Very interesting thanks - I'm curious if how often gates are direction switched is a factor too - if they switch them back and forward too often that could decrease the efficiency - it's also very annoying if the gate you're waiting to go through suddenly turns from green to red and you have to move across to another one, screwing up the queueing algorithms!
The docs mention that the recalculation was done once every minute - though you're right about this being annoying if it changes right in front of you. I also wonder about what happens where there are (movable) physical barriers dividing entries and exits, like at Stratford – you'll still need humans to move them. Though I guess this could be something they just don't bother to barrier off in the future.
Just to really emphasize how important capacity and frequency are:
I got the bus to the hospital today (routine blood tests). There's a direct bus (the 53), that runs from outside my front door to literally outside the hospital, and it only takes 10 minutes. Or I can get any of an 86/253/256 and change onto a 41/42/43/141/142/143; this takes about 15 minutes as I have to cross the road between bus stops when changing bus.
The 53 runs every half-hour, so I generally just change buses, as the combined service on both routes is more frequent than every 5 minutes. A single return journey is £4, a day pass is £5, so it's only a quid extra and I can get any bus I like.
Thinking to my usage of the Tube, the gatelines are clearly one constraint on the capacity, but I'd have thought that train crowding was more generally the problem: if you can't get people onto the train, it doesn't matter how efficient the station is at getting them to/from the platform.
I suspect that TfL are being conservative because removing a bottleneck in one place (the relatively safe ticket hall) could create a new one elsewhere (a dangerous tube platform, or a narrow pavement with a bus stop). As you rightly point out in your introduction (but fail to apply consistently!) it's capacity all the way down.
It is ever the case that you cannot totally eliminate bottlenecks in a system, only move them around and/or mitigate them until either a goal is achieved or the overall capacity is maximised to support a specific scenario.
The case for HS2 was made first on journey time, capacity was grasped afterwards when it became clear that the benefit in shorter journey times alone would not support the cost of the lavish specification that engineers had been allowed to define.
The West Coast Main Line is not full throughout its length, and most of its trains are not freight but passenger. Taking a sample hour between 1200 and 1300 today, 30 passenger trains and 1 freight train passed through Cheddington. A four track route, properly planned, with junction conflicts avoided by grade separation at either end (in this case, Euston and Rugby), and with margins to assure performance, could feasibly carry at least third more; and many of the trains which did run were shorter than their maximum length.
The WCML has five stretches where managing capacity and performance are tight,, between Euston and Tring, between Coventry and Birmingham, between Colwich, Stafford and Norton Bridge, at Crewe station (rebuilt to match 1980s levels of demand) and on the immediate approach to Manchester Piccadilly. If investment were to be made to increase capacity, improvements over, or paralleling, these relatively short stretches would be quite sufficient.
And, as has been pointed out, increasing the peak throughput of ticket barriers does not increase the capacity of a Tube line. Escalator, platform (where those getting on and those getting off collide), and train capacity all can. On the Victoria Line, which serves Blackhorse Road, the binding constraint is the platform capacity at Kings Cross southbound in the morning peak, when a high numbersof people are leaving trains from Walthamstow and another high number are boarding them having arrived at Kings Cross on the main line.
A railway system has to be seen as a whole, not broken into unconnected parts which are not optimised together.
The impetus to build HS2 was always based on capacity, but the business case required attracting travellers from air routes, which meant that speed was important to getting the project financed.
This is well documented back to the original Greengauge 21 proposals and was explored at length by Roger Ford in Modern Railways at the time.
In terms of specification, Gareth Dennis has covered this, but essentially it's important to build for the future, not for the present. Many Victorian railways were built "on the cheap", and we're still suffering from that legacy in terms of high maintenance costs (think landslips) and congestion at level crossings for example. In the 1950s many argued that Britain's motorways were being overspecified but the foresight from that era continues to serve us well today - it was worth the extra expense.
It is partly a circular argument, as shorter journey times generate more passengers, so demand an increase in capacity. But if capacity was the prime motive, why design for 400km/h?
Much the largest part of HS2's revenue was to come from London-Birmingham, Liverpool, Manchester, Leeds, Sheffield and Nottingham/Derby, routes on which airline volumes competition is very slight.
It was that generated traffic, and the economic growth enabled by it, which made the case look attractive. But that was with disastrously underestimated costs: the construction industry thought that it had found something that the government wanted to want, and could be seduced gradually into overpaying for it.
