Driverless trains - why limited progress on the national rail network?

Discussion in 'Traction & Rolling Stock' started by deltic, 6 Jan 2017.

Thread Status:
Not open for further replies.
  1. deltic

    deltic Established Member

    Messages:
    1,532
    Joined:
    8 Feb 2010
    This is not a discussion about the rights and wrongs of driverless trains but the technology which they require

    It is nearly 50 years since driverless trains were introduced on the Victoria Line and over time they have slowly been introduced on metro lines around the world including systems such as DLR.

    The automotive and software industries are investing heavily in driverless car/truck technology with breakthroughs and advances being announced all the time.

    As a lay person I would have expected the rail industry to be way ahead of the automotive sector in this regard but apart from metro systems little progress seems to have been made.

    What are the reasons for this and what techology is required to enable driverless trains to operate on the national rail network?
     
  2. Failed Unit

    Failed Unit Established Member

    Messages:
    5,010
    Joined:
    26 Jan 2009
    Location:
    Central Belt
    To me it is the obstacle detection,

    Does the technology exist yet to stop a train before it hits a tree that has fallen over the line? Or even a motor vehicle that has decided to run a red light. I know a driver can't stop such collisions happening but I suspect that they can start taking action quicker than a computer.

    Even the basics when we are driving - there is a hazard - I will watch it closely just in case.
     
  3. IKB

    IKB Member

    Messages:
    290
    Joined:
    15 Nov 2013
    Metros are usually undercover, low speed, one traction type, few conflicting movements, segregated from roads/public....all the things mainline railways are not.

    Comparions with the DLR are infuriating. This system was built to be driverless. The mainline is victorian railway with a slowly evolving signalling system.

    Oh and don't forget money. The train would still have human oversight, so there would still be a wage cost, although perhaps reduced over time as those on old contracts retire in favour of 'operatives' over drivers. And then there is the huge capitol outlay to upgrade the trains/signalling, which isn't particularly attractive if the net capacity benefit is minimal.
     
  4. Bletchleyite

    Bletchleyite Veteran Member

    Messages:
    20,839
    Joined:
    20 Oct 2014
    Location:
    Up and down the south WCML (mostly)
    A driver doesn't necessarily either.

    FWIW at Manchester Airport there are "trip lines" above the track which are intended to cause all signals to go to red in the event of an aircraft fouling them. Something like that could deal with the tree issue - line broken? All trains stop. It could cause disruption, but then you could allow for DLR-style manual driving by the "guard" on sight at low speed to avoid the system stopping entirely.
    --- old post above --- --- new post below ---
    This is true. HS2, OTOH, is new-build - is there any reason that should not be fully automated from day one?

    I'd expect to see DLR-style operation, i.e. "guard-only operation", in time. The ability to do this would revolutionise the cost base of branch lines while retaining the customer service role that people like to have and lose on DOO.

    If it's possible to make a self-driving car, it's possible to make a self-driving train that would by definition be safer than a self-driving car and probably just as safe as a manually driven train. It's inevitable - there are far fewer variables with a train, even a branch-line train on Victorian infrastructure.

    Anyone for giving it a go with the Stourbridge shuttle, perhaps?
     
    Last edited: 6 Jan 2017
  5. Bodiddly

    Bodiddly Member

    Messages:
    573
    Joined:
    7 Feb 2013
    Location:
    The people's republic of Inverness
    The technology for the rolling stock is already there, has been for years. In the case of national heavy rail it's the infrastructure that's nowhere near ready for automation. As for the human v computer argument, computers will almost always out do humans. A camera linked to a computer will 'see' an obstruction on the line and start evasive action far quicker than a human ever could.
    Automation is coming in most sectors, like it or not. I worry about the future my kids will grow up in. The workplace will change beyond recognition in the next 20-30 years and the difficulty in gaining willful employment will become a stark reality for most.
     
  6. bramling

    bramling Established Member

    Messages:
    2,925
    Joined:
    5 Mar 2012
    Location:
    South-East / Barnard Castle
    Aside from obstacle detection, the simple reality is we don't have an ATO system which is capable of driving trains as good as humans do in differing weather / adhesion conditions.

    I think ATO on some of the sub-surface lines will really show up this weakness, currently most of the ATO on London Underground is in tunnel sections.

