Hacker News Reader

Comment by t_mann | Hacker News Reader

This is urgently missing some discussion of queuing theory / stochastic processes. Discussing passenger flow in terms of time averages is misleading, passengers don't arrive in continuous streams. The peak load will be many times the average, which impacts power consumption, while at other times they will be running almost empty with just a handful of passengers.

View on HN

It's actually not that bad to assume a crowd is a fluid. For example it allows you to deduce where the hotspots will be when there's a panic.

Disagree. Firstly there aren’t many situations in fluid dynamics where there is a steady constraint on mass flow. Velocity just increases. (The exception is choked flow from a nozzle IIRC)

Now think about what happens when 6000 people/hr leave the station but arrive on 12 trains over the course of the hour, i.e. in surges. Each peak demand will comfortably exceed the flow rate of escalator.

You get a queue; the equivalent in fluid dynamics would be an increase in density upstream of a choked nozzle at Mach 1 but how to turn that into a relevant metric for eg queue length?

Thats also the reason - I suspect - that in many Tube stations two escalators go up and only one goes down.

Plenty of researchers went the fluid route. Also intuitively, it's viable. A simple example. If you want to step onto a train, the best place to position yourself is next to the doors of the train, not in front of them. The stream of people coming out of the train makes it so that the pressure is lowest where you stand. Anyway, there are plenty of papers on the topic.

fe this paper: https://www.sciencedirect.com/science/article/abs/pii/004116...

Yes, it can make sense for some applications (when used with care).

But check the article, they're not using fluid dynamics, they actually try to estimate capacity/load using period averages and basic arithmetic. Such an approach can lead to a massive underestimation of required capacity even in contexts where arrivals aren't as batched as they are in train stations (like a call center).

Reply on news.ycombinator.com