Applications Domain Some Interesting Applications
Applications Domain Some Interesting Applications
Crisp Set/Subset
µA =0
µ A =1
Conventional or crisp sets are
µ B =1
binary. An element either belongs to µC =1 Subset A
the set or doesn't.
Universe (X) Subset B
{True, false}
{1, 0} Subset C
1
Fuzzy Indicators
• Examples of fuzzy measures include close, • Can you distinguish between American and
heavy, light, big, small, smart, fast, slow, French person?
hot, cold, tall and short. • Some Rules:
– If speaks English then American
– If speaks French then French
– If loves perfume then French
e.g. On a scale – If loves outdoors then American
of one to 10,
how good was 10 9 – If good cook then French
the dive? – If plays baseball then American
9 9.5
Fuzzy Indicators
• Rules may give contradictory indicators • Fuzzy ≠ Probability
{good cook, loves outdoors, speaks French}
• Probability deals with uncertainty and
• The right answer is a question of a degree of
association likelihood
• Fuzzy logic resolves these conflicting indicators • Fuzzy logic deals with ambiguity and
– Membership of the person in the French set is 0.9 vagueness
– Membership of the person in the American set is 0.1
Example #2
• Fuzzy ≠ Probability – A bottle of liquid has a probability of ½
of being rat poison and ½ of being pure
• Example #1 water.
– A second bottle’s contents, in the fuzzy
– Billy has ten toes. The set of liquids containing lots of rat
probability Billy has poison, is ½.
#1 – The meaning of ½ for the two bottles
nine toes is zero. The
clearly differs significantly and would
fuzzy membership of impact your choice should you be
Billy in the set of people dying of thirst. #2
– 50% probability means 50% chance
with nine toes, however, that the water is clean.
is nonzero. – 50% fuzzy membership means that the
(cite: Professor Marks) water has poison.
(cite: Bezdek)
2
• Crisp membership functions (µ) are either one or • The set, B, of numbers near to 2 can be
zero. represented by a membership function
• e.g. Numbers greater than 10.
µB(x)
Fuzzy Sets
• A fuzzy set, A, is said to be a subset of B if
µ A ( x) ≤ µ B ( x)
• e.g. B = far and A=very far.
• For example...
µ A ( x ) = µ B2 ( x )
Tall Tall or Short? Short
6’
Medium
5’
Short 4’ 5’ 6’ 7’
4’
Very Short
µ = [µvs , µ s , µ m , µt , µvt ]
3
Membership Function
µ Fuzzy Logic Operations
Medium Tall Very Tall
Very Short
1.0
Short Fuzzy union operation or fuzzy OR
4’ 5’ 6’ 7’
Complement operation
µ A ( x) = 1 − µ A ( x)
0 1 2 3 x
4
µ A+ B ( x) = max [ µ A ( x), µ B ( x)]
µB(x)
µA(x)
1
0 1 2 3 x 0 1 2 3 x
µA+B (x)
0 1 2 3 x
• Fuzzifier converts a crisp input into a fuzzy • Assume we want to evaluate the health of
variable. a person based on his height and weight.
• Definition of the membership functions must • The input variables are the crisp
– reflects the designer's knowledge numbers of the person’s height and
– provides smooth transition between member and weight.
nonmembers of a fuzzy set • Fuzzification is a process by which the
– simple to calculate numbers are changes into linguistic
• Typical shapes of the membership function are words
Gaussian, trapezoidal and triangular.
5
• Rules are tabulated as fuzzy words
• Rules reflect experts decisions. – Healthy (H)
• Rules are tabulated as fuzzy words – Somewhat healthy (SH)
• Rules can be grouped in subsets – Less Healthy (LH)
• Rules can be redundant – Unhealthy (U)
• Rules can be adjusted to match desired • Rule function f
results
f = {U , LH , SH , H }
f Very
Weight
Very
Slim Medium Heavy
Slim Heavy
Very
Short H SH LH U U
Height
Short SH H SH LH U
Medium LH H H LH U
Tall U SH H SH U
f = {U , LH , SH , H } Very
Tall U LH H SH LH
Membership of Height
µ
• For a given person, compute the Very Short Short Medium Tall Very Tall
membership of his/her weight and height 0.7
• Example:
– Assume that a person height is 6’ 1” 0.3
– Assume that the person’s weight is 140 lb
4’ 5’ 6’ 7’
µheight = {µVS µS µM µT µVT}
µheight = {0 0 0.7 0.3 0}
6
Membership of Weight
µ Weight
Very Very
Very Slim Slim Medium Heavy Very Heavy Slim Medium Heavy
Slim Heavy
0.8 Very
Short H SH LH U U
Height
0.2 Short SH H SH LH U
Weight Weight
Very Medium Heavy V.Heavy
0.8 0.2 Medium Heavy
Heavy 0.8 0.2 (0) (0) (0)
Very V. Short
Short H SH LH U U 0 0 0 0 0
(0)
Height
Height
Short
Short SH H SH LH U (0) 0 0 0 0 0
0.7 LH H H LH U 0.7 0.7 0.2 0 0 0
0.3 U SH H SH U 0.3 0.3 0.2 0 0 0
Very V. Tall
Tall U LH H SH LH 0 0 0 0 0
(0)
f
Weight Weight 0.7
0.8 0.2 0.8 0.2
V. Short
(0)
0 0 Very Short H SH
Short (0) 0 0 Short SH H 0.3
0.7 0.7 0.2 0.7 LH H 0.2
0.3 0.3 0.2
0.3 U SH
V. Tall
0 0
(0) Very Tall U LH
f = {U , LH , SH , H } f = {U , LH , SH , H }
f = { 0.3, 0.7, 0.2, 0.2} f = { 0.3, 0.7, 0.2, 0.2}
7
• Use the fuzzified rules to compute the • Fuzzy set with the largest membership value is
selected.