    Sure you can have it, but can the system drive the train is fast as possible for the conditions prevailing, which is what a human is trained to do? The higher the line speed, the more this starts to bite as you have to allow for worst-case scenarios. You don't want trains going unnecessarily slowly (bad because it increases journey times), but equally you don't want them sliding (may exceed their limit of movement authority which in the worst cases could mean a collision, plus you don't want wheels being flatted all the time either).
     
    Last edited: 6 Jan 2017
  7. paulweaver

    paulweaver Established Member

    Messages:
    2,332
    Joined:
    3 Oct 2015
    Location:
    Nantwich
    Stourbridge, the Abbey Line, some Cornish Branch lines, all feel rather self-contained and ripe for driverless.


    If a car can detect people crossing lines, other cars at 4-way stop signs, etc, I'm sure the technology to detect such problems is relatively trivial. Human reaction times will be far slower than a computer.

    Computers are already moving into areas that people thought were safe:


    https://www.theguardian.com/technol...-intelligence-ai-fukoku-mutual-life-insurance
    Doesn't matter how little you pay your staff, there will always be a point when it makes sense to replace them. Fundamentally a robot is better than a slave, the robot works tirelessly, any mistakes can be fixed - and the fix applies to the whole workforce.

    http://www.bbc.co.uk/news/technology-36376966
    The McDonalds quote is stupid of course, it's cheaper to buy a $10,000 robotic arm than to hire an empoyee who is inefficent making $5 an hour. It's cheaper to buy a $1,000 robotic arm than to hire an empoyee who is inefficent making 50c an hour.

    Technology prices are constantly falling, it's just a matter of time.
     
  8. Stow

    Stow Member

    Messages:
    55
    Joined:
    12 Oct 2015
    Some of the main issues;
    - Segregation/fencing on all lines
    - Level crossings
    - Freight
    - Cost of re signalling and new rolling stock
    - Standards (interoperability)
     
  9. paulweaver

    paulweaver Established Member

    Messages:
    2,332
    Joined:
    3 Oct 2015
    Location:
    Nantwich
    This is because the code hasn't been written, not because of any inherent problem with automation. The main area of work is in replacing cars. 75 million cars get sold each year, if you make a £1000 profit on selling automation on those cars, that's a £75b a year profit.

    How many trains are there in the UK? What's the average replacement time on those trains?
     
  10. Bletchleyite

    Bletchleyite Veteran Member

    Messages:
    20,839
    Joined:
    20 Oct 2014
    Location:
    Up and down the south WCML (mostly)
    Don't we? A computer can make decisions far more quickly than a human can, and is far less likely to make those decisions incorrectly.

    If a self-driving car is possible, a self-driving train *absolutely* is.
     
  11. bramling

    bramling Established Member

    Messages:
    2,925
    Joined:
    5 Mar 2012
    Location:
    South-East / Barnard Castle
    Yes it is because of an inherent problem, not so much with automation, but with the concept of steel wheel on steel rail.

    Feel free to explain how you would design such a system ...

    You could put in bare-minimum brake rate all over the place, but this would heavily extend journey times. This is what the Jubilee and Northern lines do.

    You could tolerate having overruns all over the place. This is what the Central Line does.

    DLR varies their braking rate on a daily basis, but this wouldn't work on the mainline as it would wreck the timetable, and how do you decide which brake rate to use and where?

    All this hassle, for what benefit?
     
  12. Bletchleyite

    Bletchleyite Veteran Member

    Messages:
    20,839
    Joined:
    20 Oct 2014
    Location:
    Up and down the south WCML (mostly)
    The Stourbridge shuttle would to me be a sensible first place to try mainline "guard-only operation". Basic and standalone. No level crossings.
     
  13. bramling

    bramling Established Member

    Messages:
    2,925
    Joined:
    5 Mar 2012
    Location:
    South-East / Barnard Castle
    None of the ATO systems we currently have in Britain adequately address the issue, in fact they're not even close.
     
  14. gordonthemoron

    gordonthemoron Established Member

    Messages:
    5,336
    Joined:
    4 Sep 2006
    Location:
    Milton Keynes
    From my understanding of Nuremburg U2 & U3, making them driverless did not result in any job losses. However, U2 & U3 now have 2 car trains which run twice as frequently as U1 which has drivers and 4 coach trains. So it appears that the remote drivers are controlling 2 trains at a time
     
  15. Bletchleyite

    Bletchleyite Veteran Member

    Messages:
    20,839
    Joined:
    20 Oct 2014
    Location:
    Up and down the south WCML (mostly)
    A human also does that. If they didn't, they would be driving dangerously. And (unlike car drivers) a professional train driver does not drive dangerously.

