Analysis on Random Fuzzy Queueing Systems with Finite Capacity

Yufu Ning1 , Ruiqing Zhao2∗,

1
Department of Computer Science and Technology, Dezhou University, Dezhou 253023, China
2
Institute of Systems Engineering, Tianjin University, Tianjin 300072, China

Abstract More precisely, Li and Lee [13] proposed a general ap-

proach for analyzing fuzzy queueing systems. However,
This paper discusses random fuzzy queueing systems with as commented by Negi and Lee [15], this approach is very
finite capacity, where the interarrival times and service complicated and generally unsuitable for computational
times are characterized as random fuzzy variables. Fuzzy purposes. Negi and Lee [15] proposed an approach by us-
simulation techniques are designed to estimate the mem- ing the α-level sets and two-variable simulation to analyze
bership degree, the expected value of system length, and fuzzy queueing systems. Kao et al [6] adopted paramet-
the credibility measure that the system length does not ex- ric programming to construct the membership functions
ceed a predetermined level. Furthermore, the rough fig- of the performance measures for fuzzy queueing systems,
ures of the membership function and credibility distribu- and this approach was successfully applied to four sim-
tion function of the system length can be obtained. Fi- ple fuzzy queueing systems with one or two fuzzy vari-
nally, an example is given to illustrate the effectiveness of ables. Chen [2] developed a method that could construct
the presented techniques. the membership functions of the performance measures in
Keywords: Queueing systems; Fuzzy variable; Random finite-capacity queueing systems with fuzzified exponen-
fuzzy variable; Fuzzy simulation; Credibility distribution tial arrival rate and service rate.

However, the possibility measure is not self-dual

1 Introduction while a self-dual measure is very necessary in both the-
ory and practical applications. This property of possibility
The finite-capacity queueing systems have been widely
measure will result in the phenomenon that an event may
studied by many researchers such as Gouweleeuw and Ti-
not occur even if its possibility is 1. This is the greatest
jms [4], Bretthauer and Cote [1], Wagner [16], and Gross
shortcoming for possibility measure since it is very hard
and Harris [5]. In the traditional queueing theory, the in-
for a decision maker to make decision based on possibility
terarrival times and service times are characterized as ran-
measure. In order to overcome the drawback of possibility
dom variables. It means that statistical data are needed to
measure, Liu and Liu [12] proposed the concept of cred-
determine the distribution functions of interarrival times
ibility measure of a fuzzy event, and introduced the con-
and service times. However, in many practical queueing
cept of expected value of fuzzy variable based on the cred-
systems, it is difficult or impossible to obtain the statisti-
ibility measure. Liu [8] presented the concept of credibil-
cal data needed. In fact, in modelling queueing systems,
ity distribution of a fuzzy variable. Furthermore, to handle
one would describe the interarrival times and service times
the phenomena combined with randomness and fuzziness,
by linguistic terms such as very fast, fast, moderate or
Liu [8, 9] introduced the concept of random fuzzy vari-
slow rather than by random variables. Thus, fuzzy queue-
ables. This paper discusses random fuzzy queueing sys-
ing systems are much more realistic than the traditional
tems with finite capacity, where the arrival rate and service
queueing systems [2].
rate are fuzzy variables with arbitrary membership func-
To handle fuzzy phenomena, Zadeh [17] initiated the tions, then presents fuzzy simulation techniques to ana-
concept of fuzzy set via membership function. In order lyze the queueing system performance.
to measure a fuzzy event, Zadeh [18] proposed the con-
cepts of possibility measure and possibility space, which
have been well developed by many researchers such as The rest of this paper is organized as follows. Some
Nahmias [14], Dubois and Prade [3], Klir [7], Liu [8–11], concepts and results on fuzzy variables and random fuzzy
and Liu and Liu [12]. Nowadays, the issue of applying variables are introduced in Section 2. The queueing sys-
the possibility theory to queueing systems with finite ca- tems with finite capacity containing random fuzzy vari-
pacity has attracted a lot of researchers such as Li and ables are discussed in Section 3. The fuzzy simulation
Lee [13], Negi and Lee [15], Kao et al [6], and Chen [2]. techniques are designed to analyze the queueing system
performance in Section 4. Finally, an example is given to
∗ Correspondence Author: zhao@tju.edu.cn illustrate the effectiveness of the presented techniques.

