How can computer technology be integrated into an elevator system for a hundred story office
building? How do you optimize for availability? How would variation of traffic over a typical
work week, floor, or time of day affect this? How would you test this system?
Implement a model similar to thread scheduling models found in most multitasking OS. Some
points to consider:
Some floor may have higher priorities, like the executive's floor for instance. Calls from
those floors would always be serviced first.
The elevator should only stop at a floor for x number of seconds, to avoid monopolizing the
elevator. This time would be decreased during busy hours and increased during slow hours.
If, for some reason, a floor has an elevator call outstanding for some long period of time,
that floor's priority gets boosted so that it's serviced immediately.
Based on weight (or # of people or some other similar metric) the elevator car could know
when it's full and ignore calls until there is room for at least 2 people. Stopping for just one
person isn't worth it for a full elevator.
When the elevator is idle, it can go to the floor that will need the elevator the most or that
which has the highest priority. This will save on the wait time at that floor (e.g.: the ground
floor).
Since it is a 100 story bldg, it is highly likely that many elevators are servicing it. So there
needs to be coordination between the elevators to service properly. The elevators could be
programmed so that they only serve certain floors or floor ranges. This divide and conquer
strategy can be optimized by reducing the number of floors an elevator serves based on floor
location (e.g.: the higher the floor, the less number of floors an elevator serves).
In front of the entrance door of each floor, there could be placed a BIG poster/caricature
depicting the health benefits of climbing the stairs!! This could reduce the elevator traffic.
An alternative solution could be to optimize for availability using the Poisson statistical model. The
arrival of the people can be modeled as a Poisson distributed random variable with a set of curves
being collected over a period of time (one set of curves for every day of the week) and keep on
optimizing these set of curves using linear regression so as to obtain a single curve that is the "best-
fit" for that day.The elevator would thus stand at the floor where there is a maximum probability of availability of
people. A variation of traffic would mean that the set of points over which the curve fitting is done
would change, and thus the computer system would adapt to that change. This is closer to hard disk
operation: reading the track/sector that is closer to the head position. Then apply the rule to avoid
starvation.
building? How do you optimize for availability? How would variation of traffic over a typical
work week, floor, or time of day affect this? How would you test this system?
Implement a model similar to thread scheduling models found in most multitasking OS. Some
points to consider:
Some floor may have higher priorities, like the executive's floor for instance. Calls from
those floors would always be serviced first.
The elevator should only stop at a floor for x number of seconds, to avoid monopolizing the
elevator. This time would be decreased during busy hours and increased during slow hours.
If, for some reason, a floor has an elevator call outstanding for some long period of time,
that floor's priority gets boosted so that it's serviced immediately.
Based on weight (or # of people or some other similar metric) the elevator car could know
when it's full and ignore calls until there is room for at least 2 people. Stopping for just one
person isn't worth it for a full elevator.
When the elevator is idle, it can go to the floor that will need the elevator the most or that
which has the highest priority. This will save on the wait time at that floor (e.g.: the ground
floor).
Since it is a 100 story bldg, it is highly likely that many elevators are servicing it. So there
needs to be coordination between the elevators to service properly. The elevators could be
programmed so that they only serve certain floors or floor ranges. This divide and conquer
strategy can be optimized by reducing the number of floors an elevator serves based on floor
location (e.g.: the higher the floor, the less number of floors an elevator serves).
In front of the entrance door of each floor, there could be placed a BIG poster/caricature
depicting the health benefits of climbing the stairs!! This could reduce the elevator traffic.
An alternative solution could be to optimize for availability using the Poisson statistical model. The
arrival of the people can be modeled as a Poisson distributed random variable with a set of curves
being collected over a period of time (one set of curves for every day of the week) and keep on
optimizing these set of curves using linear regression so as to obtain a single curve that is the "best-
fit" for that day.The elevator would thus stand at the floor where there is a maximum probability of availability of
people. A variation of traffic would mean that the set of points over which the curve fitting is done
would change, and thus the computer system would adapt to that change. This is closer to hard disk
operation: reading the track/sector that is closer to the head position. Then apply the rule to avoid
starvation.
Comments
Post a Comment
https://gengwg.blogspot.com/