Proc.of the Intl.Conf.on Future Engg.

Trends(ICFET 2K10),SSE,Saveetha University,India

FPGA Implementation of Fuzzy Logic Controller For

Elevator Group Control System
Rajesh Kumar Patjoshi Kamala Kanta Mohapatra
Electronics and Commonucation Engineering, Electronics and Commonucation Engineering,
National Institute of Technology, Rourkela, India National Institute of Technology, Rourkela, India
Email : [email protected] Email : [email protected]

Abstract : - waiting time (AWT) of passenger is presented in the

This paper concerns on implementation of a fuzzy logic literature[1].
controller (FLC) on a Field-programmable Gate Array (FPGA) Fuzzy Controller described in [1], has evaluated six
system for intelligent control of elevator system. The search for
set of rule and each rule set consists of a different number of
an intelligent group controller that can satisfy multicriteria
requiements of an elevator group control system has become a rules (between 12 and 14). Hence fuzzy computation time is
great challenge for researchers. This proposed approach is based large for response to hall call even if it is a self tuning fuzzy
on algorithm which is developed to reduce the amount of controller.
computation required by focusing only on the relevant rules and In the present investigation we present a conventional
ignoring those which are irrelevant to the condition for better FLC that uses reduced number of rules that facilitates
performance of the group of elevator system. Simulation was reduction of computation time of the FLC. In this work,
carried out by considering two inputs i.e. elevator car distance
Simulation was carried out by considering two inputs i.e.
and number of stops. Based on these data the Fuzzy Controller
can calculate the Performance Index (PI) of each elevator car. elevator car distance and number of stops.
And the car which has maximum PI gives the answer to the hall The design is fit on Spartan 3E (XC3S500E) having
calls. This would facillitate reducing the average waiting time maximum operating frequency is 35.754 Mhz. the design uses
(AWT) of the passenger. 35% slices, 15% slices Flip flop, 19% 4 inputs LUTS and 35%
Key words :- Fuzzy Logic, Intelligent control, Performance Index IOBs.
Average Waiting Time (AWT).
II. DESIGN METHODOLOGY OF PROPOSED FPGA
I. INTRODUCTION BASED FUZZY LOGIC CONTROL FOR
ELEVATOR GROUP SYSTEM
In control engineering analysis and design, FPGA
based Fuzzy logic control (FLC) for elevator group system, A conventional FLC is designed on a simple concept
has been attractive because it offers a compromise between by reducing the number of rules that facilitates reduction of
special purpose ASIC hardware and general purpose computation time of Fuzzy Logic Controller for better
processors. This purposed FPGA based Fuzzy Logic performance of elevator group control system. Basically the
Controller for elevator group system can reduce the design FLC is divided into four components. Figure-1 shows the
development cycle, simplify the design complexity, and also diagram of the Basic Structure of Proposed FPGA based
improve the control performance that simplifies imple- Fuzzy Logic Controller for intelligent control of Elevator
mentation and reduces the hardware costs. System. It consists of Fuzzification Module, Rule Base, Fuzzy
Many researchers have reported about Fuzzy Logic Interface Engine, Defuzzification Module.
Controller for group control of elevator system[2]. They P,
described the validation of five dispatching algorithms for Rule P2
elevator system that were implemented on spartan-3 FPGA Base
based board in an integrated approach reducing the area and D1, N1
Pn
improving performance. The overall system is composed of
Dn Nn Fuzzi- Fuzzify Elevator
several LCS, which implement the dispatching algorithm. The fication inte-
Defuzz-
ification
Control
Output
group
rface Control
EGCS-based Fuzzy Logic (FEGCS) runs on a PC and under Module
Engine
Module Selector System

different traffic situation determines the best algorithms to be Crisp

run in each LCS in order to reduce the avergae waiting time of input Crisp output

passenger and also reduces the power consumption. D : Car Distance N : No.of Stops
Elevator group controller based Fuzzy Logic Fig. 1 : Basic Structure of the proposed Fuzzy Logic
Framework with self tunning scheme for reducing the average Cotroller for Elevator Group Control System