The 9th International Conference on Electronic Business, Macau, November 30 - December 4, 2009
Analysis on Random Fuzzy Queueing Systems with Finite Capacity 1097

2 Fuzzy variables and random fuzzy vari- Remark 2 (Liu and Liu [12]) The credibility measure is
ables self-dual, i.e.,
Cr{A} + Cr{Ac } = 1.
In this section, we review some basic concepts and results
on fuzzy variables and random fuzzy variables, which are Example 2 Let ξ = (1, 2, 3) be a triangular fuzzy vari-
used throughout the paper. able. It is easy to obtain that Pos{ξ ≤ 2} = 1, Nec{ξ ≤
2} = 0, and Cr{ξ ≤ 2} = 0.5, while Pos{ξ > 2} = 1,
Definition 1 (Zadeh [18]) Let Θ be a nonempty set, P(Θ) Nec{ξ > 2} = 0, and Cr{ξ > 2} = 0.5.
the power set of Θ, and Pos a function from P(Θ) to the
set of real numbers. Then Pos is called a possibility mea- Proposition 1 (Liu [10]) Suppose that (Θi , P(Θi ), Posi )
sure if it satisfies the following three axioms: are possibility spaces, i = 1, 2, · · · , n. Let Θ = Θ1 ×
1) Pos{Θ} = 1; Θ2 × · · · × Θn and Pos = Pos1 ∧ Pos2 ∧ · · · ∧ Posn .
2) Pos{∅} Then the set function Pos is a possibility measure, and
S = 0;
3) Pos{ i Ai } = supi Pos{Ai }, for any set Ai in P(Θ). (Θ, P(Θ), Pos) is a possibility space. Furthermore,
Furthermore, the triplet (Θ, P(Θ), Pos) is called a possi- (Θ, P(Θ), Pos) is called the product possibility space.
bility space.
Definition 6 (Liu and Liu [12]) Let ξ be a fuzzy vari-
Definition 2 (Nahmias [14]) A fuzzy variable is defined able on the possibility space (Θ, P(Θ), Pos). Then the
as a function from the possibility space (Θ, P(Θ), Pos) to expected value E[ξ] is defined as
the set of real numbers. Z +∞ Z 0
E[ξ] = Cr{ξ ≥ r}dr − Cr{ξ ≤ r}dr (1)
Definition 3 Let ξ be a fuzzy variable on the possibility 0 −∞
space (Θ, P(Θ), Pos). Then its membership function is provided that at least one of the two integrals is finite. Es-
derived from the possibility measure by pecially, if ξ is a nonnegative fuzzy variable, then
 Z +∞
µ(x) = Pos θ ∈ Θ ξ(θ) = x , x ∈ <.
E[ξ] = Cr{ξ ≥ r}dr. (2)
Example 1 Let ξ = (a, b, c) be a triangular fuzzy vari- 0

able with a < b < c. Then its membership function is Example 3 (Liu and Liu [12]) The triangular fuzzy vari-
able ξ = (a, b, c) has an expected value
x−a

 , if a ≤ x ≤ b 1
 b−a


 E[ξ] = (a + 2b + c).
4


µ(x) = x−c
, if b < x ≤ c Definition 7 (Liu [8]) The credibility distribution Φ:

 b−c
R → [0,1] of a fuzzy variable ξ is defined by




0, otherwise.
 
Φ(x) = Cr θ ∈ Θ ξ(θ) ≤ x .
Definition 4 (Zadeh [19]) Let (Θ, P(Θ), Pos) be a pos- That is, Φ(x) is the credibility that the fuzzy variable ξ
sibility space, and A a set in P(Θ). Then the necessity takes a value less than or equal to x.
measure of A is defined by
Definition 8 (Liu [8]) A random fuzzy variable is defined
Nec{A} = 1 − Pos{Ac }, as a function from the possibility space (Θ, P(Θ), Pos) to
the set F of random variables.
where Ac is the complementary set of A.
Example 4 In a queueing system, let ξ be the interarrival
Remark 1 (Liu and Liu [12]) It is obvious that neither
time of customer arrivals. Usually, the probability distri-
the possibility measure nor the necessity measure is self-
bution of ξ is assumed to be known completely except for
dual, i.e., the possibility (necessity) of a fuzzy event plus
the values of one or more parameters. For example, the in-
the possibility (necessity) of its opposite event is not equal
terarrival time ξ might be known as an exponentially dis-
to 1.
tributed random variable with an unknown mean λ, whose
Based on possibility measure and necessity measure, probability density function is
Liu and Liu [12] proposed the concept of credibility mea- (
λe−λx , if 0 ≤ x < +∞
sure. φ(x) =
0, otherwise.
Definition 5 (Liu and Liu [12]) Let (Θ, P(Θ), Pos) be a
possibility space, and A a set in P(Θ). Then the credibil- If λ is characterized as a fuzzy variable defined on the
ity measure of A is defined by possibility space (Θ, P(Θ), Pos), then ξ is just a random
fuzzy variable denoted by
1
Cr{A} = (Pos{A} + Nec{A}). ξ(λ(θ)) ∼ EX P(λ(θ)).
2

