Level boarding versus driverless trains

[father_jack]

Both a pipe dream- but one is way more deserving/realistic/achievable/moral than the other.

But the other will always get more newspaper column or Internet bandwidth wastage.......

Sad really.

Any opinions ?

 

[HSTEd]

Both a pipe dream

I don't see why either are a pipe dream.
The technology for both exists today.

Indeed both are in service at various places in the world.
The UK railway may chose not to adopt one or both, but that is not to say they are implausible.

A bigger question is whether the UK railway is truly organisationally capable of the modernisation required for either.

 

[Nottingham59]

The technology for driverless trains is getting close. Controlling an autonomous taxi is very much more computationally complex than driving a train, and yet that happens every day now in San Francisco and Phoenix and many cities in China.

What, if any, aspects of route knowledge or visual perception or driving decision-making is now beyond the scope of autonomous vehicle technology? I can't think of any. Others may know better of course.

 

[edwin_m]

What, if any, aspects of route knowledge or visual perception or driving decision-making is now beyond the scope of autonomous vehicle technology? I can't think of any. Others may know better of course.

Responding to an emergency. Or at least the perception that passengers will be more at risk from an incident if there is no driver present.

 

[stevieinselby]

The technology for driverless trains is getting close. Controlling an autonomous taxi is very much more computationally complex than driving a train, and yet that happens every day now in San Francisco and Phoenix and many cities in China.

What, if any, aspects of route knowledge or visual perception or driving decision-making is now beyond the scope of autonomous vehicle technology? I can't think of any. Others may know better of course.

An autonomous road vehicle will always have alternative route options. A train won't.
If a robotaxi loses communication with the mother ship and shuts down, or there's a blip that prevents it from using a particular route, it's a nuisance for the one passenger in it, but they can get out and go and find another robotaxi, or walk. Other vehicles can go around it. No big deal.
If a robotrain loses communication with the mother ship and shuts down, or there's a blip that prevents it from using a particular route, that could be several hundred passengers stranded and the line blocked. Yes, I know that can happen now with a breakdown, but you're significantly increasing the risk of it.

So far, trials of autonomous control in low adhesion rail conditions have been poor. The onboard control usually errs on the side of caution – which is better than the alternative! – but leads to trains going far slower than they would with a competent human driver. This isn't a problem on the Underground or Thameslink core, as adhesion conditions don't change much when you're underground, but can be a big problem on lines above ground in autumn and winter.

 

[Bletchleyite]

So far, trials of autonomous control in low adhesion rail conditions have been poor. The onboard control usually errs on the side of caution – which is better than the alternative! – but leads to trains going far slower than they would with a competent human driver. This isn't a problem on the Underground or Thameslink core, as adhesion conditions don't change much when you're underground, but can be a big problem on lines above ground in autumn and winter.

Such as the DLR?

The DLR gives a good concept for "guard only operation" where the guard is trained to drive if necessary but mostly doesn't.

 

[HSTEd]

Responding to an emergency. Or at least the perception that passengers will be more at risk from an incident if there is no driver present.

Most trains will end up with some form of on board crew anyway, especially with the continuing desire for wide deployment of catering facilities.
Since the question is merely driverless and not uncrewed, I would assume in this situation that someone will be available for response anyway.

Indeed, by moving the crew member from the cab into (say) a buffet, they are less likely to be incapacitated by various relatively high likelihood accidents.
ie. hitting something at a level crossing.

Driverless operation also has various operational advantages, especially in disruption, as trains can be turned as soon as passengers have been disembarked/reembarked.
More generally, reversing at a station need not take any more time than a regular station stop.

Whilst on board safety staff will require some form of route training, prima facie it is difficult to imagine it being as extensive as "remember every signal and speed restriction on the route". At which point it may be easier to roster crew efficiently.

 

[renegademaster]

trains can be turned as soon as passengers have been disembarked/reembarked

While not with passengers, the Elizabeth line does this at Paddington which is quite cool.

 

[Nottingham59]

Just to illustrate how good AV technology is getting, this is a video of an autonomous car driving itself around central London.

