- Who in the railway industry is heading the drive for greater automation?
- Is it organisations such as TfL, or is it from the public who simply want train to run more efficiently? (I realise a number of people don't agree that ATO increases efficiency)
- What opinions do railway workers hold regarding their skills and the effect that automation will play on them?
- Do you see automation to be eroding your skills?
- Do you feel your skills are no longer being made valuable in the world of work?
- Where have you first noticed automation throughout your career as a train driver?
- What is new, what is missing?
- How do railway workers see themselves maintaining the skills necessary in the face of automation to continue within or beyond the railway sector?
- What part do you want your employers and/or trade unions to play?
- Do you think automation is eroding skills (like tacit skills) and if so, who do you think should have responsibility for ensuring skills are upheld?
Not off topic at all! In fact I'm quite interested in how the transition from driving older to new rolling stock feels for drivers.
Would you say that, given all the computerised equipment that is now present on trains, drivers can somewhat feel detached from the very vehicle that they are operating?
I ask this because interviews with Tube drivers on Geoff Marshall's videos discussing the end of old sub-surface rolling stock shows the drivers somewhat upset at having computers almost make the tacit feeling of driving a train become almost artificial.
Having talked with my dissertation supervisor, the term we've adopted is 'gamification' - instead of drivers using their senses and skills to identify faults, these faults are now fed directly to them through the Train Management System.
You just have to look how frequently the PIS goes wrong on trains to realise we are decades away from any chance of having systems reliable enough for full automation.
If they can't even get non-critical information systems operating properly right now, what hope have we got that a safety critical system can be relied upon without a driver keeping an eye on it ?
Just quickly on your final point I think it’s more that I wonder if engineering out the human should come after “upskilling” society so that it doesn’t result in leaving an ever wider wage gap. But I’m not sure if that would be overly relevant to your studies. As I said, more a political point of view.Very glad to hear from someone who came from the aviation industry! I happen to be an aviation enthusiast myself and have studied the way that the Instrument Landing System (ILS) operates (Glideslope, Localiser etc. etc.)
Having watched numerous Air Crash Investigation episodes with my father, one of the points that we always speak about is the gradual overreliance on automation, which I'm very interested to see that you have mentioned. Would you say that the gradual implementation of automation (both railway and aviation) has resulted in pilots and drivers alike sometimes being robbed of the knowledge and skills the job requires of them? (The second Airbus Golden Rule is to 'Use the appropriate level of automation at all times', which can be viewed in both polar opposites, one telling pilots to heavily focus on the use of automation, and the other telling them to use it less so when abnormal situations arise)
Your last paragraph refers very closely to a term called 'para-professionalism', where in this context people who are experts in one field tend to infiltrate another by claiming to offer some fancy solution that solves everything. This is most common in IT consultancy, where such a company claims to produce a piece of software that improves efficiency and workflow and reduces human labour etc. etc. Do you think this is strongly the case when it comes to automated railways, where companies like Thales with its SelTrac product try to offer a solution that is 'apparently' win-win when actually there may be so many hidden problems in the works?
That is a total fallacy. The reason non critical IT systems fail is that they are non critical, so money is not spent on making them resilient. Signalling IT is safety critical so has that resilience, and so doesn't tend to fail in that way. Same with the Airbus control software.
I presume the train is opening the doors a human has programmed it to at that location? In other words a data error rather than a 'dumb train'?On the trains I drive, the PIS is linked to a critical safety system. Without the PIS the train is too dumb to know how many doors it should open at a platform.
And sometimes even when it does know where it is, it still is too dumb to open the right amount of doors.
But agree it's down to poor investment in train design, due to the fact they knew the driver would be there to overcome these problems.
No, the programme knows how many doors to open at the station. It's just the train computer sometimes doesn't know where it is. Or other times, the PIS system crashes. Neither is good for something that is safety critical.I presume the train is opening the doors a human has programmed it to at that location? In other words a data error rather than a 'dumb train'?
