Fuzzy Logic: Human-Like Decision Making: Lecture 9 of Artificial Intelligence

Lecture 9 of Artificial Intelligence
Fuzzy Logic:
Human-like decision making
Produced by Qiangfu Zhao (Since 2008), All rights reserved © AI Lec09/1

Topics of this lecture
• Definition of fuzzy set • ファジィ集合の定義
• Membership function • メンバシップ関数
• Notation of fuzzy set • ファジィ集合の記述方法
• Operations of fuzzy set • ファジィ集合の演算
• Fuzzy number and • ファジィ数とその演算
operations • 拡張原理
• Extension principle • ファジィルール
• Fuzzy rules • 非ファジィ化
• De-fuzzification • ファジィ制御
• Fuzzy control

History of fuzzy logic
• The opposite word (antonym) of

fuzzy is crisp. Fuzzy means un-clear
or ambiguous, and crisp means
clear, clean, and sharp.
• Fuzzy logic was proposed by Zadeh
in 1965, and was applied in steam
engine control by Mamdani in 1974.
• Fuzzy logic was made practically
useful in Japan in the 1990s.

Conventional set vs. fuzzy set
• X: universe of discourse. Set A is defined by
A={ x |m(x)=T ⋀ x in X}
where m(x) is a membership function.
– For conventional set, the range of m(x) is {T, F}
– For fuzzy set, the range of m(x) is [0,1]. So, the above definition
cannot be used for fuzzy set. We cannot say clearly if x is in A or
not when 0<m(x) < 1.
Examples of fuzzy set

 young people, old people, kind people
 temperature is hot, just good, a little bit cold

Description of fuzzy set
• Membership function of a fuzzy set A: mA: X[0,1]
• If the universe of discourse X={x1,x2,…,xN}, A is
described as follows:
A = mA(x1)/x1+ mA(x2)/x2+…+ mA(xN)/xN

= SmA(xi)/xi
where / is a separator, and ＋ is the logic OR.
• If X is a continuous space, integral is used instead of
the summation.

Operations of fuzzy sets
Meaning Set notation Logic notation Membership function
Equivalence
A B m A ( x)  mB ( x) m A ( x)  mB ( x)
Implication
A B m A ( x)  mB ( x) m A ( x)  mB ( x)
Complement
(negation) A m A (x) m A ( x)  1  m A ( x)
Union, OR
(disjunction) A B m A ( x)  mB ( x) m AB ( x)  max{m A ( x), mB ( x)}
Intersection, AND
(conjunction) A B m A ( x)  mB ( x) m AB ( x)  min{ m A ( x), mB ( x)}

Examples
Upper-left: AND
Lower-left: OR
Upper-right: Negation

Example 5.1 p. 89
• The universe of discourse:
X={Masahiro, Tsuyoshi, Takuya, Saburo, Masami}
• Fuzzy sets A=“young persons” and B=“Tall persons” are

defined by
– A = 0.4/Masahiro+0.6/Tsuyoshi+0.8/Takuya+1.0/Saburo+0.9/Masami
– B = 0.3/Masahiro+0.5/Tsuyoshi+0.9/Takuya+0.6/Saburo+0.9/Masami
• The OR and AND of A and B are as follows:

– A∪B=0.4/Masahiro+0.6/Tsuyoshi+0.9/Takuya+1.0/Saburo+1.0/Masami
– A∩B =0.3/Masahiro+0.5/Tsuyoshi+0.8/Takuya+0.6/Saburo+0.9/Masami

Fuzzy number
• A fuzzy number is a fuzzy set of (real) numbers.
• The membership function mA of a fuzzy number A
satisfies the following properties:
– There is an x, such that mA(x)=1;
– The support {x| mA(x)>0} is bounded;
– The a-cut {x| mA(x)>a} is closed.
• Examples of fuzzy number:
– The distance between Aizuwakamatsu city and Sendai is about
200 km.
– If we go from Aizuwakamatsu city to Sendai by train, it takes
about 2 hours.

