FUZZY CONTROL Precise control in a fuzzy world

by Musa Abubakar

Department of Electrical Engineering, Faculty of Engineering Bayero University, Kano. Nigeria February, 2013

Contents of the Presentation

Introduction

Applications of Fuzzy Control

Fuzzy Theory- Fuzzy Sets & Fuzzy Logic Fuzzy Controller Design

Example: Fuzzy Controller for Inverted

Pendulum Conclusion

Contents of the Presentation

Introduction

Applications of Fuzzy Control

Fuzzy Theory- Fuzzy Sets & Fuzzy Logic Fuzzy Controller Design

Example: Inverted Pendulum Fuzzy

Controller Conclusion

Introduction
Pioneers in Fuzzy theory and control - Lotfi Zadeh Introduced Fuzzy sets and Fuzzy logic - Ebrahim Mamdani First practical Fuzzy Controller Why Fuzzy control: Advantages of Fuzzy control - Fuzzy Control is based on natural language - Fuzzy Controllers are relatively easy to design - Systems with no mathematical models can be controlled - Fuzzy Controllers are robust - They are versatile; there are many parameters that may be adjusted to achieve required performance

Contents of the Presentation

Introduction

Applications of Fuzzy Control

Fuzzy Theory- Fuzzy Sets & Fuzzy Logic Fuzzy Controller Design

Example: Inverted Pendulum Fuzzy

Controller Conclusion

Applications of Fuzzy Control

Fuzzy washing machines by Matushita

Electric Digital Image Stabilizer by Matsushita Electric Fuzzy Automatic Transmission by Nissan Cement Kiln controller by Holmblad and Ostergaad Subway train controller in Sendai

Contents of the Presentation

Introduction

Applications of Fuzzy Control

Fuzzy Theory- Fuzzy Sets & Fuzzy Logic Fuzzy Controller Design

Example: Inverted Pendulum Fuzzy

Controller Conclusion

Fuzzy theory: Crisp versus fuzzy sets

A set H is a collection of definite, distinguishable

objects that can be treated as a whole. Crisp sets are represented by characteristic functions 1 given by () = 0 as illustrated below

Crisp sets versus fuzzy sets (cont.)

Given a collection of objects U, a fuzzy set H in U is

defined as a set of ordered pairs , ( | Fuzzy sets are represented by membership functions given by that take values in the closed interval [0, 1] as illustrated in the figure below

Operations on Fuzzy sets For fuzzy sets A and B :

Union

(OR)

= max ,

Intersection (AND)

= min ,

Contents of the Presentation

Introduction

Applications of Fuzzy Control

Fuzzy Theory- Fuzzy Sets & Fuzzy Logic Fuzzy Controller Design

Example: Inverted Pendulum Fuzzy

Controller Conclusion

Fuzzy Controller Design Fuzzy control system Fuzzy IF THEN rules Fuzzification Knowledge Base Fuzzy Inference Engine Fuzzy implication Aggregation Defuzzification

Fuzzy-Control system block diagram

Fuzzy Controller block diagram

Linguistic (Fuzzy) Rules IF - THEN - Rules

IF premise THEN conclusion
IF antecedent THEN consequent Linguistic (fuzzy) variables and

Linguistic (fuzzy) values

Fuzzy IF THEN rules

Example : Room temperature control rules: Crisp Rule: -IF temperature is 40 C THEN set fanspeed to 1000rpm

-40 C is the crisp value of the crisp variable temperature -1000rpm is the crisp value of the crisp variable fanspeed Fuzzy Rule:
-IF temperature is veryhot THEN set fanspeed to veryfast -IF temperature is medhot THEN set fanspeed to medfast

veryhot and medhot are fuzzy(linguistic) values of the

fuzzy(linguistic) variable temperature veryfast and medfast are fuzzy(linguistic) values of the fuzzy(linguistic) variable fanspeed

Fuzzy IF THEN rules (cont.)

Example : Room temperature control rules: Crisp Rule: -IF error is 10 THEN set control to 100 10 is the crisp value of the crisp variable error 100 is the crisp value of the crisp variable control

Fuzzy Rule: -IF error is positivelarge THEN set control to negativelarge -IF error is positivemedium THEN set control to

negativemedium

positivelarge and positivemedium are fuzzy(linguistic)

values of the fuzzy(linguistic) variable error negativelarge and negatvemedium are fuzzy(linguistic) values of the fuzzy(linguistic) variable control

Fuzzy IF THEN rules (cont.)

