crossing in fact every user worked crossing in the country. That's a big undertaking. Sorting out public highways should prove easy apart from AHBs as they are interlocked with the signalling but all the rest including AHBs now that's where the problems will arise.

 

[A Challenge]

Pilot over the crossings, so it is safer

 

[ComUtoR]

but as trains approach the core it'll give the driver more and more guidance. This gives a smooth transition from full driver control to ATO and is absolutely vital for the core timetable to work.

In what way ?

Full manual control till Elephant. Press ATO button at Elephant. Unit drives itself to St Pancras/Kentish.

How will it transition and what 'guidance' will it give ? I was speaking to someone who was on the test trip and when the Driver moves the controller the ATO drops out.

Cheers in advance.

 

[ComUtoR]

What including every accommodation/occupation, public footpath, bridleway crossing in fact every user worked crossing in the country. That's a big undertaking.

Could it be achieved with a massive reduction in line speed ? We have a rather infamous one where the W board has been removed. To achieve that the PSR is now 40mph. If you saw someone on the crossing you could reasonably stop.

We do have a few where you would never be able to stop in time regardless as the crossings are on blind curvatures but could a reduction in speed to 20mph give a reasonable time to stop without the need for CCTV, Interlocking, etc.