1000 Floors, 1000 Bus Stops
My crack at the classic interface design challenge, “Design an elevator for a 1000 floor building.”
Designing systems, we are like generals fighting old wars. The elevator for a 1000 floor building is a classic UI thinking question, with very pretty solutions available here, here, here, and here.
In approaching this challenge, most people assume they are building one elevator that can access all 1000 floors.
This, I will argue, is a mistake. It’s a bit like designing a bus that takes you to any of 1000 stops.
You cannot answer all interface questions with pure interface design just as you cannot answer all data engineering questions with only algorithms and data structures.
I will argue that the right design doesn’t need a different interface at all. Instead, it needs a different elevator system. Each elevator would simply present the user a smaller set of options, with the user efficiently navigating between such elevators. This way, you end up with all the elevators going to a small set of floors, like a normal elevator, and thus you can use a normal interface.
I will consider points from architecture, business, user perspective, policy, and existing elevator systems.
Considerations from Beyond Design
Architecture
Each elevator shaft occupies a fixed amount of floorspace, all the way up. For most building, people exit the building at the ground floor. Users can come and go at any hour, but the busiest traffic tends to go up in the morning and down in the early evening.
I will also assume the building has approximately equal numbers of users at each floor, though, at this height, a wider base would probably be necessary.
Business Goals
We should design to maximize ROI for the building. Toward that end, better elevator service will tend to increases rental prices. Rentable floorspace will produce revenue, while elevator shafts do not. We also want to minimize maintenance and electricity problems
Common areas (i.e. lobbies) present opportunities for services such as coffee and lunch. Such services could be included in overall rental costs and presented as amenities, or could be spaces rented out to third parties.
User Priorities
Though more research coudl help, the user’s priorities in elevator use are generally simple enough:
- Minimize trip time
- The simpler the better
- Don’t pack us in like sardines
- Do make us feel like we’re in control while navigating this crazy tower
Policy
The ADA requires elevators to serve those with accessibility issues. In most cases, this represents less than 10% of the population. Elevators must meet detailed criteria and be inspected regularly, which effectively adds to the per-elevator costs.
Conventions from Existing Elevator Systems
Double deck elevators: Two cars stacked vertically that travel together, one for even and one for odd floors. This convention fits two cars into each elevator shaft. It also requires users choose a different car for even vs odd floors.
Further, it requires an easy way to move from an even to odd floor to begin with.
Skylobbies: Used in some very tall buildings, and made popular by the World Trade Center, Skylobbies are transfer points from express elevators to local ones. They require dedicated floorspace. Skylobbies with Double Deck Elevators also need high-traffic pathways, such as escalators, from even to odd pickups.
Shared Shaft Elevators: One elevator serves one set of floors in one shaft, with another elevator above it serving a higher set of floors. Many elevators can be stacked in the same shaft this way. Recently, a more advanced system had been designed that allows different cars to use the same shaft flexibly, without predefined ranges. That convention could be adopted, but will be ignored for simplicity here.
Design Concepts
My first instinct in this challenge was to provide 1000 elevators. This is actually not a good idea because it sacrifices floorspace and would be very expensive to maintain, as explained above.
Because most elevators I ride are in buildings of less than 20 floors, I tend to assume any elevator can go to any floor. However, this is not a rule and, more importantly, it is not a good idea with 1000 floors; elevators would have to accelerate and decelerate, stopping at many inconvenient floors before dropping you off.
Here, my approach is to cut up the 1000 floor behemoth into a set of 20 floor buildings.
Design Solution
Routing individuals to 1000 different floors is like routing them to 1000 different bus stops or train stations or airports: we can save time and elevators with express and local lines.
- Build several skylobbies, perhaps every 100 floors: from the 1st to the 901st.
Skylobbies use escalators and accessibility lifts to allow traffic to even or odd level of double deck local elevators. (These accessibility lifts require extra floorspace in the Skylobby, and may be able to share a shaft with another elevator.) - Build express elevators to get to these lobbies. Each express car should have a high capacity and open at one end on the ground floor and the other end at the destination, allowing users to keep their place in “line” when traffic is bad and to continue moving in the same direction across the floor of an express elevator.
Different express lines should go to specific clusters, such as one set of elevators that goes from the ground floor to Skylobbies at 401, 501, and 601. - Build shared shaft, double deck, local elevators to get from Skylobbies to each floor. Local elevators serve less than 20 floors each, with at least 2 elevators for each subset.
- Simulate this solution in software and dial in the specific number of elevators of different types at different levels. In particular, study total trip time per user, time spent in each elevator per user, and round trip time per car.
Ideal User Flow
Going up: User enters at ground floor, takes an express elevator to closest Skylobby, then takes the correct local elevator to final destination.
Going up a little ways: User near ground level simply takes the proper local elevator.
Leaving building: User hails local elevator at specific floor, travels to Skylobby then transfers to express elevator to ground (or, if near ground floor, goes directly to ground via local elevator.)
Switching floors: User hails local elevator, arrives at Skylobby and takes either a local elevator or express to another Skylobby, then local, to target floor.
User Interface Design
Express elevators only serve a few floors, such as Ground, 101, 201, and 301. Thus they can be designed using traditional elevator design.
Local elevators serve 20 floors or less, but each cabin serves only half of those, going only to the even (or odd) floors. Thus it can also use the traditional elevator interface.
Conclusion
Elevators that can serve any floor are not efficient for very tall buildings. Local elevators are very efficient, allowing 20 shafts to serve hundreds of floors by running separate cars at different heights.
However, an even more fundamental question is probably in order. Should a 1000 floor building have elevators at all?
Perhaps each “car” should carry only a single person in an upward shaft, allowing them to move up just behind the car above, and stopping only to drop off a single passenger. When cars reach their destination, they continue to move up then switch over to a sister downward shaft where they can carry individual riders down.
This is essentially the approach of automated cars rather than buses. It has historical precedent, as an improved version of a paternoster, a funky old system still in use in some parts of Europe where one or two people hop into a car as it moves and hop out at their floor. (Of course, to do this today you’d need the car to stop for safety and accessibility reasons!)
In any case, solutions to the question that focus on interface ignore the larger questions of elevator systems in skyscrapers and deliver results that may be fun to imagine, but only by deliberately restricting the imagination.