Example : Room temperature control rules: Crisp Rule: -IF e is 10 THEN set u to 100 10 C is the crisp value of the crisp variable e

Fuzzy Rule: -IF e is PL THEN set u to NL -IF e is PM THEN set u to NM

PL and PM are fuzzy(linguistic) values of the

fuzzy(linguistic) variable e NL and NM are fuzzy(linguistic) values of the fuzzy(linguistic) variable u

Click on the cart to see the pendulum dynamics

Inverted Pendulum (cont.)

Inverted Pendulum (cont.)

Inverted Pendulum (cont.)

Inverted Pendulum (cont.)

Membership functions for input e(t)

1.2 = 0.8

1.2 = 0.15

Membership functions for input de(t)

0.6 = 0.65

0.6 = 0.35

Membership functions for output u(t)

Fuzzy IF THEN rules for the Inverted Pendulum

1 or IF e is NL and de is NL THEN set u to PL 2 or IF e is NL and de is NS THEN set u to PL 3 or IF e is NL and de is ZE THEN set u to PL 4 or IF e is NL and de is PS THEN set u to PS 5 or IF e is NL and de is PL THEN set u to ZE 6 or IF e is NS and de is NL THEN set u to PL 7 or IF e is NS and de is NS THEN set u to PL 8 or IF e is NS and de is ZE THEN set u to PS 9 or IF e is NS and de is PL THEN set u to ZE 10 or IF e is NS and de is PS THEN set u to NS

Fuzzy IF THEN rules for the Inverted Pendulum

11 or IF e is ZE and de is NL THEN set u to PL

12 or IF e is ZE and de is NS THEN set u to PS

13 or IF e is ZE and de is ZE THEN set u to ZE 14 or IF e is ZE and de is PS THEN set u to NS

15 or IF e is ZE and de is PL THEN set u to NL

16 or IF e is PS and de is NL THEN set u to PS 17 or IF e is PS and de is NS THEN set u to ZE

18 or IF e is PS and de is ZE THEN set u to NS

19 or IF e is PS and de is PS THEN set u to NL 20 or IF e is PS and de is PL THEN set u to NL

Fuzzy IF THEN rules for the Inverted Pendulum

21 or IF e is PL and de is NL THEN set u to ZE
22 or IF e is PL and de is NS THEN set u to NS 23 or IF e is PL and de is ZE THEN set u to NL

24 or IF e is PL and de is PS THEN set u to NL

25 or IF e is PL and de is PL THEN set u to NL

Table Based representation of Rules

Control u Angle-error
NL Angleerror derivative NL NS ZE PS PL PL PL PL PS ZE NS PL PL PS ZE NS

e
ZE PL PS ZE NS NL PS PS ZE NS NL NL PL ZE NS NL NL NL

de

Fuzzy Rules Activation

Fuzzy implication for Rule 17

= = =

= min (min 0.8,0.65 , = min (0.65,

Fuzzy implication for rule 18

= = =
= min (min 0.8,0.35 , = min (0.35,

Fuzzy implication for Rule 22

= = =
= min (min 0.15,0.65 , = min (0.15, )

Fuzzy implication for Rule 23

= = =
= min (min 0.15,0.35 , = min (0.15, )

Aggregate implied output membership functions to a single fuzzy membership function

Aggregate Output Membership Function

()

Defuzzification
Defuzzified (Crisp) Output, u(t)= -2.83
= 0.15 10 + 0.35 5 + 0.65 0 0.15 + 0.35 + 0.65

State-space Inverted Pendulum model

. = . . . + () .

2 =
3 4 =

The Simulink model of the Inverted Pendulum Fuzzy and PID control system

Fuzzy and PID Controller Responses

PID FUZZY
Fuzzy controller gain: output gain =100 kp=13 kd=0.1

Fuzzy and PID Controller Disturbance rejection

PID

angle (rad)

FUZZY
Fuzzy controller gain: output gain =100 kp=10 kd=0.2

time (s)

Conclusion
It has been demonstrated that the performance of a

fuzzy controller can be as good as that of other conventional controllers In some cases the fuzzy controller may perform better than other conventional controllers

Thanks to all

References
[1] K. M. Passino, S. Yurkovich. Fuzzy Control. Addison Wesley Longman, Menlo Park, California, 1998. [2] J. Jantzen, Foundations of Fuzzy Control. John Wiley & Sons, West Sussex, England,2007. [3] S. N. Sivanandam, S. Sumathi and S. N. Deepa, Introduction to FuzzyLogic using MATLAB. Springer, Berlin, 2007. [4] L. Wang, A Course in Fuzzy Systems and Control. Prentice-Hall.