ISBN:978-81-906109-7-1
Proc.of the Intl.Conf.on Future Engg.Trends(ICFET 2K10),SSE,Saveetha University,India
A. Fuzzification Design Module for the proposed FPGA small window and only four adjoining rules can be viewed
based Fuzzy Logic Controller for Elevator Group through this window at a time. Therefore, instead of having to
Control System.: - access 25 rules, the inference engine only has to access four
rules during every computation. The window can move around
The fuzzification block has five outputs, one for each the FAM table and its position is identified by an index j
fuzzy value defined in the inputs Universe of Discourse. defined as:
However, the fuzzification process entails that, for any single
crisp value of the input Xi, only two adjacent fuzzy values are ADR1 = ―00‖ & ADR2 = ―00‖  j = 0
significant (with non-zero-membership values). By ignoring ADR1 = ‖00‖ & ADR2 = ―01‖  j = 1
the insignificant fuzzy values, the number of output signals ADR1 = ―00‖ & ADR2 = ―10‖  j = 2
can also be reduced from five to two. The possible ADR1 = ―00‖ & ADR2 = ―11‖  j = 3
combinations of significant fuzzy values for an arbitrary ADR1 = ―01‖ & ADR2 = ―00‖  j = 4
inputs are : ADR1 = ―01‖ & ADR2 = ―01‖  j = 5
ADR1 = ―11‖ & ADR2 = ―11‖  j = 15
B i1 and Bi2 ; and Bi2 and Bi3 ;
There are sixteen ‗window positions‘ altogether and
It is found that using just three variables, ADRi, Bi_A the first six are shown in Fig.3. The shaded blocks are the
rules which are considered relevant for the input conditions
and Bi_B, all the combinations can be sufficiently represented
corresponding to the index j. To distinguish the ‗windowed‘
for any value of xi as shown by the following statements:
view of the FAM table from the original table, the first is
ADRi = ―00‖ : Bi_A = B i1 , Bi_B = Bi2 referred to as the ‗Mini‘ fuzzy associative memory (FAM)
table. When the window technique is applied to the FAM table
ADRi = ―01‖ : Bi_A = Bi2 , Bi_B = Bi3 in the Table I, it is observe that number of the mini-FAM
tables are identical (e.g. J=1, J=4) out of the sixteen mini–
ADRi = ―10‖ : Bi_A = Bi3 , Bi_B = Bi4 FAM tables, only seven unique tables are formed.
ADRi = ―11‖ : Bi_A = Bi4 , Bi_B = Bi5
Figure-2 illustrates how these conditions correspond Table-I ( FAM Table of the FLC Design)
with the universe of discourse.

Membership
function

VS S M L VL

Universe of
Discourse

ADRi ―00‖ ADRi ―01‖ ADRi ―10‖ ADRi ―11‖

Fig. 2 : Input Fuzzy set

B. Mini FAM Table :-

The FAM table of the FLC design is shown in table-I.

here we describes an algorithm which is developed to reduced
the amount of computation required by focusing only on the
relevant rules and ignoring those which are irrelevant to the
condition in question,. It is known that for every set of inputs,
only for fuzzy rules are relevant at any one time.
An easier way of explaining the technique is to
imagine the entire FAM table to be covered from view. Access
to the content of the FAM table is only allowed through a Fig.3 : Mini FAM Table

ISBN:978-81-906109-7-1
Proc.of the Intl.Conf.on Future Engg.Trends(ICFET 2K10),SSE,Saveetha University,India
C. Defuzzification algorithm :- call is registered, relevant data from all cars in the group are
needed for computation of the value of the input variables.
The function of the defuzzificaition is to convert the The data required from every car are, its present position in
fuzzy output value of the control system into the the building, motion status, speed direction of travel.
corresponding crisp value of the membership function shown
in fig.4. This is achieved using the weighted average Here we use two input variables for fuzzy evaluation
defuzzification method. This defuzzification operation they are the car distance and number of stops. Their
requires several multipliers and a divider. Behavioural definitions are as follows : -
modelling in VHDL supports multiplication and divi-sion but 1) Car distance : The distance travelled by car to move
these operations are complicated to realise in the systhesis and from its present position to a hall call floor when a hall
implementation stages. call is registered.
NVB NB N NS Z PS P PB PVB 2) Number of stops : The number of stops of floors a car
has to stop at to load or to unload passengers before a
Single ton
Membership
hall call floor is reached.
Function
For every car in the group, the values of the two
input variables are calculated based on the data supplied by
each car controller. The calculated values of each car are then
processed by the FLC and from fuzzy evaluation, a perfor-
mance index (PI) is assigned to each car. The PI denotes the
-1 0 1
suitability of the car to answer a hall call at a particular instant,
Universe of
the car with the highest PI value is considered the car most
Fig. 4 : Output Fuzzyset
Discourse suitable to attend to the hall call.

In the proposed approach, only cosider the four

significant rules consequents are considered. Therefore, the
number of rule-consequnets to be aggregated is reduced but
the allocation of correct weightings for the significant output
values becomes slightly more complicated. From the tables in
Fig.3 it is obvious that regardless of the WIN value, the
consequents C2 and C3 always point to the same fuzzy value
(e.g. when WIN = ―0000‖: C1PVB, C2PB, C3PB,
C4P). This implies that only C2 and C3 have to be
aggregated, hence:
DA = C1
DB = max[ C2, C3]
DC = C4
Fig. 6 : The process flow for the elevator dispatching strategy with
Where DA, DB and DC represent the membership the use of a fuzzy logic controller
function of the output fuzzy values. Thus the modified output IV. SIMULATION RESULT AND FPGA
fuzzy set is shown in figure 5, IMPLEMENTATION
DA DB DC
A. Simulation Detail
Single ton
Membership
Parameters
Function Number of floors : 40.
Number of elevator Car : 8.
Floor height is 4 mtrs.
Number passenger per floor in 5 minutes : 40.
Universe of Number of FLC run per Traffic stuation : 13.
Discourse Number of Stops to hall call floor is 20.
Fig.5 : Modified output fuzzy set Elevator casr distance from hall call floor is 100mtr.
III. DESIGNED METHODOLOGY FOR ELEVATOR We consider here two inputs i.e. Elevator Car
GROUP CONTROLS Distance and Number of Stops. Based on these datas the
Figure-6 shows the process of selecting the most fuzzy controller can calculate the performance index (PI) of
suitable car to answer a hall call through the FLC system. This each elevator‘s car and evaluate the maximum (PI) of each
paper considers the up-down call button system when a hall car, the car which has the largest PI donates the suitable car to