No, the programme knows how many doors to open at the station. It's just the train computer sometimes doesn't know where it is. Or other times, the PIS system crashes. Neither is good for something that is safety critical.
So if they can't get that right on 3-year-old trains what hope have we got for full automation ?
But agree it's down to poor investment in train design, due to the fact they knew the driver would be there to overcome these problems.
Just as an aside this is a common misunderstanding. Alpha floor (stall) protection is not available on an A320 in the failure state of the aircraft that landed on the Hudson. What it did have is a more intuitive way of interacting with the flight controls that made holding the necessary angle of attack easier.I already answered that. If the system is overseen by a person or non-safety-critical it is simply not worth spending the money on the levels of testing or redundancy required to make it safety-critical (see also the reason for buying TrainFX at all - it's cheap). If however it is not, then you do need to spend that money.
As an example, ASDO is often done using GPS, which isn't 100% but it's good enough if it can occasionally pack in and have the human override it. But it's cheap. Whereas if you wanted to do a "perfect" SDO system that would near-never fail, you would do it using unpowered beacons at the start and end of the platform.
When did you last see an automated airport people mover (or something similar like the Lausanne metro) open its doors in the wrong place? I'd venture that far more drivers have given a wrong-side DOO release than has ever happened on that type of system. Or indeed the DLR? (Doors are released automatically on that).
A better illustration of safety critical systems going wrong is the 737 Max, where those elements were skipped, which to me is criminally negligent. I trust you avoid Airbuses because of the automation there? Personally, I'd rather Airbus than Boeing - I trust the computer far more than the squishy thing at the front, and the computer can do things the squishy thing can't, e.g. fly an aircraft on the edge of a stall without engines (I believe the captain of the A320 that went in the Hudson said he couldn't have done it in a Boeing, it would have been too easy to stall it).
Just as an aside this is a common misunderstanding. Alpha floor (stall) protection is not available on an A320 in the failure state of the aircraft that landed on the Hudson. What it did have is a more intuitive way of interacting with the flight controls that made holding the necessary angle of attack easier.
Whilst you may not trust the squishy thing as much as the computer it only takes two relatively low level failures for said computer to hand the aircraft back to that squishy thing and get them to solve the problem.
I’m not sure... yet. I’d expect single pilot opps then enhanced cabin crew first. Technology approval in aviation is glacial. Currently Centralised monitoring systems aren’t good at multiple failures. They almost need a form of AI to be able to think out the system and come up with a solution. Granted these failures are incredibly rare but they still happen.Ah, cheers.
Of course that is the case because the squishy thing is known to be there. If it wasn't even that approach would differ.
Again I don’t know enough to say how similar problems might affect rail services. The saving grace for the railway is that if a problem is too complex they can just stop.
Now how that translates into automation on the railway network? I can theorise but thankfully as yet, I’m not aware of any serious incidents that have resulted from a degradation in skills due to ATO. Predominantly because there isn’t really that much ATO in use and, none in complex systems. But I wonder if a driver suddenly needing to stop in low visibility or low adhesion conditions would be able to do so as well if they hadn’t actually handled the controls in a long time.
That of course makes a massive difference - in almost* every case on the railway, the correct answer in the event of any operating emergency is to throw the brake into emergency and worry about it afterwards. You'd also be unlikely to have totally unstaffed trains, so you likely do have a "squishy thing" there to intervene if needs be - it's the DLR model I'd expect to see (i.e. what I've referred to elsewhere on the Forum as "Guard-Only Operation"), not a totally unstaffed one. Even on roads there are considerable numbers of cases where stopping dead is dangerous - on railways there are almost none.
We are fortunate to have a well established case study for a "guard-only operated" automated railway - the DLR.
* There is the issue of stopping on viaducts etc, but the guard could drive forward manually on sight and at walking pace to clear that. The only other one I can think of is a runaway (e.g. the recent Caledonian Sleeper incident where I seem to recall a train in front was asked to run fast so the errant CS didn't catch it up) but it's incredibly rare.