The 9th International Conference on Electronic Business, Macau, November 30 - December 4, 2009
1098 Yufu Ning, Ruiqing Zhao

Remark 3 (Liu [8]) Roughly speaking, if Θ consists of denote them. Such a case is very similar to the case of
one single element only, then the random fuzzy variable using a deterministic value instead of a random variable
degenerates to a random variable. If F is a collection in stochastic environments. Therefore, it is much more re-
of real numbers (rather than random variables), then the alistic that the arrival rate and service rate are charac-
random fuzzy variable degenerates to a fuzzy variable. terized as fuzzy variables when they can not be obtained
exactly.
3 Random fuzzy queueing systems with fi-
nite capacity 4 Analysis on random fuzzy queueing sys-
tems with finite capacity
Consider a stochastic queueing system with finite capac-
ity M/M/1/F CF S/K/∞, in which there is one server. There are many indices to measure the queueing system
Customers arrive at the single server as a Poisson pro- performance, such as system length, queue length, wait-
cess with an arrival rate λ, and all service times are in- ing time, etc. Without loss of generality, we take system
dependent and identically distributed exponential random length as the performance measure of queueing systems.
variables with service rate v. The queue discipline is first In stochastic environments, the expected value of system
come, first served (F CF S). There is a limit K placed on length of the queueing system M/M/1/F CF S/K/∞ is
the number allowed in the system at any time, and the size  
λ KλK+1 − v(K + 1)λK + v K+1
of source population is infinite. L(λ, v) = , (3)
In statistics, the values of λ and v are estimated by the (λ − v)(λK+1 − v K+1 )
interval estimate or point estimate method. However, in where λ < v (see [5]).
many practical queueing systems, the data needed in these In the random fuzzy queueing system
methods can not be obtained. Therefore, it is very diffi- RF/RF/1/F CF S/K/∞, λ and v are character-
cult or impossible to obtain the exact values of λ and v. In ized as fuzzy variables on the possibility spaces
such a case, it is reasonable that λ and v are characterized (Θ1 , P(Θ1 ), Pos1 ) and (Θ2 , P(Θ2 ), Pos2 ), respec-
as fuzzy variables according to experts’ experiences. By tively. Therefore, L(λ, v) is a fuzzy variable defined
Definition 8, the interarrival times and service times are all on the product possibility space (Θ, P(Θ), Pos), where
random fuzzy variables. Such queueing systems are called Θ = Θ1 × Θ2 and Pos = Pos1 ∧ Pos2 . Apparently,
random fuzzy queueing systems with finite capacity, which L(λ, v) is a nonnegative fuzzy variable. By Definition 6,
is denoted by RF/RF/1/F CF S/K/∞, where RF de- the expected value of L(λ, v) can be written as
notes that interarrival times and service times are random
fuzzy variables. The following example explains the con- Z +∞
cept of random fuzzy queueing systems with finite capac- E[L(λ, v)] = Cr{L(λ, v) ≥ r}dr. (4)
ity. 0

The credibility distribution Φ: R → [0,1] of the sys-

Example 5 There is a new petrol station with a single
tem length L(λ, v), by Definition 7, is
pump and finite space for no more than 10 cars (9 wait-
ing, 1 being served ), i.e., K = 10. Cars arrive at the Φ(x) = Cr {L(λ, v) ≤ x} .
petrol station according to a Poisson process, and ser-
vice time has an exponential distribution. Due to lack of That is, Φ(x) is the credibility measure that the system
historical data on arrivals and service time, the arrival length L(λ, v) does not exceed x.
rate λ and service rate v can not be determined exactly. In many real queueing systems, it is very hard to com-
In such a case, by some relevant experts’ experiences, λ pute the membership degree µL (x), the expected value
and v are characterized as the fuzzy
 variables with the E[L(λ, v)] and the credibility measure Cr{L(λ, v) ≤ x}.
membership functions µ1 (x) = 1 − (x − 1)2 ∨ 0 and Therefore, it is necessary to design fuzzy simulation tech-
2
µ2 (x) = e−(x−2) , respectively. Then the queueing sys- niques to estimate these values. For more details on fuzzy
tem is just a random fuzzy queueing system with finite ca- simulation techniques, see [12].
pacity.
4.1 Fuzzy simulation for the membership degree
Remark 4 If λ and v degenerate to crisp num- µL (x)
bers, then the random fuzzy queueing system
RF/RF/1/F CF S/K/∞ is just the traditional queue- The fuzzy simulation technique for estimating the mem-
ing system M/M/1/F CF S/K/∞. bership degree µL (x) is described as follows.