Computation with fuzzy numbers
• Based on the extension principle, a binary operation ◦
between two fuzzy numbers can be calculated by
C  A  B : m AB ( y )  sup {m A ( x1 )  m B ( x2 )}
x1  x2  y
Superior (上限)
• The operation ◦ can be , , , or .
About 2 About 4 About 6
About 6 is more ambiguous!
1 2 3 4 5 6 7 8 9 10 11 12

Linguistic values of a fuzzy number
• To inference (reasoning) with fuzzy numbers, we may read a fuzzy
number using “about”, “approximately”, or “roughly”, but these
representations may not help us to understand the physical meaning.
• For example, to say a body temperature is about 39 degree, we may
not understand that this number if relatively dangerous for an adult
person. Rather, if we say the body temperature is “very high” or
“very dangerous”, we can get the meaning more easily.
Very Young Young Old Very old

1
0 10 20 30 40 50 60 70 80 90 100 Age

Fuzzy rules
Number of attributes : N
Number of rules : K
The k - th fuzzy rule :
Rk : if ( x1  Fk 1  x2  Fk 2    x N  FkN )then ( y  Bk )
, k  1,2,, K ; j  1,2,, N , are

where Bk , Fkj 　
linguistic values.
Example of fuzzy rule
Nurui On
T S
10 20 30 40 50 1 2 3 4 5
If (Temperature = Nurui)Then (Oidaki = On)

Pattern matching for fuzzy rules
For a given input datum x  ( x1 , x2 ,, x N ), the similarity
between x and the condition of the k - th rule is
S ( x, Rk )  m Fk 1 ( x1 )  m Fk 2 ( x2 )    m FkN ( x N )
where  is defined as the " min".
The similarity can also be considered

the degree of matching.
Fuzzy inference
For a given input datum x  ( x1 , x2 ,, x N ), we can make
a fuzzy decision as follows :
Step 1 : Find the membership function of the result B* ,
which is also a fuzzy number using the following equation :
mB ( y )  mR ( x )  mR ( x )    mR ( x )
*
1 2 K
where  is " max" , and m Rk ( x )  S ( Rk )  m Bk ( y ).

Fuzzy inference
Step 2: The final output is calculated as follows:
b *

 Y
y  m B* ( y )dy
m
Y B*
( y )dy
• This is the gravity center of the fuzzy number B*.

• We may also use median or the maximum value.
• The process for finding a concrete number from a fuzzy
number is called de-fuzzification.

Example 5.3 pp. 93-96
• According to Japanese
Road laws, road around the
corner must be constructed
based on the relation
between the curvature and
the speed limitation.
• For example, if the speed
limit is 50 km/h, the radius of
the curve must be 100m or
more (see Table 5.4 in p.
94).

Fuzzy rules for speed control
• R1: If(v=Normal∧ Speed (v) Curvature Break (b)
r=Sharp curve) (radius r)
Normal Slow curve As is
Then(B=Weak)
• R2: If(v=A little fast∧ Normal Normal As is
r=Sharp curve) Normal Sharp curve Weak
Then(B=Weak)
A little fast Slow curve As is
• R3: If(v=Fast∧
r=Normal) A little fast Normal As is
Then(B=Weak)
A little fast Sharp curve Weak
• R4: If(v=Fast∧
Fast Slow curve As is
r=Sharp curve)
Then(B=Strong) Fast Normal Weak
Fast Sharp curve Strong

Linguistic values of the fuzzy numbers

Proper speed for a given condition
Step 1: First, find the similarity (degree of matching) for
each rule as follows:
• S(R1)=min(mnormal(65), msharp(90))=min(0,0.8)=0
• S(R2)=min(ma little fast(65), msharp(90))=min(0.75,0.8)=0.75
• S(R3)=min(mfast(65), mnormal(90))=min(0.25,0.2)=0.20
• S(R4)=min(mfast(65), msharp(90))=min(0.25,0.8)=0.25
mB ( y )  max[min(0.75, mweak ( y )), min(0.25, mstrong ( y ))]

*

Proper speed for a given condition
• Step 2: Defuzzification
b *

 Y
y  m B* ( y )dy
=20.7 km/h
m
Y B*
( y )dy
• That is, if the speed is reduced from 65 to 44.3

km/h, the car can pass the curve safely.

Problems in using fuzzy control
• Fuzzy control is a kind of soft control that uses human
knowledge. The performance depends on the
membership functions of the linguistic values.
• To find the optimal membership functions, we must use
some other tools.
• For example, we may genetic algorithm to fine-tune the
parameters of the membership functions (e.g. to
determine the shape or type).
• We may also use neural networks to capture the best
membership functions through learning.

Homework for lecture 9 (1)
• Solve problem 5.4 in p. 97 in the textbook.
• That is, for the membership functions of the linguistic
values in Example 5.3, find their equations and draw the
figures.
• Submit before end of the exercise class.

