What this chart doesn't show you is how much more expensive it is to advance the grade of automation. Systems like the Docklands Railway are only safe because all their tracks are grade separated with exclusive track. They still need a person on board to be sure that nobody is stuck in the doors or wedged between the train and platform before the train starts moving.
"Have computers control everything" is four words to say, but it's millions or billions more in money, and adds years to the projects. London's Crossrail, for example, was years late due to the complexity of switching between 4 separate signalling systems (ETCS on the Heathrow branch, AWS on the GWML, TPWS on the GEML, and CBTC in their tunnel and exclusive track.)
We're still using TPWS on the ECML, for example, because:
> TPWS was developed by British Rail and its successor Railtrack, following a determination in 1994 that British Rail's Automatic Train Protection system was not economical – ATP would cost £600 million (equivalent to £1.2 billion in 2024) to implement, compared to value in lives saved: £3 – £4 million (£6 – 8 million in 2024), per life saved, which was estimated to be 2.9 per year.
TPWS is where a little light comes on saying it would be a good idea to brake now because you passed a signal at danger. Only very fancy versions of TPWS will actually apply the brakes for you.
> Both crashes would have been prevented by an operational automatic train protection (ATP) system, wider fitting of which had been rejected on cost grounds. The crash severely damaged public confidence in the management and regulation of safety of Britain's privatised railway system.
But even then, an inquiry asked to look specifically into ATP determined we were right to reject it, it was just Too. Fucking. Expensive.
> Since both the Paddington and Southall crashes had reopened public debate on ATP, a separate joint inquiry considering the issue in the light of both crashes was also held in 2000; it confirmed the rejection of ATP and the mandatory adoption of a cheaper and less effective system, but noted a mismatch between public opinion and cost-benefit analysis.
> TPWS is where a little light comes on saying it would be a good idea to brake now because you passed a signal at danger. Only very fancy versions of TPWS will actually apply the brakes for you.
You're confusing AWS and TPWS – AWS is the older system (based on pre-WW II technology but only deployed nationwide after WW II) which sort of roughly fits your first sentence, while TPWS will actually stop you if you're trying to pass a signal at danger (with the caveat that most automatic signals on the plain line aren't fitted on the basis that a train will relatively rarely actually have to stop at one – statistically that's absolutely sensible if you only have a limited amount of money and want to cover the higher risk signals first, but unfortunately the recent Elstow train crash did manage to hit that particular gap in the system…).
> But even then, an inquiry asked to look specifically into ATP determined we were right to reject it, it was just Too. Fucking. Expensive.
But you did eventually get TPWS which does deliver a quite large fraction of the safety benefits of ATP at a significantly smaller fraction of the costs.
> "Have computers control everything" is four words to say, but it's millions or billions more in money, and adds years to the projects.
The delays, the reduced amount of infrastructure built, and passing on of those costs in ticket prices causes people to use other forms of transport that are less safe thereby causing more casualties than a slightly less safe rail system would.
> They still need a person on board to be sure that nobody is stuck in the doors or wedged between the train and platform before the train starts moving.
Why do they need someone?
Skytain in Vancouver is fully automated and driverless with no attendant on board. Grade separated, doors automatically close, and there are sensors that trigger if someone enters the guideway.
That's a "nobody's dead yet" scenario. Safety standards require full platform doors at all stations for attendant-free operation, i.e. the platform has wall-to-ceiling barriers, the train arrives perfectly aligned (which means zero variation in train configuration, which rules out almost all mainline stations that accept different operators services), and the platform doors open and close in sync with the train doors.
Anything less is begging to get someone wedged in there as the train drives off, even if it's just a limb or a trailing bag strap.
> In June 2023, the operator of the Vancouver SkyTrain, TransLink announced a feasibility study into installing platform screen doors on the Expo and Millennium lines. Such installation was previously deemed infeasible, due to SkyTrain's diverse fleet and different door positions. However, with the acquisition of the Alstom Mark V trains, which will replace the ageing Mark I, the door positions allow for a feasibility study to proceed.
EDIT: I found some SkyTrain stats for you. 2016-2025, 1 fatal train-passenger interaction on the platform, 36 fatal in the guideway. Most were intentional suicide attempts.
I am aware, iirc the rate doubled after covid with most being self harm which is why the platform door discussion has started again. Primarily to prevent suicides and the resulting system shutdowns as the number of unintentional major injuries or deaths has been pretty low over its 40 years of no drivers or attendants and being the 4th busiest in canada/usa. The two new lines currently under construction right now will also not have platform doors.
So despite being fully automated is comparable with most other major metros/train systems that have a driver/attendant with the only metro's performing significantly better being the ones with platform doors - hence my comment of asking "why is an attendant needed for automation?" when it doesn't significantly improve safety in the same way doors would.
in the end we still may not end up with platform doors even thou the study can proceed now as the cost will be astronomical and always with safety systems there might not be the appetite for it.
