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. NSEFAN

    NSEFAN Established Member

    Messages:
    2,827
    Joined:
    17 Jun 2007
    Location:
    Southampton
    A sufficiently high reliability is always possible if you're willing to pay for it. The sensors may have redundancy to cover for failure, but does the communication link have similar redundancy, for example?
     
  2. Billy A

    Billy A Member

    Messages:
    81
    Joined:
    9 Jan 2017
    Because they're usually coming from those other countries with a larger supply of working age people? The girl dishing out the coffee in your local Costa is from Poland or somewhere near, not Germany or Sweden (unless your Costas are very different to mine).
     
  3. Dave1987

    Dave1987 Established Member

    Messages:
    3,728
    Joined:
    20 Oct 2012
    Yes if the CPDLC fails then the pilots revert back to broadcasting their long/lat on HF at specified points. There are many real world examples of how the tech gurus in their perfect worlds have tried to engineer things like Mother Nature out of the equation then 5 yrs down the line the fitters discover that 5 yrs of being in service 20 hrs a day and being battered by the elements they start failing on a regular occurrence.
     
  4. najaB

    najaB Veteran Member

    Messages:
    15,392
    Joined:
    28 Aug 2011
    Location:
    Scotland
    Except there was never even a Pan situation, never mind mayday. The aircraft was flown into the ocean. They were likely under 5000 feet before they even left 1G flight.
     
  5. GB

    GB Established Member

    Messages:
    4,867
    Joined:
    16 Nov 2008
    Location:
    Somewhere
    I feel we have drifted more towards flight than railways but how would a pilotless plane have coped with Flight 1549 that ditched in the Hudson with not one loss of life?
     
  6. najaB

    najaB Veteran Member

    Messages:
    15,392
    Joined:
    28 Aug 2011
    Location:
    Scotland
    That depends on if anyone had thought it a likely enough situation to actually program for - I'm guessing that they would not. My guess is that the plane would have impacted the ground somewhere between where the engines failed and the closest airport, likely with major loss of life both on the plane and on the ground.
     
  7. NotATrainspott

    NotATrainspott Established Member

    Messages:
    2,394
    Joined:
    2 Feb 2013
    This is how planes have operated in the past and how they operate now. This does not mean they will operate that way in future. Technology is developing faster and faster and it will mean that things that haven't been possible suddenly will be. Unless you're looking forward to the future and seeing how a small startup in San Francisco can pull together a few technologies and build something remarkable you'll simply be left behind. Just as big auto makers have gone on shopping sprees buying up small autonomous driving shops in the Bay Area, Airbus and Boeing will have little option but to do the same as otherwise their competition will and they will lose out. As I said, the commerical advantage of using autonomous aircraft for cargo planes is enormous and in a thin-margin commoditised market with big customers like Amazon now competing themselves you have little choice but to adopt them as soon as they're available.

    That isn't how I read the Wikipedia summary. Also, remember that with the constant communication the plane's progress can be compared against its planned flightplan and analysed for statistical anomalies. An AI would have noticed something was wrong here from the data even if the pilots didn't. If the ground co-cockpit AI can tell the pilots that their pitot tubes have probably failed (remember, it'll have all of the weather data as well!) then the pilots can start to rectify the situation. If your pilots are no longer doing high-level problem solving then they really don't have much of a purpose any more.

    You don't program AI like that. Modern AI systems learn from experience and use that learning to handle novel situations, just like a human pilot would. The AI systems have an ultimate goal of landing the plane safely and they'll work out what resources they've got available to manage that - again, just like a human would. It would be able to take into account all of the information it had about the situation and work out how much flight time it had left, and whether it would be able to make it back to a runway. It would do all of this in a fraction of the time needed for a human to do the same. If the only option would be to ditch, then it would have learned already how best to ditch a plane, just as human pilots are taught to ditch aircraft. The AI systems can be trained up on huge numbers of simulations, and unlike a human pilot there is no need to train up each aircraft individually.