I think the most likely long term solution is indeed the DLR model - a "Train Manager" sort able to do customer service, revenue and competent in emergency movement of the train and safety procedures in degraded / failure situations .
Yes, definitely. This is why I've long said DOO is a blip, and GOO is likely to be the future. Slowly, though, so I can't imagine any current driver finding their job threatened.
I already answered that. If the system is overseen by a person or non-safety-critical it is simply not worth spending the money on the levels of testing or redundancy required to make it safety-critical (see also the reason for buying TrainFX at all - it's cheap). If however it is not, then you do need to spend that money.
As an example, ASDO is often done using GPS, which isn't 100% but it's good enough if it can occasionally pack in and have the human override it. But it's cheap and provides a reasonable level of guard against the human mucking up. Whereas if you wanted to do a "perfect" SDO system that would near-never fail, you would do it using unpowered beacons at the start and end of the platform. But that is much more costly.
When did you last see an automated airport people mover (or something similar like the Lausanne metro) open its doors in the wrong place? I'd venture that far more drivers have given a wrong-side DOO release than has ever happened on that type of system. Or indeed the DLR? (Doors are released automatically on that).
A better illustration of safety critical systems going wrong is the 737 Max, where those elements were skipped, which to me is criminally negligent. I trust you avoid Airbuses because of the automation there? Personally, I'd rather Airbus than Boeing - I trust the computer far more than the squishy thing at the front, and the computer can do things the squishy thing can't, e.g. fly an aircraft on the edge of a stall without engines (I believe the captain of the A320 that went in the Hudson said he couldn't have done it in a Boeing, it would have been too easy to stall it).
Yes, exactly. It's a bit "if we were going there, we wouldn't start here" - safety critical systems design and testing is totally different to non-safety-critical*, and costs a lot more.
* In a simple sense, if you have safety-critical software you test EVERY possible path through the code, for business applications you usually don't as if there's a bug reported you just fix it later. PIS is really just a business application.
Thank you for your detailed comments @PupCuff.1) Typically the desire for full automation is driven by the general public, or politicians. Yes, they are correct in saying that a fully automated system would increase efficiency, what they do not say (conveniently) is who is going to pay for it. The cost to implement national full automation, I would say astronomical, but perhaps 'astronomical to the power of astronomical' would be more appropriate. There is a drive towards 'digital railway' which isn't automation as such but sees more effective use of digital technology to automate processes that otherwise would require significant resource (eg monitoring equipment to identify maintenance requirements, planning, and backend admin work etc). This is a more effective and affordable compromise. Often, the desire for full automation is for ideological reasons (eg reducing union bargaining power) than actual benefit.
Quite an interesting take on your second point. Suppose we were dealing with a fully autonomous railway (e.g. parts of the Paris Metro); should an incident arise, who would be held responsible in your opinion on said incident occurring? Would you agree that the undertaking of such a responsibility could be accomplished through a change in skills which you have stated? If so, what skills (type of skills) would you suggest?2) From my perspective, currently if a train driver is involved in an operational incident, following the event an investigation will take place to look at the circumstances which surrounded it, with a particular focus on human factors. Putting decisions in the hands of a computer will remove the risk of the human element of operating a railway, but will increase the risk that the computer will make a poor decision based on it's inputs and programming. In such cases where the theoretical automated train is involved in a train accident or incident, my role in the investigation ceases to be about understanding human factors, but more about understanding the technology 'under the hood' and what combination of inputs and system decisions led to the system making the decision that it did. Although it might sound like a simple transition, and it's one that I myself would be comfortable with, I have colleagues who struggle when they are asked to open a spreadsheet in Excel. This therefore may change the skills required of operational railway managers.