final decision. • Fuzzy decision:
• Two methods are often used. f = {U , LH , SH , H }
– Maximum Method (not often used) f = { 0.3, 0.7, 0.2, 0.2}
– Centroid • Final Decision (FD) = Less Healthy
• If two decisions have same membership max,
use the average of the two.
f FD =
∑µ D = µ U DU + µ LH DLH + ... ...
0.7 ∑µ µU + µ LH + ... ...
f f
0.7
0.75
0.3
0.2
0.25
Crisp Decision Index (D) is the centroid Fuzzy Decision Index (D)
D = 0.4429 75% in Less Healthy group
25% in Somewhat Healthy group
8
• Assume that we need to evaluate student • Assume that we need to evaluate student applicants
applicants based on their GPA and GRE based on their GPA and GRE scores.
scores. • For simplicity, let us have three categories for each
score [High (H), Medium (M), and Low(L)]
• Let us assume that the decision should be
• Let us assume that the decision should be Excellent (E),
Excellent (E), Very Good (VG), Good (G), Fair Very Good (VG), Good (G), Fair (F) or Poor (P)
(F) or Poor (P) • An expert will associate the decisions to the GPA and
• An expert will associate the decisions to the GRE score. They are then Tabulated.
GPA and GRE score. They are then
Tabulated.
Excellent = 95-100%
Very Good = 90 - 94%
Good = 80 - 89%
Fair = 70 - 79%
Poor = 0 - 69%
variable.
• Definition of the membership functions must
– reflects the designer's knowledge
– provides smooth transition between member and
nonmembers of a fuzzy set
– simple to calculate
• Typical shapes of the membership function are
Gaussian, trapezoidal and triangular. µ GRE = {µL , µM , µH }
9
µ GPA
Fn
µ GPA = {µL , µM , µH }
GRE=900
µ GRE
GPA=3.6
µ
GRE = {µL = 0.8 , µM = 0.2 , µH = 0}
µ GPA = {µL = 0 , µM = 0.6 , µH = 0.4}
10
Fn
0.6
0.4
0.2
• Converting the output fuzzy variable • Fuzzy set with the largest membership
into a unique number value is selected.
• Two defuzzifier methods are often • Fuzzy decision: Fn = {P, F, G,VG, E}
used. Fn = {0.6, 0.4, 0.2, 0.2, 0}
• Final Decision (FD) = Poor Student
– Maximum Method (not often used)
• If two decisions have same membership
– Centroid max, use the average of the two.
Fn ∑ µ fn µ f
× E + µVG ×VG + .....
FD = i = E
∑µ f
µ E + µVG + .....
i
11
F Input
Reference
u Output
Fuzzy System
Controller
Speed Regulation
CE 3.0
LN MN SN ZE SP MP LP 2.5
LN LN LN LN LN MN SN SN
2.0
MN LN LN LN MN SN ZE ZE
Position
SN LN LN MN SN ZE ZE SP 1.5
E ZE LN MN SN ZE SP MP LP 1.0
SP SN ZE ZE SP MP LP LP 0.5
MP ZE ZE SP MP LP LP LP
0.0
LP SP SP MP LP LP LP LP 0 1 2 3 4 5 6 7 8 9 10
Time
Conventional non-adaptive
Tracking (HPD) Tracking Control
• Requires mathematical models
• Nonlinear processes are linearized
3
Time
12
Conventional non-adaptive
Tracking Control
Merits of Conventional Control
Irony
• Neural and Fuzzy Control
• Stability and reliability studies • Based on intuitions and judgments.
are based on linearized models • Relatively simple, fast and adaptable
• Most systems do not behave • Can implement design objectives, difficult to
express mathematically, in linguistic or
linearly
descriptive rules.
• Most systems do drift
13
MULTI-LAYER FUZZY CONTROL
System Inputs
µE LN MN SN ZE SP MP LP
1
Execution Layer
ELC ELC ELC 0
-6 -3 -1 0 1 3 6
µCE
Supervisor Layer 1
LN MN SN ZE SP MP LP
MLFC outputs
0
-3 -2 -1 0 1 2 3
14