 

[Sorcerer]

I think level boarding is vastly more likely than driverless trains on the network. All you would need to do for that to be widespread is legislate that all new rolling stock must be built with step-free access and within a generation you'll have level boarding on practically the entire network. Stadler have already proven it's possible on the UK network so it's really just a question of political will. Driverless trains though? Considerably more difficult and expensive to convert the entire network to enable such operation.

 

[zwk500]

Driverless trains requires the rollout of ETCS Level 2, CBTC or other cab-communication driving systems to actually move the train. Using Autonomous Technology to read signals would be unlikely to be considered sufficiently fail-safe in the event of misreading signals in busy or complex areas (or too cautious to the extent of confusion and blowing performance to smithereens). Given how much of the network still runs on train staff/token/Absolute Block I can't see cab-signalling being universal any time soon - although I can seen increasing automation on main lines accelerating as ETCS L2 is rolled out. In the long run maybe, but not for a lifetime, IMHO.

Level boarding is, in contrast, relatively straightforward to achieve - it requires the alignment of industry, politics and funding, but it's still a much more viable project to align standards for platform and train floor height and then carry out a rolling programme of track/platform alignment. There are a lot of platforms to touch, but it would not be impossible for the vast majority of platforms at or about 915mm to be rebuilt to 950mm and for all new trains to be ordered at 960mm floor height with gap fillers. 1,100mm stock would be unaffected by the lifting of the platform, so a crossover period could be relatively easily accommodated.
In my mind, such an programme would also take the opportunity to rebuild platform structures and track spacings to UIC-compliant clearances with a readily adjustable top surface should UIC-GB ever be implemented. But I think that's further away than driverless trains!

 

[swt_passenger]

Is the title meant to imply that we need level boarding to even think about automatic operation?

I’m not really sure about the use of the word ‘versus’, the two matters aren’t in competition surely?

 

[Bletchleyite]

Is the title meant to imply that we need level boarding to even think about automatic operation?

I'm not sure we do if "guard only operation" was to be the method.

I’m not really sure about the use of the word ‘versus’, the two matters aren’t in competition surely?

Agree, they're not.

 

[zwk500]

Is the title meant to imply that we need level boarding to even think about automatic operation?

I’m not really sure about the use of the word ‘versus’, the two matters aren’t in competition surely?

I read it as an 'either/or' hypothetical decision. As in limited resources to do both, so choose one to implement.

 

[brad465]

I think level boarding is vastly more likely than driverless trains on the network. All you would need to do for that to be widespread is legislate that all new rolling stock must be built with step-free access and within a generation you'll have level boarding on practically the entire network. Stadler have already proven it's possible on the UK network so it's really just a question of political will. Driverless trains though? Considerably more difficult and expensive to convert the entire network to enable such operation.

This should be first and foremost the main talking point with driverless trains. The technology can be there, but if the money is not, both in terms of physical installation cost and the cost via disruption during installation, it won't happen. It should come as no surprise that driverless trains only exist in the UK on segregated networks (like the DLR and underground lines), or lines that are new build and/or dominated by a single train class/type (Elizabeth line core and Thameslink core).

 

[Sorcerer]

It should come as no surprise that driverless trains only exist in the UK on segregated networks (like the DLR and underground lines), or lines that are new build and/or dominated by a single train class/type (Elizabeth line core and Thameslink core).

Or quite possibly even on HS2 once completed, although much like the Elizabeth Line and Thameslink, this will only be limited to the core services. On the mainline it will simply be a logistical nightmare to replace everything, yet despite that many seem to think that these small segregated sections of railway somehow set a precedent for full automation on the entire network. Just go to any social media article reporting on strikes and you are guaranteed to find several comments saying "will be automated soon anyway".

 

[father_jack]

Is the title meant to imply that we need level boarding to even think about automatic operation?

I’m not really sure about the use of the word ‘versus’, the two matters aren’t in competition surely?

I meant in the context of the frequent media subject of driverless trains and the current passenger assistance matters. Which is more realistically able to and/or likely to, happen ?

 

[BrianW]

I imagine that driverless trains will be shown to be safer than trains with fallible drivers.
Similarly, I imagine they will be shown to be cheaper to run, with a multi-skilled train/ customer-facing manager.
Level boarding MUST come as a priority- it's well overdue.
Electrification/battery/hydrogen must also happen.

