MultiCar Elevator System using Genetic Network Programming

1
Multi-Car Elevator System using Genetic Network
Programming

• Lu Yu

2
Contents

• Background
• Proposed method
• Simulations
• Future Works

3
Elevator Group Control Systems
The system consists of cages, shafts, base floor,
general floors, call button (hall/cage),
passengers and a group controller.

• The elevator systems are the most important
  transportation systems for handling passengers in
  the building.
• The elevator systems provide safe, fast and
  economical movement for people and goods.

4
Double-deck Elevator Systems
The different from the traditional elevator is
that it has the structure of two connected cars
in an elevator shaft.
5
Multi-Car Elevator System
Multi-Car Elevator System is the system which has
two separated cars in an elevator shaft. They can
move freely.
Garage Floor
6
The movement rules of MCES
rule 1
The cars can only move vertically and cannot pass
each other.
A
A
Reversal Floor
rule 2
The cars can only move in the same direction.
B
B
Upward Operation
Downward Operation
7
The movement rules of MCES
rule 1
The cars can only move vertically and cannot pass
each other.
A
A
rule 2
The cars can only move in the same direction.
B
B
8
Destination Floor Guidance System (DFGS)
Passengers can input their destinations at
elevator halls.
Elevator Indicator
Destination Call Button
ELVATOR No.1
SERVING FLOORS 3, 15
15
16
1
3
13
14
11
12
9
10
7
8
3
5
6
3
4
1
2
1
Assigned Elevator Indicator
9
Office Building
30th Floor

• For office buildings, one elevator group can
  generally serve 15 to 20 floors.
• With more than 20 floors, the elevator system for
  buildings can be separated into low rise service
  and high rise service.

High Rise Group
Low Rise Group
The elevator has no stop in the lower floors.
Lobby
10
Background

• Problems to solve
• Due to the a large amount of uncertainties, the
  stochastic dynamic problem of MCES should be
  solved.
• MCES requires specific controls due to the
  separated cars and the need for securing
  comfortable rides.

11
Background

• Genetic Network Programming (GNP)

GNP is constructed by Initial Node, Judgment
Nodes and Processing Nodes.
Judge the specific inputs from the environment
Judgment Node
Processing Node
Process a certain function depending on the
judgment
The current node moves to the next node according
to its transition rule and generate the control
sequences.
The Time Delay
The time spent for each transition
GNP works well even in a dynamic environment such
as EGSCS.
12
Proposed Method

• Structure of the proposed method

Controller (GNP)
13
Proposed Method
Controller (GNP)

• Degree of the variance of the elevator position.
• Origin floor and direction of the call.
• The destination floor of the calls.

14
Proposed Method
Controller (GNP)

• How many and what evaluation items are to be
  selected out of evaluation items are determined
  by evolution.

transited node num.
weight
normalized value
15
Proposed Method
Controller (GNP)

• The candidate car is evaluated again by
  individual evaluation items each by each.

16
Proposed Method
Controller (GNP)

• The new call is assigned to the candidate car by
  car assignment node( Processing Node).
• After assignment, the node transition returns to
  the System Information Judgment Part, and
  identical procedures are execute for the next new
  hall call.

17
Proposed Method
Controller (GNP)
18
Proposed Method
Fitness Function
Elimination of the loop gene of GNP
Minimization of waiting time
N Total num. of passengers
w weight
19
Simulation Conditions
Evolutional Conditions of GNP
Specifications of Elevator
20
Simulation
Fitness Curves of the Proposed Method
21
Simulation
30
Regular Traffic Average Waiting Time of DDES and
MCES
22
Simulation
30
Up-peak Traffic Average Waiting Time of DDES and
MCES
23
Simulation
30
Down-peak Traffic Average Waiting Time of DDES
and MCES
24
Simulation
LWP denotes the percentage of the passengers
waiting more than 60 seconds.
25

• Thank you for your attention.