Really pleased to hear people talking about the PIS, ASDO, DAS and other automated systems found on trains! One of the points that I am reading is to do with 'routinisation' - while you mention that these systems take some of the workload off of the driver, do you think these innovations run the risk of no longer requiring the driver to have certain skills related to these points? Fault-finding is definitely my key interest, as train drivers on older rolling stock would have required the engineering and technological know-how to identify faults, whereas nowadays the plethora of sensors on newer rolling stock allow for faults to be detected and notified to the driver via the TMS (in addition to the depot now automatically being made aware via GSM-R, as is the case with the class 700, 707, 717). Moreover, would you agree that, while DAS can help advise drivers manage their performance, these systems can somewhat turn the job into more of a repetitive exercise (i.e., know exactly how much to accelerate/decelerate at xxx location)?3) Again I'm not a train driver, but drivers on our more modern trains now have access to systems like DAS, which helps them manage their braking and acceleration performance amongst other things, onboard computers which as well as looking after the train's customer information systems also look after things like fault finding, assisting them with timely resolution of problems rather than spending an excessive amount of time undertaking manual diagnostic procedures.
Your last sentence on this point was quite intriguing. Do you think there is need for much greater collaboration between the union(s) and the industry and that the ideological motives of those in the making of policy are making matters more difficult than they ought to be?4) If you don't adapt, you get left behind, is my view on that. Employers (as a collective) need to constantly review the changing requirements of the driver role and adjust recruitment processes accordingly (if we do get to a situation where the train mostly drives itself with a human simply supervising, do we need to have recruitment processes which assess crossing out dots and whatnot, or is something more akin to the hazard perception test when you get your car licence more appropriate?). Trade unions are a vital tool in the industry, so much good work as a result of union/management co-operation comes out and it's a shame that their role gets overshadowed by the occasional strike. But a minority of trade union folk at individual company level really ought to take a good long look at the direction of the industry and present their case for how they see things going accordingly - for instance, there are staff out there on the railway who are given a mobile phone, with access to relevant industry IT systems, telephone, email etc but because the union has not given agreement, it means they are not supposed to use it. Most staff, luckily, are sensible enough to know when they are being asked to cut their nose off to spite their face and appreciate the benefits to their role that technology can bring, going against union advice. Equally though, the sooner some companies within the industry ditch the management who are only in the job for the sense of satisfaction they get from causing a barmy with the unions, the better - there's circumstances I'm sure many of my colleagues will have seen where neither side comes out smelling of roses.
In a nutshell, do you think you could describe upskilling as the one-size-fits-all method that would help continue evolution in the railway while also improving the occupational capacities of those in the industry?Managing training needs is always the responsibility of the company and companies need to dedicate sufficient resource to updating the training requirements regularly. This should be captured through existing validation of change processes, does it always? Probably not. Automation will remove some skills, that's virtually inevitable, however in my view is that we mustn't automate and then put full faith in said automation - we need to then train people (whether that's recruiting new specialist staff or upskilling our existing workforce) to do the jobs which the automation will require, such as additional monitoring or maintenance requirements.
Oh no your point is very relevant to what I am doing my research on @Mattydo - thank you very much!Mattydo said:
Just quickly on your final point I think it’s more that I wonder if engineering out the human should come after “upskilling” society so that it doesn’t result in leaving an ever wider wage gap. But I’m not sure if that would be overly relevant to your studies. As I said, more a political point of view.
Funny you mention the German and French systems... If I remember correctly it was a disagreement between the two in the design of the A380 that resulted in it's release being pushed back (the cables weren't long enough for items such as the camera on the vertical stabiliser).Mattydo said:
I can see ATO having a purpose in increasing capacity such is on the tube or through the Thameslink core as is planned. On the mainline is it worth the investment? I’m not sure I know enough about the technology involved or the cost to implement to say whether doing so would be beneficial or ultimately political really.
As to automation creating gaps in skills, I’ve spent 12 years in the flight deck of various Airbus including the A380 and it is an aircraft designed by engineers with less pilot input. I did my 380 type rating with Airbus and the trainers would often exclaim “engineering concept” for any new system they didn’t approve of. (That or claim it must be one of the German systems as the French ones were all much better lol).