Skytrain is from the 1990s. However elevators have been fully automated since before I can remember (1960s is my guess, but that is a guess if you care to look it up I'll stand corrected)
[–] FeteCommuniste 1 day ago
[–] paxy 1 day ago
[–] bluGill 1 day ago
[–] lebuffon 1 day ago
I would think that the software is the weakest link in the chain.
[–] amiga386 1 day ago
What this chart doesn't show you is how much more expensive it is to advance the grade of automation. Systems like the Docklands Railway are only safe because all their tracks are grade separated with exclusive track. They still need a person on board to be sure that nobody is stuck in the doors or wedged between the train and platform before the train starts moving.
"Have computers control everything" is four words to say, but it's millions or billions more in money, and adds years to the projects. London's Crossrail, for example, was years late due to the complexity of switching between 4 separate signalling systems (ETCS on the Heathrow branch, AWS on the GWML, TPWS on the GEML, and CBTC in their tunnel and exclusive track.)
We're still using TPWS on the ECML, for example, because:
https://en.wikipedia.org/wiki/Train_Protection_&_Warning_Sys...
> TPWS was developed by British Rail and its successor Railtrack, following a determination in 1994 that British Rail's Automatic Train Protection system was not economical – ATP would cost £600 million (equivalent to £1.2 billion in 2024) to implement, compared to value in lives saved: £3 – £4 million (£6 – 8 million in 2024), per life saved, which was estimated to be 2.9 per year.
TPWS is where a little light comes on saying it would be a good idea to brake now because you passed a signal at danger. Only very fancy versions of TPWS will actually apply the brakes for you.
And then we had the Southall and Ladbroke Grove crashes a few years later: https://en.wikipedia.org/wiki/Ladbroke_Grove_rail_crash
> Both crashes would have been prevented by an operational automatic train protection (ATP) system, wider fitting of which had been rejected on cost grounds. The crash severely damaged public confidence in the management and regulation of safety of Britain's privatised railway system.
But even then, an inquiry asked to look specifically into ATP determined we were right to reject it, it was just Too. Fucking. Expensive.
> Since both the Paddington and Southall crashes had reopened public debate on ATP, a separate joint inquiry considering the issue in the light of both crashes was also held in 2000; it confirmed the rejection of ATP and the mandatory adoption of a cheaper and less effective system, but noted a mismatch between public opinion and cost-benefit analysis.
[–] iggldiggl 11 hours ago
You're confusing AWS and TPWS – AWS is the older system (based on pre-WW II technology but only deployed nationwide after WW II) which sort of roughly fits your first sentence, while TPWS will actually stop you if you're trying to pass a signal at danger (with the caveat that most automatic signals on the plain line aren't fitted on the basis that a train will relatively rarely actually have to stop at one – statistically that's absolutely sensible if you only have a limited amount of money and want to cover the higher risk signals first, but unfortunately the recent Elstow train crash did manage to hit that particular gap in the system…).
> But even then, an inquiry asked to look specifically into ATP determined we were right to reject it, it was just Too. Fucking. Expensive.
But you did eventually get TPWS which does deliver a quite large fraction of the safety benefits of ATP at a significantly smaller fraction of the costs.
[–] graemep 1 day ago
The delays, the reduced amount of infrastructure built, and passing on of those costs in ticket prices causes people to use other forms of transport that are less safe thereby causing more casualties than a slightly less safe rail system would.
[–] katbyte 1 day ago
Why do they need someone?
Skytain in Vancouver is fully automated and driverless with no attendant on board. Grade separated, doors automatically close, and there are sensors that trigger if someone enters the guideway.
[–] amiga386 1 day ago
Anything less is begging to get someone wedged in there as the train drives off, even if it's just a limb or a trailing bag strap.
https://en.wikipedia.org/wiki/Platform_screen_doors
> In June 2023, the operator of the Vancouver SkyTrain, TransLink announced a feasibility study into installing platform screen doors on the Expo and Millennium lines. Such installation was previously deemed infeasible, due to SkyTrain's diverse fleet and different door positions. However, with the acquisition of the Alstom Mark V trains, which will replace the ageing Mark I, the door positions allow for a feasibility study to proceed.
EDIT: I found some SkyTrain stats for you. 2016-2025, 1 fatal train-passenger interaction on the platform, 36 fatal in the guideway. Most were intentional suicide attempts.
https://www.technicalsafetybc.ca/technologies/railways/skytr...
[–] katbyte 1 day ago
So despite being fully automated is comparable with most other major metros/train systems that have a driver/attendant with the only metro's performing significantly better being the ones with platform doors - hence my comment of asking "why is an attendant needed for automation?" when it doesn't significantly improve safety in the same way doors would.
in the end we still may not end up with platform doors even thou the study can proceed now as the cost will be astronomical and always with safety systems there might not be the appetite for it.
[–] bluGill 1 day ago