A modern paternoster elevator (for cars and maybe people)
Earlier this week, I wrote about making a subway for robotic vans which just has tunnels and ramps to the surface, rather than the vastly more expensive system of giant stations we use for today's underground transit. It offers the chance to save immense amounts of money because stations are expensive to build and maintain.
The Boring Company (Elon Musk's tunnel effort) just released a video with their very similar version. Instead of stations, elevators take buses from street level down to the exit lanes of the tunnels. They are stating that they want to think more about group vehicles than having private cars in the tunnels, but the right answer is to have both, and maximize use of the tunnels, with mostly vans in rush hour and mostly cars outside of that.
The simplest design, if starting from scratch, is probably just a ramp to the surface. It's fast and very simple, and uses the energy of travel of the vans to get to the surface, and gravity to speed up when going down. This follows the "stupid network" principle that made the internet great -- make the network stupid and put the smarts in the cars.
In dense existing cities, however, it may not be possible to create ramps, because they need real estate to merge with the road network, and also just below the ground as they come to the surface -- in dense areas that space just below ground is usually building basements.
The Boring Co design suggests standard elevators. Elevators do not have good throughput -- it probably takes a least a minute for an elevator to take a round trip. Ramps can input our output vehicles almost as fast as you like, just like roads.
There is one design of elevator which has high throughput called the Paternoster. These elevators are quite rare today because they are a bit dangerous for people, as you will see if you look at some of the videos. The Paternoster design has a series of cabins which travel in a loop, always moving. You jump into and out of them as they go by. Don't time your jump well and you might get crushed. (In a modern design, sensors would shut the system down but that's also not very useful.)
It seems to me you could make a paternoster for robots a bit more easily. In this design, imagine you want to enter a downward cabin. Your robocar would drive onto a platform, and the door behind it would close. As a cabin descended, the door to the moving cabin would open as they became level. Then, the platform would start to lower into the ground at the same speed as the cabin. As it lowered, the car would drive into the cabin and the doors close behind it. The platform would then stop hard and return quickly to the surface, ready to take another car.
Your cabin would then descend to the underground level. As it approached another platform would have risen up, and would start a descent to match your cabin. Your car would drive onto the lowering platform, which would then slow and stop so you could drive off it and into the tunnels.
If, for some reason, a car did not complete its transfer during the time, the entire system would slow down to give it time, or it would come to a stop if something got stuck.
Because the cabins are on a loop, you can have one transfer at the top and one at the bottom, for use in both directions. If you get into a downward car at the bottom, it just loops around immediately and starts going up.
For greater throughput, you could also build a double-paternoster design. At the top would be a very small paternoster loop which sometimes moved at the same speed as the main loop, but also stopped and started. You would enter and exit the boarding loop from the outside while it was stopped, and in fat it would not start moving until all parties had entered their boarding cabins. Then it would start up and match speed and sync cabins with the main loop, allowing a decent amount of time for transfer. This would allow 2 or more simultaneous transfers for higher throughput.
At the other end of simplicity, since these are robots, there could just be a ramp that is fixed at one end and can go up and down at the end near the cabins. The robot would drive onto the ramp and by the time it got to the end, the ramp would be in sync with the ascending or descending cabin, and it would keep going until all the way in. Scary perhaps but with the right safety interlocks (ie. the system stops unless weight is off the ramp by the time it goes close to the endpoint) it could work and be fairly low cost and even more mechanical than electronic.
Just a short elevator
As I said above, ramps are probably a much better idea than elevators, but you can't always have them. It may however make the most sense to have only a short elevator which takes you down just far enough that a ramp can then go from the elevator base down to the tunnel. Perhaps just 20 to 30 feet. With a short journey, and the rapid and reliable load/unload that robocars would do, you might get adequate throughput from a small number of traditional elevators. One van would drive into the cabin as the other is leaving, for a very short turnaround at either end.
Paternoster for people
I got some inspiration from this Hitachi Paternoster design which has a larger gap between cabins and an effective ability for the cabins to stop for a while during that gap. Another attempt at something similar is the design of most ropeway gondolas which slip off the track at the stations so they can move slowly (bunching up) for boarding and departure. You can't do that with an elevator because it's hard to safely go off the track when vertical.
This does inspire a design for a pedestrian elevator. Again, you would have a paternoster circulating in the building. Passengers would enter a box next to it which is in another cabin. Their cabin would then match speed with the paternoster cabin, and slide from one cabin to the next. The big issue here is that the sideways acceleration (especially the stopping) might be quite jarring, since you want to do it as quickly as possible. Since the entry cabins need to be able to climb a floor or two and return, they take up the whole room of another elevator shaft, making this useful only in larger setups.
Why is this so important? Traditional elevator design is the largest barrier to the construction of skyscrapers. If you want enough elevator capacity to handle a very tall building, soon, the entire building has to be elevator shafts if you stick to the one elevator per shaft approach. Super-tall buildings solve this by having elevator lobbies, where you transfer elevators. The problem can also be solved if more than one elevator can occupy a shaft, which is what many people have tried to do. It's hard to do with full safety.
One downside of the paternoster approach is that it doesn't get to be super fast, because you need enough time for the transition. If you want 6 seconds for a transition over two floors you can only go 20 floors per minute, which obviously isn't acceptable going to the 100th floor of a skyscraper. You could build a Paternoster of this design which sped up when nobody needed to get on or off and slowed down when they did. This would be somewhat jarring to occupants (though of course a modern elevator is starting and stopping all the time rather than just slowing down) but would mean the elevator could go high speed when demand was low, and lower speed when demand (and need for throughput) is high.