There exist systems where, instead of selecting a floor when you enter an elevator, you select a floor in the hall and then are told which elevator to take. I wonder how much this advance knowledge about who is going where helps with scheduling. Intuitively I feel like it should help substantially, but in practice people get confused by these systems and end up on wrong floors.
It can significantly increase the efficiency in a high-rise elevator bank, but like many things it takes a little bit of time to train people to utilize it properly.
It’s a right pain when you only visit a building occasionally. You push a button, ignore the number, then just move to the next loft which opens along with everyone else, then realise you’re in the wrong lift.
One of our offices has this system (it makes no sense too - there’s only 5 floors) and it catches me every time. The building is designed to discourage use of stairs too (they’re behind hidden doors)
The Terminus building in Atlanta works like this. You select your floor as you walk in (it's possible to do mid stride) and then the screen says "ELEVATOR 4" so when you get there you just walk in to an elevator that has no buttons, just a screen that says "TODAYS WEATHER: JSON FEED NOT FOUND".
A problem with the floor selector in the hall is when some people enter the elevator together without selecting a floor. The system thinks that there is still space in the elevator, but in fact it is full, so it stops at different floors without anyone being able to enter. I've seen this happen quite often in a corporate building during the end of business hours and lunch hours.
At Twitter/X HQ they have these, and you can "trick" it into going express to your floor by pressing the button in the hall repeatedly so the car then thinks it is full and won't stop at other floors.
Having said that, the default system is also ignorant to the current weigh of the elevator so it is still subject to stopping on floors without there being any space left.
Trying to figure out space available from the number of people seems unreliable, but I wonder if you can use a set of heuristics such as having a max weight at which point the elevator is assumed full and if it stops to let someone on but the weight doesn't increase assume the elevator is full (maybe require this to happen twice in a row?, IDK, false positives probably aren't too expensive.)
My assumption is that, above a certain number of floors, the floor selector in the hall is better because it has more advance information for any planning. The downside being that it's unfamiliar to anyone who doesn't use the system regularly.
We have this in my building. The lifts are terrible. I suspect the algorithm is actually better, but we can’t A/B test to see how bad it would be otherwise, and the feeling is that you have less control, especially when a lift goes past you and doesn’t stop. The result is that people hate it and blame the lifts.
I've used a system that each worker of the building would be given a card. In the hall, there's a card reader. There's multiple elevators in the same building. To use the elevator, one would need to tap the card against the card scanner. Then the screen on the scanner would tell you which elevator's gonna serve you.
I've also been in buildings with double-decker elevators. So it stops at 1&2, 3&4 etc. So if you enter the building at floor 1, but need to go to floor 12, you have to take the escalator and enter the elevator at floor 2 so that you later can jump off at floor 12 when it stops at 11&12.
> We have installed these elevators in Roppongi Hills Mori Tower. Mori Tower’s “super double-deck elevators” are equipped with a pantograph joint on the part of the elevator that connects two passenger cars that can expand and contract by up to two meters, making it possible for the elevator to serve floors with varying heights.
> Nippon Otis Elevator Company and Mori Building have a joint patent on this pantograph joint structure, the first of its kind in the world.
There are also lobbies where you choose an elevator for a specific range of floors. I wonder if that's a legacy solution and what you're describing has become the new standard.
Both work, but I think the range elevators simply have no exits for the unsupported floors and can go high speed to the first supported elevator. They probably do have exits though, in case another elevator breaks down.
Taller buildings will require you to transfer to a different elevator once you've reached a certain floor; one because else you get cables that are hundreds of meters long (with the weight to match), and two because else you'll need more elevators and -shafts to service all hundred+ floors in a timely fashion, which takes up real estate / floor space.
> They probably do have exits though, in case another elevator breaks down.
Floor space is generally too precious for that. In a typical multi-bank elevator setup, lower floors get their toilets or other utility spaces placed between two banks of elevators that don't stop at that floor; the corresponding point in the upper floors is the elevator lobby for an upper-floor elevator bank.
There's a widely reported story that a group of people stuck in an elevator in the WTC on 9/11/2001 used a window washer's squeegee to cut their way through multiple layers of drywall to escape into a 50th floor bathroom.
I've seen that in several buildings in Dubai. Considering that most of Dubai was built since around 2000, these kinds of advanced elevator controllers should've been available when at least some of these buildings were planned and built. So no, probably not legacy. Maybe those hall-floor-selector systems are simply much more expensive. Or maybe they only make sense when you need a really high throughput.
I swapped in VS Code’s editor component (Monaco) and added TypeScript type declarations someone had posted. Note: The official game isn’t accepting PRs anymore, but details and code are here: https://github.com/magwo/elevatorsaga/pull/137
A long time ago I took an AI class where we wrote an elevator simulation. The elevators were intelligent agents that tried to learn strategies for moving people faster / reduce wait times / etc. This was long before neural networks came into fashion. We used logic, rules, but also decision trees. Fun times.
Hehe, from time to time I encounter an elevator like this:
You step in, select floor, doors close, and then nothing happens.
Turns out elevator controls want the doors to close first, and only then look at what floor you select. And no that wasn't a mis-press or something. Where encountered, it's repeatable.
I was curious to see if someone would mention this. I would play that game as a teenager for a day and then ragequit only to come back a week later. I hate-loved that game.
I've used something like this (my code, Java, ~1k LOC) for code interviews,
because I was always thinking about what would be the best algorithm for an elevator when going in on - to this day.
1. Explain what is going on
2. Fix a bug in the code
3. Add a small feature
Agreed, but I find the JavaScript to be difficult to work with. (This is less about he language and more about the fact that the API is a little vague and errors are not well reported.)