Homework for lecture 9 (2)
• Based on the skeleton program, write a program to
implement a fuzzy system for speed control, and test
your program using v=70 km/h and r=90 m.
• Plot the results using “gnuplot”, and save the 3-D figure.
• The name file should be “plot.png”, and the coordinates
are as follows:
– (X, Y, Z) = (v, r, b)
• From the figure, try to describe the relation between b

and (v, r), and add your description in “summary_09.txt”.
Quizzes for lecture 9
(1) Find the compliment set of A given in Example 5.1. (3) Below is a fuzzy rule for controlling a robot based
on light sensors.
Rs : if ( x1  VeryBright  x2  VeryBright 
(2) Using the extension principle, find the membership x3  VeryDark  x4  VeryDark )
function of “fuzzy 5”, when “fuzzy 2” and “fuzzy 3”
are defined below. then( y  GoForwardQuickly)
Suppose that the minimum and maximum values of each

Fuzzy 2 sensor are 0 and 1024. Try to define the membership
Fuzzy 3 functions for the linguistic values VeryBright and
？
VeryDark.
1 2 3 4 5 6 7 8 9 10 11 12
0 512 1024

Fuzzy Logic: Human-Like Decision Making: Lecture 9 of Artificial Intelligence

Uploaded by

Copyright:

Available Formats

Fuzzy Logic: Human-Like Decision Making: Lecture 9 of Artificial Intelligence

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Fuzzy Logic: Human-Like Decision Making: Lecture 9 of Artificial Intelligence

Uploaded by

Copyright:

Available Formats

Lecture 9 of Artificial Intelligence

Produced by Qiangfu Zhao (Since 2008), All rights reserved © AI Lec09/1

Produced by Qiangfu Zhao (Since 2008), All rights reserved © AI Lec09/2

• The opposite word (antonym) of

Produced by Qiangfu Zhao (Since 2008), All rights reserved © AI Lec09/3

Examples of fuzzy set

Produced by Qiangfu Zhao (Since 2008), All rights reserved © AI Lec09/4

A = mA(x1)/x1+ mA(x2)/x2+…+ mA(xN)/xN

Produced by Qiangfu Zhao (Since 2008), All rights reserved © AI Lec09/5

Produced by Qiangfu Zhao (Since 2008), All rights reserved © AI Lec09/6

Produced by Qiangfu Zhao (Since 2008), All rights reserved © AI Lec09/7

• Fuzzy sets A=“young persons” and B=“Tall persons” are

• The OR and AND of A and B are as follows:

Produced by Qiangfu Zhao (Since 2008), All rights reserved © AI Lec09/8

Produced by Qiangfu Zhao (Since 2008), All rights reserved © AI Lec09/9

Produced by Qiangfu Zhao (Since 2008), All rights reserved © AI Lec09/10

Very Young Young Old Very old

Produced by Qiangfu Zhao (Since 2008), All rights reserved © AI Lec09/11

, k  1,2,, K ; j  1,2,, N , are

If (Temperature = Nurui)Then (Oidaki = On)

Produced by Qiangfu Zhao (Since 2008), All rights reserved © AI Lec09/13

where  is defined as the " min".

The similarity can also be considered

where  is " max" , and m Rk ( x )  S ( Rk )  m Bk ( y ).

Produced by Qiangfu Zhao (Since 2008), All rights reserved © AI Lec09/15

• This is the gravity center of the fuzzy number B*.

Produced by Qiangfu Zhao (Since 2008), All rights reserved © AI Lec09/16

Produced by Qiangfu Zhao (Since 2008), All rights reserved © AI Lec09/17

Produced by Qiangfu Zhao (Since 2008), All rights reserved © AI Lec09/18

Produced by Qiangfu Zhao (Since 2008), All rights reserved © AI Lec09/19

mB ( y )  max[min(0.75, mweak ( y )), min(0.25, mstrong ( y ))]

Produced by Qiangfu Zhao (Since 2008), All rights reserved © AI Lec09/20

• That is, if the speed is reduced from 65 to 44.3

Produced by Qiangfu Zhao (Since 2008), All rights reserved © AI Lec09/21

Produced by Qiangfu Zhao (Since 2008), All rights reserved © AI Lec09/22

Produced by Qiangfu Zhao (Since 2008), All rights reserved © AI Lec09/23

• From the figure, try to describe the relation between b

Suppose that the minimum and maximum values of each

Produced by Qiangfu Zhao (Since 2008), All rights reserved © AI Lec05/25

You might also like