Hacker News Reader

Comment by dilap | Hacker News Reader

> Turns out, strangely, that it actually reduces throughput compared to just having everyone stand

Wait, what? How can this be true?

Imagine an escalator w/ both lanes filled with people just standing. Now imagine the left lane starts walking: clearly, higher throughput.

Now, maybe you're imagining a scenario where people don't want to walk. Now everyone stands on the right, and the left lane is completely empty. OK, now we've reduced throughput by half compared to "everyone standing".

So as we consider moving between the extremes of "no walkers" and "walker saturation", there will be some break-even point of throughput. And beyond that having a walking lane will be more efficient.

(Also, besides raw throughput, having a walking lane gives you a better "fastest possible traversal time", which is also valuable, for people in a hurry.)

View on HN

My first reaction to learning this was also about the same, haha.

The issue with the "everyone starts walking" is the same reason why stationary cars are closer together than moving cars. If people walked with military trained precision you could get the same density and vastly higher throughput by walking. But for civilian strangers, it turns out two lanes of standing gives high enough density to trump the increased speed of walking, in terms of total throughput. I'm fairly certain there's even a study showing this. They tested it in london underground at least.

But yes, it's certainly better for the few walkers.

Ah, I didn't consider density, that's a good point!

Taking a look at the study you mention (linked in other comments), it seems like the escalators they studied were well to the "too few people want to walk" side of the curve to be efficient.

I wonder how close the break-even point is to the simple metric, "do you have to wait to enter on the right?"

Or thinking of the queue length for standers, clearly when the length is 0 allowing walking is more efficient (you've taken nothing away from the standers, and the walkers are more efficient). But there is some point where the stand-queue length, let's call it N, is big enough that the corresonding wait-time means overall throughput is lowered by allowing walking.

I think it would be quite reasonable to require standing in those situations.

So the rule could be something like, "stand on the right, walk on the left, unless there is a queue of more than N people waiting to stand on the right, in which case you may stand on the left." Easy right? ;-)

(I wonder if N would vary w/ escalator length? Intuitively, almost certainly. So we'd have to post signs next to each escalator w/ its individual stand-queue limit. :-)

That makes a lot of sense to me.

Transport for London experimented with this, and found higher overall throughput during rush hour with two standing lanes:

https://www.independent.co.uk/news/science/it-is-faster-to-s...

> Standing-only escalators had a "throughput" of up to 151 passengers per minute, while an escalator where commuters were still allowed to walk saw around 115.

Your thought experiment's completely right, but in real world circumstances, once an escalator is above a certain height, people rarely use the walking lane:

> But a 2002 study of escalator capacity on the Underground found that on machines such as those at Holborn, with a vertical height of 24 metres, only 40% would even contemplate it [walking up]. By encouraging their preference, TfL effectively halves the capacity of the escalator in question, and creates significantly more crowding below, slowing everyone down.

(from https://www.theguardian.com/uk-news/2016/jan/16/the-tube-at-... , which has a great graphic showing the two systems)

Transport for London should politely refrain from any statements like that, I believe. They hardly can allow people to enter the platform because there's no space to fit the crowd. There's sea of heads almost every morning in front of most popular stations' gates. Not because they have a gates issue, or any other unusual problem - they are just regulating flow that way. So, I'd kindly ask them start modernizing platforms to allow more throughput in the first place. Then let them experiment and play with statistics, by all means.

That’s in one place, which is a particular bottleneck. The general rule is that two lanes are better.

Article on the London experiment: https://www.theguardian.com/uk-news/2016/jan/16/the-tube-at-...

It's a long escalator, so few people walk (40% according to the Guardian article). And then during rush hour there's a crowd in front of the escalator.

This blog post estimates walking is 26 seconds and standing is 40 seconds: https://www.capgemini.com/gb-en/2016/04/would-standing-on-th...

The optimal policy is to have everyone walk 2 abreast as densely as possible, so this about debating two suboptimal policies. Presumably if it was actually a problem they would just build another escalator.

It sounds to me that the walk/stand system is nice because it creates a priority queue. Some people would rather wait and stand than not wait and walk. The people who are in a hurry get to walk and not wait. If the norm was "always stand", the 40% of people who walk would be much more annoyed. Even when the escalator was only at 5% capacity, there would be enough side-by-side standers that they wouldn't be able to walk.

The volume of the escalator is reduced by greater than half, and the departure rate is a function of density, so reserving half of the escalator for people who want to depart slightly faster than others (in single file, with space) apparently doesn't overtake what a single, densely packed transit could carry.

The problem is that people who want to walk are limited in reality by how quickly the slowest person ahead of them is traveling as well as conventions that make it unlikely that people would get within 1 full step of the person in front of them, reducing the lane's volume to 1/3 (the step you're on plus the step you're taking + space) of the possible occupancy of the standing lane.

So I could definitely see that assertion being true.

Standing people stand closer together than walking people. However, I think they should optimize for minimum annoyance, not maximum throughput. With a walking lane, there's lower latency for people willing to walk.

> Now, maybe you're imagining a scenario where people don't want to walk.

Agreed that this is an imagined scenario because people in a hurry will want to maximize their speed by walking/running up the escalator.

More importantly, people generally (I'm imagining) do not want to stand on the same escalator step as another person, especially if they are strangers to each other.

That means in a 2-lane standing escalator there will be breaks between people standing to the left or the right which, effectively, will turn a 2-lane standing-only escalator into a "1-lane" standing escalator.

That's actually how it should be. People should stand alternating left-right-left-... Not having one empty step in front/back of you is quite dangerous especially with packed escalator.

Reply on news.ycombinator.com