4 ms·

Elevator Expertise from an Expert

16 points by danw 19y ago

neilc 19y ago
Wow, that's very cool -- I remember thinking about elevator scheduling in the past as an interesting theoretical problem. Scheduling algorithms for taxis, bus services, and similar systems is another related problem that seems interesting. A friend of mine spent some time writing software and setting up a small compute cluster to solve variants of the traveling salesman problem for a courier service. More broadly, I think there are many situations in which CS can be applied to yield an incremental improvement in some every-day task where the standard solution is perfectly reasonable, but could be made by efficient by an intelligent algorithm. As the cost of building in these algorithms continues to decrease (and as WiFi + RFID becomes more pervasive), the number of problems for which it will be feasible to apply a sophisticated solution will continue to increase. An incremental improvement in a single elevator's dispatch time might not mean all that much, but when it's applied to tens of thousands of elevators that operate every day, that's a nice productivity hack.
Tichy 19y ago
I think that notion of "experience design" is very important, and that people's perception might differ from reality. Like I think I have heard that Menus are actually faster on Linux than on Windows, but on Windows it feels faster because they are animated immediately. I also read somewhere of a case where designers solved the waiting problem by putting up mirrors next to the elevators. That way, people were busy looking at themselves and didn't notice the wait time as much.
- fnord123 19y ago
  >but on Windows it feels faster because they are animated immediately. Context menus take between 3 and 6 seconds on my Windows XPSP2 machine to load the desktop context menu as it thrashes the hard drive. That's after a memory upgrade where it took half a minute to load the desktop context menu. My slower KDE machine loads context menus instantly. Where are these Windows machines with instant animating menus?
edw519 19y ago
"it's the intangible quality of what users are experiencing that becomes the toughest nut to crack" Good thing Microsoft isn't in the elevator business. Imagine what a splash screen would look like.
danteembermage 19y ago
This reminded me of a keynote speech at the dynamical systems conference at Snobird a few years back about the limits of computational mathematics, i.e. the human limits of asking the wrong question. The aged speaker was apparently a consultant for the British transportation authority; they wanted a model that would optimize their highway expenditure to put the money where it would do the most good along multiple criteria, time on the road, overall budget effects, safety, etc. The team spent a significant amount time getting all the dynamic pieces of the model to interact and then started running solutions. Strangely though, no matter what reasonable weights or parameters they threw at it, it always recommended pedestrian safety expenditure of zero. After a lot of puzzling they finally figured out what was going on. You see, a significant portion of the pedestrians on English crosswalks were aged pensioners. Since the model incorporated overall budget effects into its preferences, every dollar spent on pedestrian safety lead to the decreased risk of accidental death of net-revenue-drain individuals, i.e. little old ladies walking to the corner shop. So the models did an excellent job of solving the wrong problem, GIGO. Likewise, actual time spent from start to finish is most definitely not the same at perceived time from start to finish, with the second probably being to most important. Which, in case you didn't know, is why hotel elevators almost invariably have mirrors on either side of the door, you reflection gives you something to do other than thinking about how long your elevator is taking, a considerably cheaper solution than trying to solve the computational problem of maximal overall trip efficiency. That said I'm totally jealous they get paid to play with that problem. If I were there I'd be like the guys in the arm chairs from the Collins College advertisements doing the "dude, I can't believe we get paid to play games!"

16 points by danw 19y ago

neilc 19y ago

Wow, that's very cool -- I remember thinking about elevator scheduling in the past as an interesting theoretical problem. Scheduling algorithms for taxis, bus services, and similar systems is another related problem that seems interesting. A friend of mine spent some time writing software and setting up a small compute cluster to solve variants of the traveling salesman problem for a courier service. More broadly, I think there are many situations in which CS can be applied to yield an incremental improvement in some every-day task where the standard solution is perfectly reasonable, but could be made by efficient by an intelligent algorithm. As the cost of building in these algorithms continues to decrease (and as WiFi + RFID becomes more pervasive), the number of problems for which it will be feasible to apply a sophisticated solution will continue to increase. An incremental improvement in a single elevator's dispatch time might not mean all that much, but when it's applied to tens of thousands of elevators that operate every day, that's a nice productivity hack.

Tichy 19y ago

I think that notion of "experience design" is very important, and that people's perception might differ from reality. Like I think I have heard that Menus are actually faster on Linux than on Windows, but on Windows it feels faster because they are animated immediately. I also read somewhere of a case where designers solved the waiting problem by putting up mirrors next to the elevators. That way, people were busy looking at themselves and didn't notice the wait time as much.

fnord123 19y ago
>but on Windows it feels faster because they are animated immediately. Context menus take between 3 and 6 seconds on my Windows XPSP2 machine to load the desktop context menu as it thrashes the hard drive. That's after a memory upgrade where it took half a minute to load the desktop context menu. My slower KDE machine loads context menus instantly. Where are these Windows machines with instant animating menus?

edw519 19y ago

"it's the intangible quality of what users are experiencing that becomes the toughest nut to crack" Good thing Microsoft isn't in the elevator business. Imagine what a splash screen would look like.

danteembermage 19y ago

This reminded me of a keynote speech at the dynamical systems conference at Snobird a few years back about the limits of computational mathematics, i.e. the human limits of asking the wrong question. The aged speaker was apparently a consultant for the British transportation authority; they wanted a model that would optimize their highway expenditure to put the money where it would do the most good along multiple criteria, time on the road, overall budget effects, safety, etc. The team spent a significant amount time getting all the dynamic pieces of the model to interact and then started running solutions. Strangely though, no matter what reasonable weights or parameters they threw at it, it always recommended pedestrian safety expenditure of zero. After a lot of puzzling they finally figured out what was going on. You see, a significant portion of the pedestrians on English crosswalks were aged pensioners. Since the model incorporated overall budget effects into its preferences, every dollar spent on pedestrian safety lead to the decreased risk of accidental death of net-revenue-drain individuals, i.e. little old ladies walking to the corner shop. So the models did an excellent job of solving the wrong problem, GIGO. Likewise, actual time spent from start to finish is most definitely not the same at perceived time from start to finish, with the second probably being to most important. Which, in case you didn't know, is why hotel elevators almost invariably have mirrors on either side of the door, you reflection gives you something to do other than thinking about how long your elevator is taking, a considerably cheaper solution than trying to solve the computational problem of maximal overall trip efficiency. That said I'm totally jealous they get paid to play with that problem. If I were there I'd be like the guys in the arm chairs from the Collins College advertisements doing the "dude, I can't believe we get paid to play games!"