1) Set k = 1.
Remark 5 If we ignore the fuzziness of the arrival rate
and service rate when they can not be obtained exactly, 2) Let xk = kKM , and µk = 0, where M is a sufficiently
it is reasonable for us to employ their expected values to large integer.

The 9th International Conference on Electronic Business, Macau, November 30 - December 4, 2009
Analysis on Random Fuzzy Queueing Systems with Finite Capacity 1099

3) Uniformly generate a number λ1 from the support of 1) Set e = 0.

λ, then calculate the membership degree of λ1 ac-
cording to the membership function of λ, written as 2) Uniformly generate (θk1 , θk2 ) from Θ1 × Θ2 such
v1 . that pk ≥ ε for k = 1, 2, · · · , N , where pk =
Pos{θk1 } ∧ Pos{θk2 }, ε is a sufficiently small posi-
4) Replace L(λ, v), λ and v in Eq. (3) with xk , λ1 and tive number, and N is a sufficiently large integer.
y, respectively, then Eq. (3) becomes the following
polynomial equation 3) Set
K+2 K+1
xk y − (λ1 + xk λ1 )y
a = L(λ(θ11 ), v(θ12 )) ∧ · · · ∧ L(λ(θN 1 ), v(θN 2 )),
+ (K + 1 − xk )λK+1
1 y (5)
+ (xk − K)λK+2
1 = 0. b = L(λ(θ11 ), v(θ12 )) ∨ · · · ∨ L(λ(θN 1 ), v(θN 2 )).

Solve Eq. (5) by the dichotomy method, and calcu- 4) Uniformly generate r from [a, b]. Set e = e +
late the membership degree of the solution according Cr{L(λ, v) ≥ r}, where
to the membership function of v, written as v2 .

5) Set temp = v1 ∧ v2 .  v) ≥ r}
Cr{L(λ,
1 
= max pk L(λ(θk1 ), v(θk2 )) ≥ r
6) If µk < temp, then set µk = temp. 2 1≤k≤N

7) Repeat Steps 3 to 6 N times, where N is a suffi-

ciently large integer. 

+ min1≤k≤N 1 − pk L(λ(θk1 ), v(θk2 )) < r .
8) Set k = k+1. If k ≤ M , return to Step 2. Otherwise,
output (xi , µi ), where i = 1, 2, · · · , M . 5) Repeat the fourth step for N times.
Example 6 Let us continue to discuss the random fuzzy
6) Compute E = a ∨ 0 + b ∧ 0 + e · (b − a)/N , then
queueing system in Example 5. It is easy to obtain that
output E.
10λ12 − 11vλ11 + λv 11
L(λ, v) = ,
(λ − v)(λ11 − v 11 ) Example 7 Let us continue to discuss the random fuzzy
queueing system in Example 5. After 5000 cycles, the es-
where λ and v are the fuzzy variables with the member-
 timate of E[L(λ, v)] is reported by fuzzy simulation tech-
ship functions µ1 (x) = 1 − (x − 1)2 ∨ 0 and µ2 (x) =
2 nique as 3.5201. The variations of the estimate with dif-
e−(x−2) , respectively. The rough figure of the member- ferent numbers of cycles are shown in Figure 2, where the
ship function of L(λ, v) is shown in Figure 1. straight line represents the real value of E[L(λ, v)], and
the curve represents the variations of the estimates with
different numbers of cycles. It is easy to see that the esti-
µL.(x) mates approach the real value when the number of cycles
...... is larger than 2000.
....
1 .......................................................................................................................................................................
... .......... ........
0.75 ..... ........... ......
.....
... ... .....
0.5 .. . .
. .. .... E[L(λ, v)]
.... .. .... .......
. ...
0.25 ... .
. . ... ....
....................................................................................................................................................... 4 ....
0 .. x ....................................................................................................................................................
5 10 .............. ..
3 ...................
...
..
Figure 1: Rough Shape of µL (x) in Example 6 2 .....
...
..
1 ................................................................................................................................................. N
1000 2000 3000 4000 5000
4.2 Fuzzy simulation for the expected value
E[L(λ, v)] Figure 2: Variations of the Estimates of E[L(λ, v)] in Ex-
The fuzzy simulation technique for estimating E[L(λ, v)] ample 7
is summarized as follows.