    With this sort of technology, costs do not scale linearly with the number deployed. The best way for NR to pay for the cost of developing one of these CCTV posts would be to cover the entire railway network with them, as the cost of developing one is a huge amount but the cost of building the second one is really just the cost of materials. Adding some amount of overlap in sensing areas isn't going to break the bank and could even be more affordable if the cost of fixing a single failure is high.
     
  8. GB

    GB Established Member

    Messages:
    4,867
    Joined:
    16 Nov 2008
    Location:
    Somewhere
    How does your AI cope with crosswinds that are above operating range or what happens if there is a problem with the ILS (either on the plane or on the ground) requiring a visual approach?

    Let me guess, the AI will be programmed or the AI will learn?...:roll:
     
  9. najaB

    najaB Veteran Member

    Messages:
    15,392
    Joined:
    28 Aug 2011
    Location:
    Scotland
    That's the problem. Ditching was a counter-intuitive action. Instructions from ATC were to head to Teterboro Airport and technically given the height and velocity they should have been able to make it. But Sully felt how the aircraft was flying and decided to go into the river - in part because he could see that there were vessels on the water able to provide assistance, and due to a reluctance to take his aircraft over densely populated areas.

    Good luck codifying those sorts of judgement calls into an algorithm.
     
  10. FenMan

    FenMan Member

    Messages:
    658
    Joined:
    13 Oct 2011
    Simple demographics. The UK needs significant net immigration for the next 15 years merely to cope with the baby boomer generation leaving the workforce. The historically low UK birth rate means there are insufficient new locals joining the working population, so the gap has to be bridged somehow. There are already significant shortfalls of suitably qualified and capable workers in parts of the UK economy and the trend is for this to be a growing problem. Employers (notably the NHS and the construction industry) are trying to plug the gaps by attracting workers from overseas.

    Very few politicians want to talk about this, for some reason or other.

    Anyway, back to the thread ...
     
  11. NotATrainspott

    NotATrainspott Established Member

    Messages:
    2,394
    Joined:
    2 Feb 2013
    However, the AI would have an even better feeling of how the plane is flying because it would have every single data point created by every single sensor on board at its disposal. It would know exactly what state the engines were in and whether or not any thrust was possible, or how likely it would be that the engines would fail before landing. Since it would be able to make all of these decisions faster than the human it would have even more time to get to an airport.

    Remember that I did say the first pilot-less planes will be cargo ones, so that there will be zero humans on board rather than the cockpit crew alone. Should a cargo plane have this sort of situation you can simply write it off and purposely ditch or crash it into the water as appropriate. After years of fault-free autonomous flying even when these sorts of situations happen there will be no good reason to have human pilots on board passenger ones.
     
  12. carriageline

    carriageline Established Member

    Messages:
    1,698
    Joined:
    11 Jan 2012


    I agree with what you say to an extent, but for the railways to go over to true automation would require a complete rewrite on how it designs, procures and operate. Of course, not a impossible task.

    Oh and the app you talk of? It would not be deployable today, too much would have to change, and billion would have to be spent to even get to a stage where that information would even be ready for an app to pick it and display it.


    Sent from my iPhone using Tapatalk
     
  13. 455driver

    455driver Veteran Member

    Messages:
    11,332
    Joined:
    10 May 2010
    You are basing your knowledge on what is written on Wikipedia.

    Righto then!
     
  14. 455driver

    455driver Veteran Member

    Messages:
    11,332
    Joined:
    10 May 2010
    And yet we have hundreds of thousands of able bodied people out of work!
     
  15. carriageline

    carriageline Established Member

    Messages:
    1,698
    Joined:
    11 Jan 2012
    The though of a driverless and staffless railway scare me, but it's gonna happen, hopefully long after I have gone. Technology is advancing at a crazy rate. I know the railway isn't ready for automation, too much has to change, but to answer some of your questions!

    Why would it need to? Wheel speed sensors on each wheel could soon work out there is a dragging brake. Then a man in the control room can decide on information the train is giving him.