Really pleased to see you mention the situation that pilots find themselves in when Airbus's Flight Envelope Protection (FEP) goes bye-bye. Indeed, if I remember correctly the pilots were made aware through the ECAM that the PITOT tubes had frozen and the plane degraded out of Normal Law. Yet, despite being made aware by the ECAM, along with the 50+ stall alarms, this muscle memory really was thrown out the window. Do you think that the reason the FO stalled the aircraft was not only because of the false belief that an Airbus can't be stalled, but also because the pilot(s) had become so reliant on the FEP that they thought a time would never come where they would need to use their training on different Airbus laws of flight?Mattydo said:
This means that a lot of the systems are built so that even in a “manual” flight mode, the computer systems are preventing erroneous pilot input (Airbus refers to this as “Normal” control law).
The trouble with this sophisticated automation is its limitations. The moment a failure is introduced the computers become less capable of controlling the aircraft. Multiple failures result in degradation of control law protection to either alternate or direct laws. In direct law the Airbus is just a big cessna and you can stall it, over speed it or exceed load tolerances without the computers stopping you. This happened to Air France 447 when unreliable air speed indications resulted in the aircraft “giving up” and wishing the pilot good luck. The second officer then proceeded to stall the aircraft into the southern Atlantic. His doing so was not entirely his fault. He was trained to believe that you could not stall a modern Airbus and in the dead of night, in a thunderstorm with erroneous airspeed indications the computer clicked into direct law and simultaneously disengaged the autopilot. He believed that alpha floor protection was still in place and so rather than respond to what he was seeing like a pilot should, he acted as if the automation would save him, not realising that there was none left. He had lost the muscle memory that every new pilot should be engrained with in a stall. “push, roll, power, recover”.
Great to see someone bring up the GOO model used on the DLR @Bletchleyite.That of course makes a massive difference - in almost* every case on the railway, the correct answer in the event of any operating emergency is to throw the brake into emergency and worry about it afterwards. You'd also be unlikely to have totally unstaffed trains, so you likely do have a "squishy thing" there to intervene if needs be - it's the DLR model I'd expect to see (i.e. what I've referred to elsewhere on the Forum as "Guard-Only Operation"), not a totally unstaffed one. Even on roads there are considerable numbers of cases where stopping dead is dangerous - on railways there are almost none.
We are fortunate to have a well established case study for a "guard-only operated" automated railway - the DLR.
* There is the issue of stopping on viaducts etc, but the guard could drive forward manually on sight and at walking pace to clear that. The only other one I can think of is a runaway (e.g. the recent Caledonian Sleeper incident where I seem to recall a train in front was asked to run fast so the errant CS didn't catch it up) but it's incredibly rare.
Great to see someone bring up the GOO model used on the DLR @Bletchleyite.
Given that the New Tube for London (NTfL) trains are soon to arrive on certain LU lines, which apparently may not have a driver's cab, do you think that the solution for current Tube Drivers is to be transitioned from the cab into the saloon and become train captains?
@Horizon22 I'm so glad you mentioned how drivers were originally hired based on their technical and mechanical skills. The picture that I've placed on here is the TBC of the now-retired C-Stock trains formerly used on the sub-surface LU lines. Would you agree that tacit skills and training were vital in those days for someone to even be able to understand how such trains even worked?Well I'm not a driver, although I work around them, but yes I suppose so. I know there's added frustration by many "this poxy system" etc etc. Some of the younger drivers appear (completely anecodtal of course) seem to have less of these grievances. Perhaps they've grown up in a world of IT / video games, so it feels more natural to them. Drivers were originally hired for their technical and even mechanical experience and expertise as well as seeming more subjectively as a "good fit". Thus a certain type of person was often hired (many ex-military and police roles and those who worked up through the ranks). It's certainly more varied now with NTS (non-technical skills) and customer experience being additional values that are seen as important and - I would guess - some computational skills.
I know some signallers feel the same in modern ROCs "I don't want to be playing video games in an office" haha.