The suggestion that infrastructure should be overspecifed to give it a long life disregards the effects of discounting and of risk: saving renewal costs a long time in the future has a very low NPV; and building for a market which may not exist in that long distant future requires the discount rate to be increased to allow for the risk.
Those who conceived 1960s motorways understood that: the roads have needed very substantial improvement to meet demand – for volume and vehicle weights – much higher than was foreseen at the time of their construction.
I just got back from Tokyo, where the Metro gates can be passed in either direction. I'm struggling to find out which other metro systems employ these gates, but it got me thinking about them. Presumably they help flow during rushes in one direction, but maybe cause problems if it's equally busy in both directions.
Interesting trial, thanks for the write up! Shame they never went further but possibly a couple of reasons:
- as Richard Gadsden mentioned above, queuing time at gates is irrelevant if enough people are getting into the station to pack the trains still. Not sure we have that information here? And this will vary greatly between stations, I assume
- the trial was at a time when people were getting very wary of personal data collection, especially in relation to “AI” systems. They might have decided the (still unclear) benefits weren’t worth that battle (yes you can do the tracking by other means but the CV approach would be the most precise/accurate I think - and assume a reasonable accuracy is required here?)
Btw the absolute time saved by auto gate switching is unclear to me - the y-axis on the graph doesn’t have units. Is an 80% time saving a matter of a few seconds? Tens of seconds? Minutes?
My friend is a TfL engineer and he was saying the HS2 line would stop the existing London to Birmingham line being at 90% capacity. Most of it is freight, not passengers, which is something most people probably don't realise. I've even made the "who wants to get to Birmingham 30 minutes faster?" joke but honestly, there's so much stuff being moved around by rail, to take pressure off the roads. Maybe we need to upgrade the canals and move things by barge again.
Blackhorse Road station happens to have a large increase of nearby high rise flats being built so capacity is definitely already an issue there. Also can the knock on effect of platform overcrowding be monitored in a similar way or is it already?
Very interesting thanks - I'm curious if how often gates are direction switched is a factor too - if they switch them back and forward too often that could decrease the efficiency - it's also very annoying if the gate you're waiting to go through suddenly turns from green to red and you have to move across to another one, screwing up the queueing algorithms!
The docs mention that the recalculation was done once every minute - though you're right about this being annoying if it changes right in front of you. I also wonder about what happens where there are (movable) physical barriers dividing entries and exits, like at Stratford – you'll still need humans to move them. Though I guess this could be something they just don't bother to barrier off in the future.
Just to really emphasize how important capacity and frequency are:
I got the bus to the hospital today (routine blood tests). There's a direct bus (the 53), that runs from outside my front door to literally outside the hospital, and it only takes 10 minutes. Or I can get any of an 86/253/256 and change onto a 41/42/43/141/142/143; this takes about 15 minutes as I have to cross the road between bus stops when changing bus.
The 53 runs every half-hour, so I generally just change buses, as the combined service on both routes is more frequent than every 5 minutes. A single return journey is £4, a day pass is £5, so it's only a quid extra and I can get any bus I like.
Thinking to my usage of the Tube, the gatelines are clearly one constraint on the capacity, but I'd have thought that train crowding was more generally the problem: if you can't get people onto the train, it doesn't matter how efficient the station is at getting them to/from the platform.
I suspect that TfL are being conservative because removing a bottleneck in one place (the relatively safe ticket hall) could create a new one elsewhere (a dangerous tube platform, or a narrow pavement with a bus stop). As you rightly point out in your introduction (but fail to apply consistently!) it's capacity all the way down.
It is ever the case that you cannot totally eliminate bottlenecks in a system, only move them around and/or mitigate them until either a goal is achieved or the overall capacity is maximised to support a specific scenario.
Some corrections are needed.
The case for HS2 was made first on journey time, capacity was grasped afterwards when it became clear that the benefit in shorter journey times alone would not support the cost of the lavish specification that engineers had been allowed to define.
The West Coast Main Line is not full throughout its length, and most of its trains are not freight but passenger. Taking a sample hour between 1200 and 1300 today, 30 passenger trains and 1 freight train passed through Cheddington. A four track route, properly planned, with junction conflicts avoided by grade separation at either end (in this case, Euston and Rugby), and with margins to assure performance, could feasibly carry at least third more; and many of the trains which did run were shorter than their maximum length.