    A massive financial saving, primarily. And allowing single-person operation without a large swathe of the problems DOO causes.
    --- old post above --- --- new post below ---
    That does not preclude developing one that does adequately address the issue. In particular one based on the concept of self-driving cars, i.e. giving the vehicle a large degree of automation.
     
    Last edited: 6 Jan 2017
  16. jcollins

    jcollins Veteran Member

    Messages:
    25,692
    Joined:
    23 Jan 2009
    I think to allow that ideally you would need some mileage driven by the on board member of staff so they aren't out-of-practice to take over the driving duties when the computer fails. Also the chance of the system failing would have to be lower than the chance of a healthy human driver collapsing.
     
  17. bramling

    bramling Established Member

    Messages:
    2,925
    Joined:
    5 Mar 2012
    Location:
    South-East / Barnard Castle
    The difference is the human does it on the spot, and generally has a high level of success is making the right decision.

    To clarify, the DLR sets one brake rate for whole system. You can get away with this on a small system that is concentrated in a small geographic area. You simply can't do this for mainline trains which cover a wide area, and travel at higher speed so the consequences of making the wrong choice could be severe - both in terms of cost, and perhaps safety.
     
  18. Bletchleyite

    Bletchleyite Veteran Member

    Messages:
    20,839
    Joined:
    20 Oct 2014
    Location:
    Up and down the south WCML (mostly)
    Safety-critical technology has been about for years, and so the chance of a computer making an error is *far* lower than the chance of a human doing so. And in the end, if the computer gets it wrong[1], unlike an aircraft, you have the option of just stopping everything.

    [1] As with some aircraft, you may have two or three parallel systems which would conclude to stop the train if they don't agree on the course of action.
     
  19. GB

    GB Established Member

    Messages:
    4,867
    Joined:
    16 Nov 2008
    Location:
    Somewhere
    ...and I'd argue a human can react to a developing situation better than a camera and computer would.
     
  20. Bletchleyite

    Bletchleyite Veteran Member

    Messages:
    20,839
    Joined:
    20 Oct 2014
    Location:
    Up and down the south WCML (mostly)
    I didn't know that - but remember the DLR system is well over 20 years old. The idea of a single 153, say, learning the appropriate brake rate as it goes, analysing the amount of wheelslip and railhead conditions, is much more feasible than it was 20 years ago.
    --- old post above --- --- new post below ---
    I'd disagree. A computer can't panic. Provided it has the correct inputs and the correct programming, it will take the correct, safe action, bearing in mind that in almost all cases where there is any doubt on the railway the correct action is to stop the train.

    That of course has issues in road traffic - drivers often take risks in aid of progress. But on a railway you want safety-above-all, and a computer is perfectly placed to deliver that.
     
    Last edited: 6 Jan 2017
  21. paulweaver

    paulweaver Established Member

    Messages:
    2,332
    Joined:
    3 Oct 2015
    Location:
    Nantwich
    I wouldn't, I wouldn't design a self driving car either, or an aircraft autopilot.

    Fundamentally you have a driver that takes inputs and adjusts their outputs based on those inputs and their experience. That's exactly what computers do.

    There's no fundamental problem with designing such a system. Getting the right equation to determine the is tricky part, and the way I'd approach it would be to install a multitude of sensors on trains (including rail adhesion, local weather (multple temperature/humidity sensors, cameras in various wavelengths, take inputs from other trains etc), location, train weight and distribution of that weight, etc, gather data of real drivers over a few years, build a model which would result in the same decisions that train drivers make based on those inputs.

    Once you've done that it's a matter of implementation.

    Detecting a tree on the track is far trickier problem.

    Machines can diagnose diseases that doctors with years of training miss, I don't see why you'd think that machines couldn't determine how hard to apply brakes.
     
  22. Chrisyd

    Chrisyd Member

    Messages:
    105
    Joined:
    16 May 2015
    There was this Long Read article on The Guardian website about just this idea and how this can go wrong: https://www.theguardian.com/technology/2016/oct/11/crash-how-computers-are-setting-us-up-disaster

     
    Last edited by a moderator: 6 Jan 2017
  23. najaB

    najaB Veteran Member

    Messages:
    15,366
    Joined:
    28 Aug 2011
    Location:
    Scotland
    Just to point out that verifying both of the highlighted terms is a non-trivial exercise. An even harder task is proving that the programming can deal appropriately with incorrect input. In aircraft this is done by having the automation bail-out and hand control back to the humans.

