Otis Elevator: combining fuzzy logic and neural networks to minimize waiting for elevators

Since long waits for elevators are one of the aggravations of working or living in a highrise building, Otis Elevator and its competitors such as Hitachi, Toshiba, and Mitsubishi have tried to improve their automatic dispatchers. These are the programs that decide which elevator to send when person presses an elevator call button in a hallway. In a simple example, one person is in an elevator going down from the 16th floor, nine are in an elevator going down from the 12th floor, and someone presses the call button to go down from the 8th floor. With older dispatching systems, the elevator nearest the new call would pick up that person, even though this choice would delay nine people rather than just one who is in the higher elevator. With a fuzzy logic system, it is easier to represent and reason with additional variables, such as how full the cars are or how inconvenienced the people in different cars would be from extra waiting or from being crammed together excessively.

Fuzzy logic alone is not enough to wring the best performance from an elevator dispatcher, however. Buildings have their own idiosyncrasies, such as different office schedules in different parts of the building through the fact that a cafeteria may be on the 27th floor. To get the best performance, the dispatcher needs the ability to learn the building's idiosyncrasies. This is an ideal application of an neural network because extensive data can be captured easily.

The first Otis Elevator using a fuzzy dispatcher was installed at a Hyatt Hotel in Osaka, Japan in 1993. During this installation the Otis team learned about some factors in elevator dispatching that go beyond mere logic. They concluded that in Japan people care more about causing an extra delay for someone else, whereas in New York people simply don't want to wait, period.

Questions:

1. Use the WCA framework to organize your understanding of this case and to identify important topics that are not mentioned.

2. Identify some of the characteristics of elevator location that the dispatching system might consider.

3. Explain why you believe fuzzy logic is or is not important in the elevator system.

Source: Pinder, Jeanne E. "Fuzzy Thinking Has Merits When It Comes to Elevators," New York Times, Sept. 22, 1993, p. C1.

Back to Useful Cases from Previous Editions

Main Page About the Book Student Resources Instructor Resources Contact Us