Rolling programmes are the key, with skilled teams, starting with a couple of pilot schemes on 'low-hanging fruit' more self-contained parts of the network, building up and out ...
The possibility of political commitment is there

 

[HSTEd]

Or quite possibly even on HS2 once completed, although much like the Elizabeth Line and Thameslink, this will only be limited to the core services. On the mainline it will simply be a logistical nightmare to replace everything, yet despite that many seem to think that these small segregated sections of railway somehow set a precedent for full automation on the entire network. Just go to any social media article reporting on strikes and you are guaranteed to find several comments saying "will be automated soon anyway".

The fundamental technologies developed for the UK ATO installations are applicable to anywhere that is signalled with ETCS.
Conceptually, there is nothing stopping them being installed as part of every ETCS scheme.

The ETCS programme is not going well, but it is (apparently) still progressing.
Meanwhile ATO is increasingly common abroad.

 

[vuzzeho]

If I had to choose one, I don't think we need driverless trains right now. As a long term goal, sure, that's fine. Level boarding, however, I view as more essential. Allowing for better access to the railway, especially for disabled people, should be a priority.

 

[Trainbike46]

The technology for driverless trains is getting close. Controlling an autonomous taxi is very much more computationally complex than driving a train, and yet that happens every day now in San Francisco and Phoenix and many cities in China.

Driverless trains exist today, both in London (DLR), and in other countries (e.g. Copenhagen metro).

Level boarding, once again, exists both in the UK (Greater Anglia, TfW, and Merseyrail all offer it at standard UK platform height), and in many other countries (e.g. the Netherlands, all local/regional trains and an increasing number of IC trains offer level boarding from the platform).

obviously driverless trains require more work, especially a signalling system that supports it.

 

[Nick Ashwell]

The fundamental technologies developed for the UK ATO installations are applicable to anywhere that is signalled with ETCS.
Conceptually, there is nothing stopping them being installed as part of every ETCS scheme.

The ETCS programme is not going well, but it is (apparently) still progressing.
Meanwhile ATO is increasingly common abroad.

Of the two heavy rail implementations of ATO I can think of, Crossrail and Thameslink, only 50% is ETCS

 

[zwk500]

Of the two heavy rail implementations of ATO I can think of, Crossrail and Thameslink, only 50% is ETCS

I think it's been mentioned on other threads that HS2 will be ATO on ETCS L2 (with driver in the cab), but I could be wrong.

Is Drayton Park to Moorgate planned for ATO when signals-away ETCS comes in? I have a feeling it's not, but am not confident either way.

 

[Vexed]

Of the two heavy rail implementations of ATO I can think of, Crossrail and Thameslink, only 50% is ETCS

The only reason Crossrail chose CBTC is because, at that point, ATO over ETCS wasn't proven technology, especially with the accuracy needed for platform edge doors. While XR did a lot of new things I think this was just one too many to try at once. TfL also had prior experience of ATO over CBTC with some of the Underground lines.

ATO over ETCS was first used in passenger service (anywhere in the world) in March 2018 so with their initial planned opening date of December 2018, it did make sense at the time to use proven CBTC. In hindsight with the actual opening being a lot later and ATO over ETCS being proven this maybe wasn't the right decision. It probably would have made some of the issues they had with transition between signalling systems easier as well if they could go from 3 to 2. ETCS is definitely the future for new ATO.

 

[boiledbeans2]

If a robotrain loses communication with the mother ship and shuts down, or there's a blip that prevents it from using a particular route, that could be several hundred passengers stranded and the line blocked. Yes, I know that can happen now with a breakdown, but you're significantly increasing the risk of it.

There is a lot of ongoing research (though not from the UK) in automating trains. In case the automated train fails, one solution to your above problem is to control the trains remotely. Consider the Shift2Rail TAURO (Technologies for the AUtonomous Rail Operation) H2020 project.

This paper was presented in the TRA 2022 conference "Remote driving and command of trains: The Shift2Rail approach.", co-written by CAF, Thales and Alstom:
https://www.sciencedirect.com/science/article/pii/S235214652300844X (paper is open-access)

Quotes from the paper:

Driving and commanding rail vehicles remotely can be applied to a wide variety of use cases, and nowadays rail operating companies and rolling stock maintainers are demanding solutions to gain remote control of their assets, in order to increase flexibility and resilience while reducing costs. The members of Shift2Rail within the TAURO project have identified three relevant use cases: Remote operation under ATO GoA4 over ETCS, remote freight shunting yard operation and remote operation of tramways in depot.