The 9th International Conference on Electronic Business, Macau, November 30 - December 4, 2009
1100 Yufu Ning, Ruiqing Zhao

4.3 Fuzzy simulation for the credibility measure

Cr{L(λ, v) ≤ x} Φ(x)
.......
...
The fuzzy simulation technique for estimating .
1 .............................................................................................................................................
Cr{L(λ, v) ≤ x} is summarized as follows.
0.75 .....
.. ............
... .................. ...... ......................
.........................
1) Uniformly generate (θk1 , θk2 ) from Θ1 × Θ2 such 0.5 ..... ......................
that pk ≥ ε for k = 1, 2, · · · , N , where pk = .....
0.25 ........
Pos{θk1 } ∧ Pos{θk2 }, ε is a sufficiently small posi- ..
tive number, and N is a sufficiently large integer. 0 ................................................................................................................................................ x
5 10
2) Compute
 Figure 4: The Credibility Distribution Function Φ(x) in
1  Example 8
M= max pk L(λ(θk1 ), v(θk2 )) ≤ x
2 1≤k≤N

+ min 1 − pk L(λ(θk1 ), v(θk2 ) > x ,
1≤k≤N variable λ with the membership function

then output M.  1 − (x − 1)2 , if 0 ≤ x ≤ 2


Example 8 Let us continue to discuss the random fuzzy µ1 (x) = 1 − (x − 3)2 , if 2 < x ≤ 4

queueing system in Example 5. We use the fuzzy simula- 

0, otherwise,
tion technique proposed to estimate the credibility mea-
sure Cr{L(λ, v) ≤ 5}. After 5000 cycles, the estimate of the service rate be the triangular fuzzy variable v =
Cr{L(λ, v) ≤ 5} is 0.6305. The variations of the esti- (3, 5, 7), and the system capacity K = 100.
mates with different numbers of cycles are shown in Fig- It is easy to obtain that
ure 3, where the straight line represents the real value of
Cr{L(λ, v) ≤ 5}, and the curve represents the variations
of the estimates. It is easy to see that the estimates ap- 100λ102 − 101vλ101 + λv 101
L(λ, v) = .
proach the real value when the number of cycles is larger (λ − v)(λ101 − v 101 )
than 2000.
Apparently, since the membership function of arrival
rate λ is multimodal, the methods presented in [2] and [6]
Cr{L(λ,. v) ≤ 5} can not apply to this queueing system. The fuzzy simula-
....... tion techniques proposed in Section 4 can be used to es-
...
.
1 ..... timate the membership degree µL (x), the expected value
... E[L(λ, v)], and the credibility measure Cr{L(λ, v) ≤ x}.
0.75 ................................................................................................................................. ..........
... . The rough figure of the membership function of
0.5 .... L(λ, v) is shown in Figure 5.
...
0.25 ....
..
0 ................................................................................................................................................ N
1000 2000 3000 4000 5000 µL (x)
.......
....
Figure 3: Variations of the Estimates of Cr{L(λ, v) ≤ 5} 1 ..........................................................................................................................................
.....
in Example 8 0.75 .......... .....
.... .....
0.5 ..... .......................................................................
... .......................................
By using the fuzzy simulation technique proposed to .........
0.25 ..... ...
estimate Cr{L(λ, v) ≤ x}, where x ∈ (0, 10), the rough ....................................................................................................................................................
0 x
figure of the credibility distribution function Φ(x) of the 50 100
system length L(λ, v) can be seen, which is depicted in
Figure 4 (5000 cycles in fuzzy simulation for every credi-
bility measure Cr{L(λ, v) ≤ x}). Figure 5: Rough Figure of µL (x) in Example 9