    Of course it can, the train will know the horn is faulty, and can then run at a set speed. TSRs won't need lights, as everything will be in cab based, and speeds will be transferred as part of the movement authority.





    I've asked this question before, which no one can really answer (apart from a man in the control centre watching the video, from a filthy camera covered in grime and dead bugs!)

    Anything is possible if you are willing to throw enough money at it!


    Sent from my iPhone using Tapatalk
     
    Last edited: 17 Feb 2017
  16. NotATrainspott

    NotATrainspott Established Member

    Messages:
    2,394
    Joined:
    2 Feb 2013
    However, that underlying technological change is happening anyway. For instance, the entirety of the new Crossrail infrastructure is included in one enormous BIM model. After completion, that model isn't just going to disappear off the face of the earth. Instead, it'll be the main plans used whenever maintenance or modifications are required. Instead of needing to go and find paper records and transcribe them into a computer model they'll be able to find them immediately at the click of a button. Since the tunnels will be fitted with 4G and WiFi, there is no good reason why its maintenance crews couldn't be given similar smart devices, which would then give them full access to that entire BIM model.

    The cost savings from having all railway records in a BIM model would be immense. When you know everything about the age, condition and formation of every part of the infrastructure you can plan maintenance and improvements far more efficiently. Of course, collecting this data in the first place isn't going to happen overnight, but remember that every time works do need done someone is adding things to a computer model anyway. You could do an enormous amount just by computer analysing all of the recordings made by the forward-facing CCTV cameras on board trains, giving you a sort of railway Street View.

    The section that I was reading is an accurate translation of the original French accident reports, as is the standard for Wikipedia articles about recent events with primary sources.

    And we'll have even more people out of work in the very near future. If we don't plan for the post-job society, then we won't have a society at all. The UBI will end up resulting in more jobs as people will end up doing things that they actually like doing rather than just whatever work they can find.
     
  17. najaB

    najaB Veteran Member

    Messages:
    15,392
    Joined:
    28 Aug 2011
    Location:
    Scotland
    No time was wasted attempting to diagnose the engine problem, they knew immediately what had happened.

    And you haven't touched on the key point - Sully made a counter-intuitive decisions to ditch the aircraft based on factors that an AI would have been unlikely to take into consideration. An AI would have tried to make it to the airport and likely would have crashed.
     
  18. Dave1987

    Dave1987 Established Member

    Messages:
    3,728
    Joined:
    20 Oct 2012
    Like I said before you don't really seem to have a true grasp on the way aircraft operate. Yes tech is developing. But it's developing to assist the human. Boeing has recently enabled the ability to pressurise the cabin down to 8000ft with a higher cabin pressure differential and with larger windows. That's a huge achievement but it's purely down to comfort. I don't believe there is any appetite to try and take the pilot out of the flight deck.

    It's really quite funny seeing your vision of how to automate everything. It's all this perfect world stuff that will easily fail in the real world.
     
  19. NotATrainspott

    NotATrainspott Established Member

    Messages:
    2,394
    Joined:
    2 Feb 2013
    What factors though? What human intuition is there about the state of the plane that would not also be apparent from sensor readings? If the NTSB say that he could have made it to Teterboro or LaGuardia, the why would an AI aircraft that tried to do the same thing not been able to either? Why, equally, would the AI discount the option of ditching any more than a human would? It's not as dumb as just saying water landing = bad. It would consider all of the same factors that led Sully to think that a ditching would be possible. It really wouldn't be that hard for an AI to pre-determine a variety of emergency scenarios ahead of time either, just as the Space Shuttle launches had a number of pre-planned emergency abort procedures. The AI could determine that based on the flight plan, if dual engine failure happened at T+X seconds it would have a 90% chance of making it back to the airport, while at T+Y seconds it would have to ditch. All of these things can be pre-planned in as a programmed or learned behaviour not just on one aircraft, but on every aircraft.
     