Remote driving and command of rail vehicles is a function which has several fields of application. First, accessing remotely the vehicle will be required when the autonomous or automatic operation is disturbed by any operational problem or faulty system and thus entering a degraded mode. For instance, if the ATO function is not available, rescuing (i.e. driving) the vehicle remotely until the next safe stop will be an expected mitigation to be provided under ATO GoA4 operation. The ability of accessing the TCMS remotely, not only to monitor (e.g. remote HMI) but also to diagnose and command some actions to solve issues is a key aspect to maximize operational availability of autonomous and automatic rail vehicles. Second, there is an increasing demand for driving rail vehicles remotel under specific operational scenarios, like depots or shunting yards, as an intermediate step before fully automatize such environments. From a virtual driving desk placed at the control center, the workshop or the office, or even from a portable device, such as a tablet, the user will drive the vehicle through the depot or the yard as a kind of train-drone.

The obstacle to driverless trains isn't technology. The issue is money and unions.

 

[zwk500]

There is a lot of ongoing research (though not from the UK) in automating trains. In case the automated train fails, one solution to your above problem is to control the trains remotely. Consider the Shift2Rail TAURO (Technologies for the AUtonomous Rail Operation) H2020 project.

This paper was presented in the TRA 2022 conference "Remote driving and command of trains: The Shift2Rail approach.", co-written by CAF, Thales and Alstom:
https://www.sciencedirect.com/science/article/pii/S235214652300844X (paper is open-access)

Remote operation has a place, but it still leaves important gaps that become more relevant for services outside of a metro area. DLR style GoA3 Guard-Only operation with training to move the train to safety if required covers far more gaps in that regard (e.g. if communication between the RBC/Control and Train is what's failed, or the on-train power control systems, or if it's something like a circuit breaker that needs resetting).
There's also the issue that if there's a major failure presumably the controllers can only 'drive' one train at a time? Fine if it's a single train that needs to be dealt with, but somewhat more problematic if it's a major 4-track main line into London.

Remote control of trains is best suited to depots and shunting IMO (which I note is how the paper starts it's focus), and shouldn't be considered a primary failure mode. Being able to do runrounds single-manned using a remote control loco on the mainline would be an example of where that sort of technology would be very useful.

Notably from the paper quoted (my bold):

However, a train driver performs many actions besides accelerating and braking and its intervention is crucial in cases where the availability of the vehicle is compromised, requiring specific actions such as manually locking doors or resetting systems. Therefore, a comprehensive analysis is still required to lead to a standardized solution, and in line with the current developments on architecture and interfaces and on regular standardized communications (FRMCS [5]), IEC 61375-2-6). The use case of degraded operational modes could imply the unavailability of some functions (e.g. due to faulty devices) and hence the remote driving could be impossible.

 

[Topological]

It would be good to move to a point where branch lines that operate on a single-train basis can be autonomous. Though such lines may not be high on the ETCS list, they would seem to be straightforward enough to switch over. That would help such lines run when there are shortages of drivers and cutting that service is seen as the path of least resistance.

Longer term services like Manchester - Southport could be run as Wigan - Southport autonomously (for example)?

I do think level boarding is a good idea though and should be rolled out as much as possible.

 

[Meerkat]

I thought they should have used the Isle of Wight line as an ATO pilot, with a guard/emergency driver in a DLR style.

 

[Bletchleyite]

I thought they should have used the Isle of Wight line as an ATO pilot, with a guard/emergency driver in a DLR style.

When that does eventually happen my money is on Stourbridge.

 

[Nottingham59]

When that does eventually happen my money is on Stourbridge.

I'd propose the Barton on Humber branch. Grimsby to Barton trains connect with the bus service over Humber Bridge to Hull. As soon as that bus service can be run with autonomous busses, like the CAV Forth trials in Edinburgh, then extend the bus route into Grimsby and close the railway, which would no longer serve any useful purpose.