    Largely, the challenges standing in the way of driverless trains are largely the same as those being faced in the development of full ETCS Level 3: ensuring train integrity, reliable accurate determination of position and velocity and continuous evaluation of rail/wheel interface conditions.
     
  24. dk1

    dk1 Established Member

    Messages:
    4,964
    Joined:
    2 Oct 2009
    Location:
    East Anglia
    Only another 8 years then I will be 56 & can happily retire with 40yrs in the pension scheme :p Can't see any movement apart from the partial XR & TL before then.
     
  25. Sunset route

    Sunset route Member

    Messages:
    572
    Joined:
    27 Oct 2015
    It's going to interesting to authorise a computer to pass a signal/marker board at danger over several sets of points that have been manually wound and thus have no detection due to a failure.
     
  26. Bletchleyite

    Bletchleyite Veteran Member

    Messages:
    20,839
    Joined:
    20 Oct 2014
    Location:
    Up and down the south WCML (mostly)
    Same on a train, potentially. I can't see much case for unstaffed mainline trains - it's "guard only operation" that I'd see as a most likely scenario. If the computer can't cope, brakes on and "Will the guard please proceed to the control panel at the front of the train" on the PIS.
     
  27. IKB

    IKB Member

    Messages:
    290
    Joined:
    15 Nov 2013
    On the subject of automation generally, there are bigger issues for mankind to consider here - i.e. what will the jobs market look like in 20, 30 or 40 years time. If robots lead to a reduction in semi or unskilled labour, what form of gainful employment does a society offer its citizens. The number of robots, for example, in Amazon warehouses has increased exponentially in recent years. Multiply that to transportation, service industries, banking etc. Will those pushed out by technology find work elsewhere? If not will the state have to support them, or will they turn to crime etc.

    A lot of tech solutions seem to be solutions looking for problems, driven by those who seek to gain financially from selling the tech and those hoping to reduce the cost of labour.
     
  28. Bletchleyite

    Bletchleyite Veteran Member

    Messages:
    20,839
    Joined:
    20 Oct 2014
    Location:
    Up and down the south WCML (mostly)
    The train would be manually driven on sight at low speed in such a case.
    --- old post above --- --- new post below ---
    And that is where things like "basic income"/"Grundeinkommen" come in. It's more a societal/political problem than a reason to stop the automation.
     
  29. IKB

    IKB Member

    Messages:
    290
    Joined:
    15 Nov 2013
    So the wage saving is minimal, as you'd still have to employ one person to oversee the train (as per a DOO driver). Unless of course that person is employed on substantially reduced T&Cs than a driver at present.

    If wage saving is limited and the net capacity gain is minimal (or worse if the braking profile is geared for worst conditions)......then the billions of cost expenditure in upgrading the system is justified and recouped how...?
    --- old post above --- --- new post below ---
    Just because you can automate something doesn't mean you should, for numerous moral and ethical reasons, among many others.
     
  30. bahnause

    bahnause Member

    Messages:
    39
    Joined:
    30 Dec 2016
    Location:
    bülach (switzerland)
    Driverless trains on the mainline are expensive. The high capacity of such trains makes the savings per passeneger / ton of cargo relativly small, nevertheless there is expensive development to be done and additional equipment for trains and infrastructure is necessary. On a completely new line without connections to other lines? Maybe for capacity reasons, mostly in urban areas. On exisiting tracks? Probably not in the next 25 years.

    A lot of people simplify the problems and tell us, how easy and cheap it must be. But No one has delivered so far.

    A much underrated Problem ist indeed trackhead conditions. Even the most modern trains struggle with changing conditions. I was driving a rather sophisticated doubledecker train today in snowy and often changing conditions (https://en.wikipedia.org/wiki/Stadler_KISS). Both acceleration and braking required a lot of manual input to make it a pleasant and safe ride for passengers.

    We still struggle as well with the ETCS Odometry, despite having rev counters on each axle of the loco and a radar. Distance measurement still gets upset when the slightest wheelslip occurs.
     
Thread Status:
Not open for further replies.

Share This Page