ISBN:978-81-906109-7-1
Proc.of the Intl.Conf.on Future Engg.Trends(ICFET 2K10),SSE,Saveetha University,India
answer a hall call at a particular instance. Base on the data
given to the input of fuzzifier, elevator car 3 has the highest PI
to response the hall calls, which is shown in simulation result
in figure-7.

Simulation Result for PI evaluation

Fig. 8 : RTL view of the FLC Arhitecture

V. CONCLUSION

De-fuzzification Result
This paper presents an approach for the
implementation of a fuzzy logic controller for elevator group
Fig.7 : Output Simulation Result
control system on FPGA using VHDL. The controller for
Based on these data the average waiting time of the elevator group control system is implemented on a Xilinx
passenger become 48 sec for 40 floors, whereas this time was Spartan-3E FPGA . The implementation of the fuzzy logic
taken to be 52 sec for 35 floors [1]. The average waiting time controller is very straight forward by coding each component
is calucalted based on the average traffic flow in particular of the fuzzy inference system in VHDL according to the
time instance i.e.morning (8am to 10am) and evening (5pm to 7pm) design specifications. The design of the FLC is highly flexible
and computation time of the fuzzy controller. Table-II show as the membership functions and rule base can be easily
the comparison result between proposed approach and 35 changed with reduced rule techniques. Because of the reduced
floors which was described in [1]. rule techniques the computation time of the fuzzy controller is
Table-II
reduced and elevator group control system gives faster
performance by reducing the average waiting time (AWT) of
AWT Time
the passenger is reduce upto 48 seconds.
Average Waiting Time of proposed approach for 40 floors. 48 Sec
Average Waiting Time of proposed approach for 35 floors 52 Sec
described in [1]. VI. REFERENCE
1) J. Jamaludin, N.A. Rahim, W.P.Hew, ―Development of self tunning
B. FPGA Implementation :
fuzzy logic controller for intelligence control of elevator system‖ Journal
The design of fuzzy logic controller for group of of Engineering Applications of Artificial Intelligence 22 (2009)1167–
elevator control system is implemented using VHDL. 1178.
Synthesis process has performed using Xilinx tools[6] for 2) Daniel M. Munoz, Carlos H. Llanos, Mauricio Ayala-Rincon,
synthesizing the compiled VHDL design codes into gate level ―Distributed Approach to group control of elevator systems using fuzzy
schematics. The VHDL codes are synthesized for converting logic and FPGA implementation of dispatching algorithms‖, Journal of
Engineering Applications of Artificial Intelligence 21 (2008)1309–1320.
into RTL view of the FLC architecture as shown in figure 8.
3) Daijin Kim, member, IEEE ―An Implementation of fuzzy Logic
The Technology mapping has chosen in this project from Controller on the Reconfigurable FPGA system,‖ IEEE Transactions on
Spartan 3E (xc3s500E) with FG320 package and a speed industrial Electronics, Vol.47, No.3, June 2000.
grade of -4. The synthesized schematic is also simulated to 4) Gudwin, R., Gomide, F., Andrade Netto, M., 1998. ―A fuzzy elevator
ensure the synthesized design functions. Table-III shows the group controller with linear context adaptation‖. In: Proceedings of
device utilisation summary of the FPGA based fuzzy logic FUZZ-IEEE98, WCCI‘98—IEEE World Congress on Computational
Intelligence, Anchorage. IEEE, Alaska, USA, pp. 481–486.
controller for inteligent control of elevator group system.
5) Daniel M. Munoz Carlos H. Lianos, Mauricio Ayala-Rincon Rudi van
Table - III : Device Utilisation Summary Els, Renato P. Almeida, ―Implementation of Dispatching Algorithms for
Elevator Systems using Reconfigurable Architectures‖ IEEE Conference
No. of slices 1610 out of 4656 35 %
2007 February.
No. of slices flip flop 1415 out of 9312 15 %
No. of 4 input LUTS 1826 out of 9312 19 % 6) Xilinx ISE 10.1 Software manuals www.xilinx.com.
No. of Bonds IOBs 81 out of 232 35 % 7) Fuzzu Control System Design and Analysis : A Linear Matrix Inequality
No. of GCLK 1 out of 8 13 % Approach by Kazuo Tanaka, Hua O. Womg, John Wiley and Sons
Max. freuqncy, 35.75MHz Publicatins.
Total memory uses 206524 kilobyte 8) Fuzzy Controller by Leonid Rezmik Victoria University of Technology,
Minimum period 28.855 ns Melbourne, Australia, Newnes Publication.

ISBN:978-81-906109-7-1