  20. NotATrainspott

    NotATrainspott Established Member

    Messages:
    2,394
    Joined:
    2 Feb 2013
    How aircraft operate today is not important here. It's how they could operate that matters, and how that will affect a company's bottom line.

    Why do you think that cargo airlines do not have an appetite to take the pilot out of the plane? Yes, of course there may be some industry executives who think the next twenty, thirty years will have changes just as gradual as the past twenty or thirty years but that's irrelevant, as they'll happily lead their companies into the ground just like so many industries that have been totally disrupted by new technology-focussed companies. Amazon has absolutely no qualms about putting money into autonomous cargo planes, because they're focussed on the bigger issue of getting things from suppliers to consumers rather than the smaller issue of running a cargo airline. That they simply do not care about doing things the way they've always been done is the reason why the company is so incredibly successful. With the amount of stuff they shift around the world and the tiny margins that they make on the things they sell, as sheer scale is a better long-term source of profit than individual margins, they have the economic incentive to spend billions on automated delivery lorries and warehouse automation and things that are well beyond the reach of smaller, less efficient companies. If you're FedEx and you've got your head screwed on you're already worried about the fact that they're willing to set up their own cargo airline just to move their own stuff around, and you know that if you don't invest in the technology yourself then you'll be overtaken completely.

    [​IMG]

    Seriously, an online bookstore has orders for 39 767-300F freighters! If I had told you this would happen only five, ten years ago then you would have thought I had gone totally mad.
     
  21. Dave1987

    Dave1987 Established Member

    Messages:
    3,728
    Joined:
    20 Oct 2012
    Ok then with dual engine failure what is powering the AI computer? The RAT only provides limited electrical and hydraulic power. How does it deploy the gear with no power to the gear hydraulics? The human would pull the manual gear extension and then visually check the gear is down, how does the AI do that? Remember you only have very limited power on the RAT.
     
  22. NotATrainspott

    NotATrainspott Established Member

    Messages:
    2,394
    Joined:
    2 Feb 2013
    You think too small! RAT obviously can't deal with the electrical load of a modern aircraft, and planes are only becoming more electrified to reduce weight and maintenance requirements. Hydraulic systems are heavy and are slowly on their way out for this reason. As a result, much more electrical power will necessarily need to be available anyway regardless of the state of cockpit automation. The most likely solution will be onboard batteries, drawing on the developments in the automotive field to have more capacity in less space with less weight. We will be seeing totally electric aircraft before too long, with one mooted possibility being to make the entire metal frame of the aircraft out of a charge-carrying material so that no mass is wasted.

    Why do you need a manual gear extension? It's like saying that a computer won't be able to drive a car because someone needs to turn the steering wheel for the wheels to change direction. Why a visual check of the landing gear if smartphone camera modules cost pennies and use milliwatts of power? All of these things might apply today but only because planes have developed so slowly.
     
  23. najaB

    najaB Veteran Member

    Messages:
    15,392
    Joined:
    28 Aug 2011
    Location:
    Scotland
    The NSTB determined that they *might* have been able to make a runway. To quote the report:
    Because ditching is normally the worse option. It was the combination of relatively still water and the fact that Sully saw that there were ferries and other craft nearby to render assistance that made it a viable option. I don't imagine that many aircraft are fitted with sufficiently advanced scene recognition technology to make that determination.
    Choice of Shuttle abort modes was largely defined by Sir Isaac Newton.
    And now you are making the point that you disagreed with so vehemently when I made it - if the logic isn't programmed in the computer isn't going to 'make it up as it goes along'.
     
  24. Dave1987

    Dave1987 Established Member

    Messages:
    3,728
    Joined:
    20 Oct 2012
    For a start the latest jets aren't made out of metal, they are made from a carbon composites which are very light but strong. You are completely delving into the world of fantasy. It all sounds completely like it works in my head so why shouldn't it work the real world. I've seen many many ideas from people in offices forced onto those on the front line because they believe it works on paper and in their head but in the real world it is either a rubbish design or simply cannot cope with real world conditions.
     
