|
Mitsubishi Electric's application of breakthrough Artificial Intelligence technology to elevator control
has resulted in improved operational efficiency and new functions that can be configured to suit a
variety of building types.
Two basic systems are available, including the  AI-22 System which is designed for small or
medium-sized buildings with two to four cars in the elevator group, and the AI-2200C System
for larger buildings with three to eight cars in the elevator group. The AI-2200C System is
especially suitable for buildings with dynamic traffic conditions throughout the day and peak
carrying times.
 |
Group Control Systems: AI-22 and AI-2200C |
|
Performance |
| Expert System and Fuzzy Logic |
An intelligent "Expert System" comprises the brains behind AI Supervisory Control. It's called
the Expert System because it incorporates the practical knowledge and experience of actual
elevator group control experts, forming a "Knowledge Database" stored inside the system's
memory. The purpose is to maximize the operational effectiveness of each car, by drawing from
the Knowledge Database and applying if-then decision rules to monitor and analyze traffic
conditions.
Additionally, the elevator control system can make decisions using fragmentary and fuzzy logic
intelligence concepts. Using its built-in "intelligence" and "common sense", the system can
determine whether or not potential car assignments will result in longer waiting times for calls in
the near future or contribute to elevator congestion. The assessment results are applied to
determine car assignments in order to improve overall service.
|
 AI-2200C System Configuration |
| Psychological Waiting Time Evaluation |
In general, people will patiently await a bus or train for ten minutes and even more, but irritation
sets in almost immediately if an elevator car doesn't arrive at once. After comprehensive study of
this phenomenon, Mitsubishi Electric has formulated a psychological evaluation function based
on the following principle:
The irritation of a passenger waiting for elevator car arrival is proportional to the square of the
actual waiting time.
Based on this principle, car assignments to hall calls are performed by the AI Supervisory Control
System on the basis of evaluation results generated by the sum of such factors as forecasted
waiting time, probability of being bypassed for a hall call, probable time required for traveling
after car assignment, current car load and others, owing to the evaluation function's coefficient
diversity.
| Cooperative Optimization Assignment |
When a hall call is registered, the Cooperative Optimization Assignment Algorithm predicts a
potential hall call that could reduce transport efficiency, and allocates the best elevator through
evaluations of the registered hall call and the forecasted call. Cars are ready for any hall call,
and cooperate for optimum elevator operation.
| Dynamic Rule-set Optimizer |
Mitsubishi Electric's Dynamic Rule-set Optimizer selects optimum car allocation through
"Rule-Set" simulations. The Neural Network technology enables the system to continually and
accurately predict the passenger traffic within intervals of several minutes.
A high speed Reduced Instruction Set Computer (RISC) runs real-time simulations using multiple
Rule-Sets and the predicted passenger traffic to select the best Rule-Set which optimizes transport
efficiency.
|
Rule-set Selection with Real-time Simulation Example |
During morning up peak time, an ideal rule-set is selected every few minutes according to building
traffic conditions.
| Determination of Traffic Flow with Neural Networks |
In order to ensure efficient conveyance of elevator users in a building, elevators must operate
consistently with optimum service patterns. Efficiency is maintained by adjusting service patterns
to passenger flows within the building, which fluctuate constantly depending on time zones such
as morning up peak, lunchtime and evening down peak.
Mitsubishi Electric's AI-2200C System applies Neural Networks technology to precisely recognize
the current traffic flow in real time, and to select the optimum service pattern in a timely manner.
It works by applying an ideal rule-set for each traffic pattern and for each day of the week, using
the passenger flow data calculated by past car load records, as well as frequency of car and hall
calls on each floor.
At the same time, it predicts traffic flow in the next several minutes based on both past and current
operating data. The selected rule-sets are simulated under the predicted flow in order to determine
the optimum service pattern. Accuracy in predicting traffic flows and selecting optimum service
patterns is enhanced through the accumulation of actual elevator operation data, in order to
achieve optimum elevator service.
| Destination Oriented Prediction System (DOAS-S) |
The Destination Oriented Prediction System (DOAS-S) is a unique allocation system that provides
passengers with the fastest and least crowded route to their destinations. When a passenger
enters a destination floor number on a hall operating panel, the best car is automatically allocated
to the passenger and the elevator number is displayed on the panel. The passenger goes to the
assigned car, confirming the elevator identification on a hall lantern with an elevator number plate.
The passenger does not need to press the floor buttons on a car operating panel in the car as the
destination floor has already been registered.
Mitsubishi Electric's Motor Drive Mix technology is designed to decrease waiting time during
congested periods of elevator use. It does so by increasing elevator acceleration between floors.
The elevator's acceleration and deceleration rates are adjusted according to car load and traffic
conditions to maximize driving system efficiency.
| Immediate Prediction Indication |
As a means to reduce user irritation while waiting for an elevator car to arrive, the Immediate
Prediction Indication system provides a measure of feedback that helps the passenger predict
car arrival.
Here's how it works: Once the user has registered a hall call, the first available car for quickest
response is designated, the hall lantern lights and a chime sounds once to indicate which elevator
door will open. As the car approaches the lantern will begin to flash and the chime will sound twice,
indicating imminent arrival.
| Energy-Saving Operation -Allocation Control |
This system selects the elevator in a group that best balances operational efficiency and energy
consumption. Priority is given to operational efficiency during peak hours and energy efficiency
during non-peak hours.
Car allocation that maximizes operational efficiency does not necessarily translate to energy
efficiency. A car uses energy efficiently when it travels down with a heavy load, or up with a light
load. Accordingly, if multiple cars have the same traveling distance, this system chooses the car
that requires the least energy.
Through a maximum 10% reduction in energy consumption compared to our conventional system,
this system allows building owners to cut energy costs without sacrificing passenger convenience.
| 