The WCML has five stretches where managing capacity and performance are tight,, between Euston and Tring, between Coventry and Birmingham, between Colwich, Stafford and Norton Bridge, at Crewe station (rebuilt to match 1980s levels of demand) and on the immediate approach to Manchester Piccadilly. If investment were to be made to increase capacity, improvements over, or paralleling, these relatively short stretches would be quite sufficient.
And, as has been pointed out, increasing the peak throughput of ticket barriers does not increase the capacity of a Tube line. Escalator, platform (where those getting on and those getting off collide), and train capacity all can. On the Victoria Line, which serves Blackhorse Road, the binding constraint is the platform capacity at Kings Cross southbound in the morning peak, when a high numbersof people are leaving trains from Walthamstow and another high number are boarding them having arrived at Kings Cross on the main line.
A railway system has to be seen as a whole, not broken into unconnected parts which are not optimised together.
JM
The impetus to build HS2 was always based on capacity, but the business case required attracting travellers from air routes, which meant that speed was important to getting the project financed.
This is well documented back to the original Greengauge 21 proposals and was explored at length by Roger Ford in Modern Railways at the time.
In terms of specification, Gareth Dennis has covered this, but essentially it's important to build for the future, not for the present. Many Victorian railways were built "on the cheap", and we're still suffering from that legacy in terms of high maintenance costs (think landslips) and congestion at level crossings for example. In the 1950s many argued that Britain's motorways were being overspecified but the foresight from that era continues to serve us well today - it was worth the extra expense.
It is partly a circular argument, as shorter journey times generate more passengers, so demand an increase in capacity. But if capacity was the prime motive, why design for 400km/h?
Much the largest part of HS2's revenue was to come from London-Birmingham, Liverpool, Manchester, Leeds, Sheffield and Nottingham/Derby, routes on which airline volumes competition is very slight.
It was that generated traffic, and the economic growth enabled by it, which made the case look attractive. But that was with disastrously underestimated costs: the construction industry thought that it had found something that the government wanted to want, and could be seduced gradually into overpaying for it.
The suggestion that infrastructure should be overspecifed to give it a long life disregards the effects of discounting and of risk: saving renewal costs a long time in the future has a very low NPV; and building for a market which may not exist in that long distant future requires the discount rate to be increased to allow for the risk.
Those who conceived 1960s motorways understood that: the roads have needed very substantial improvement to meet demand – for volume and vehicle weights – much higher than was foreseen at the time of their construction.
JM
[Mutters about Peckham Rye's three gates, which is obviously not a TfL station and being redeveloped anyway.]
Fascinating post and great detective work through the FOI request. Worthy of being reposted on www.londonreconnections.com to reach a wider audience.
I just got back from Tokyo, where the Metro gates can be passed in either direction. I'm struggling to find out which other metro systems employ these gates, but it got me thinking about them. Presumably they help flow during rushes in one direction, but maybe cause problems if it's equally busy in both directions.
Interesting trial, thanks for the write up! Shame they never went further but possibly a couple of reasons:
- as Richard Gadsden mentioned above, queuing time at gates is irrelevant if enough people are getting into the station to pack the trains still. Not sure we have that information here? And this will vary greatly between stations, I assume
- the trial was at a time when people were getting very wary of personal data collection, especially in relation to “AI” systems. They might have decided the (still unclear) benefits weren’t worth that battle (yes you can do the tracking by other means but the CV approach would be the most precise/accurate I think - and assume a reasonable accuracy is required here?)
Btw the absolute time saved by auto gate switching is unclear to me - the y-axis on the graph doesn’t have units. Is an 80% time saving a matter of a few seconds? Tens of seconds? Minutes?
My friend is a TfL engineer and he was saying the HS2 line would stop the existing London to Birmingham line being at 90% capacity. Most of it is freight, not passengers, which is something most people probably don't realise. I've even made the "who wants to get to Birmingham 30 minutes faster?" joke but honestly, there's so much stuff being moved around by rail, to take pressure off the roads. Maybe we need to upgrade the canals and move things by barge again.
Blackhorse Road station happens to have a large increase of nearby high rise flats being built so capacity is definitely already an issue there. Also can the knock on effect of platform overcrowding be monitored in a similar way or is it already?
I suspect this is why the trial was superseded by this other, broader AI vision trial - which could conceivably monitor crowds *and* gatelines.
https://takes.jamesomalley.co.uk/p/tfls-ai-tube-station-experiment-is