  25. jopsuk

    jopsuk Veteran Member

    Messages:
    11,140
    Joined:
    13 May 2008
    using 1980s Soviet technology the USSR successfully launched, orbited the planet with and landed (on a runway) their copy of the US space shuttle, unmanned in fully automatic mode in 1988
     
  26. najaB

    najaB Veteran Member

    Messages:
    15,392
    Joined:
    28 Aug 2011
    Location:
    Scotland
    Yup. And the Shuttle could do the same - with the exception of lowering the landing gear.

    Autonomous flight is relatively easy - if it's a specialist vehicle in restricted airspace. Regular, day to day commercial flying into congested airspace is a different thing altogether.
     
  27. oldsiggie

    oldsiggie Member

    Messages:
    429
    Joined:
    24 Sep 2014

    I know many people who do not have, or want a smartphone, and as for staff trackside having phones and switched on, that is a big no no, the COSS will, but no one else is allowed, Train Crew too must NOT have any mobile phone switched on whilst driving
    CCTV would also have to have back up as well, plus lighting, wifi connection are not good enough to be used, to have everything 'app' based is not a safe way of working.
    Sometimes all this technology is a step backwards not forwards,
     
  28. NotATrainspott

    NotATrainspott Established Member

    Messages:
    2,394
    Joined:
    2 Feb 2013
    And the AI would be able to work out the probability of success and the relative risk/reward of each option. It would pick the option which it had calculated had the best chance of getting everyone home safely again. This is exactly what a human pilot would do. If it determined that the risk of flying over a populated area was too high and that ditching was a less bad option, then it would ditch.

    This is a retort to what I expected you would come out with about the implausibility of an aircraft AI making all of these decisions instantaneously, especially if there were electrical problems. You can do both. Ahead of time, you can use AI to plan abort procedures using up-to-the-minute weather and status info so that if something does happen, the plan could be put into action immediately. The important point here is that these plans would be generated by AI. No airline would force their human pilots to come up with fresh abort procedures for every single flight they ever take, as that process is time-consuming and most likely wasteful.

    Of course, because the folks at NASA haven't got the slightest clue what they're on about:

    I mean seriously. Are you suggesting that Airbus are wasting their time when they're planning for electric aircraft? Michael O'Leary would fly planes powered by grass clippings if it meant he could have lower costs than the competition. Electrification allows some pretty remarkable innovation in aircraft design, as electric propulsion systems can be smaller and distributed around the aircraft in the optimum positions rather than slung low in big heavy lumps under the wings.
     
  29. oldsiggie

    oldsiggie Member

    Messages:
    429
    Joined:
    24 Sep 2014
    And that is why automation is not always a good thing, you cannot beat a human brain thinking things through. and as was proved in the film, the computer on 1549 said the plane would make it back and land ok, and in fact did it time and time again, until the pilot said, have you taken this into account, which of course they had not, then each time they ran the simulation it crashed into a populated area.
    A brain can think on its feet, a Computer can only do what its programmed to do, and in such rare events would never cope
     
  30. NotATrainspott

    NotATrainspott Established Member

    Messages:
    2,394
    Joined:
    2 Feb 2013
    You can't get rid of lineside telephones. All that mobile phone technology is dangerous and nowhere near good enough to be used.

    And then we get GSM-R. Which will then form the backbone of ETCS Level 3, so will be used as movement control for trains when there are no lineside signals or track counters whatsoever. GSM-R will be relied upon to keep trains from crashing into one another.

    If you are working for a company that issue you with a smart device, then your personal preference against smartphones is irrelevant. It's a work tool like any other at that point. If it's a company-issued rugged smartphone used for work purposes then that's a world away from a personal device that would be a dangerous distraction. The safety case for such a device is that it can report the location of all track workers and provide unambiguous information about the safety of the working environment, so that there will never be situations when workers have to jump out of the way of a train.
     
Thread Status:
Not open for further replies.

Share This Page