After 5000 cycles, the estimate of E[L(λ, v)] is re-

ported by fuzzy simulation technique as 49.8414. The
5 Numerical example variations of the estimate of E[L(λ, v)] with different
numbers of cycles are shown in Figure 6, where the
Example 9 Consider a random fuzzy queueing system straight line represents the real value of E[L(λ, v)], and
RF/RF/1/FCFS/K/∞. Let the arrival rate be the fuzzy the curve represents the variations of the estimate of

The 9th International Conference on Electronic Business, Macau, November 30 - December 4, 2009
Analysis on Random Fuzzy Queueing Systems with Finite Capacity 1101

E[L(λ, v)]. It is easy to see that the estimates approach

Φ(x)
the real value when the number of cycles is larger than .......
2000. ...
.
1 ................................................................................................................................................
.. ...............................................................................
E[L(λ, v)] 0.75 ..... ......................
... ..
....... 0.5 ......
.... ...
20 ..................................................................................................................................................................... 0.25 .....
... . ................................ . ..
................. .... .. 0 ................................................................................................................................................ x
........ .. 50 100
10 ............
........
...... Figure 8: The Credibility Distribution Function Φ(x) in
.
.................................................................................................................................................. N Example 9
1000 2000 3000 4000 5000

Figure 6: Variations of the Estimates of E[L(λ, v)] in Ex- that the system length was less than or equal to a prede-
ample 9 termined level. Finally, an example was given to illustrate
the effectiveness of the proposed techniques. The methods
After 5000 cycles, the estimate of Cr{L(λ, v) ≤ 50} can be easily employed to analyze the other performance
is 0.5001. The variations of the estimate with differ- measures such as waiting time, queue length, and so on.
ent numbers of cycles are shown in Figure 7, where the Similarly, the proposed methods can be employed to ana-
straight line represents the real value of the credibility lyze the other random fuzzy queueing systems.
measure Cr{L(λ, v) ≤ 50}, and the curve represents the
variations of the estimates obtained by the fuzzy simula- Acknowledgments
tion technique. It is easy to see that the estimates ap-
proach the real value when the number of cycles is larger This work was supported by the National Natural Science
than 2000. Foundation of China Grant No. 70571056 and China Post-
By using fuzzy simulation technique to estimate doctoral Science Foundation No. 20060400704.
Cr{L(λ, v) ≤ x}, where x ∈ (0, 100), the rough figure
of the credibility distribution function Φ(x) of the system
length L(λ, v) can be seen, which is depicted in Figure 8 References
(5000 cycles in fuzzy simulation for every credibility mea-
sure Cr{L(λ, v) ≤ x}). [1] Bretthauer, K. M., Cote, M. J.: Nonlinear programming for mul-
tiperiod capacity planning in a manufacturing system. European
Journal of Operational Research. 96 (1997) 167-179
[2] Chen, S. P.: Parametric nonlinear programming for analyzing
fuzzy queues with finite capacity. European Journal of Operational
Cr{L(λ, .v) ≤ 50} Research. 157 (2004) 429-438
...... [3] Dubois, D., Prade, H.: Possibility Theory. Plenum, New York
.... (1988)
1 ............ [4] Gouweleeuw, F. N., Tijms, H. C.: A simple heuristic for buffer
.......................................................................................................................................................
0.75 ..... design in finite-capacity queues. European Journal of Operational
... Research. 88 (1996) 592-598
0.5 .... [5] Gross, D., Harris, C. M.: Fundamentals of Queueing Theory. Wi-
.. ley, New York (1998)
0.25 .... [6] Kao, C., Li, C. C., Chen, S. P.: Parametric programming to the
..
0 ................................................................................................................................................ N analysis of fuzzy queues. Fuzzy Sets and Systems. 107 (1999) 93-
1000 2000 3000 4000 5000 100
[7] Klir, G. J.: On fuzzy-set interpretation of possibility theory. Fuzzy
Sets and Systems. 108 (1999) 263-273
Figure 7: Variations of the Estimates of Cr{L(λ, v) ≤ [8] Liu, B.: Theory and Practice of Uncertain Programming. 2nd ed.,
50} in Example 9 Springer-Verlag, Berlin (2009)
[9] Liu, B.: Uncertainty Theory. 2nd ed., Springer-Verlag, Berlin
(2007)
[10] Liu, B.: Toward fuzzy optimization without mathematical ambigu-
ity. Fuzzy Optimization and Decision Making. 1 (2002) 43-63
6 Conclusion [11] Liu, B.: A survey of credibility theory. Fuzzy Optimization and
Decision Making, 5 (2006) 387-408
This paper discussed random fuzzy queueing systems [12] Liu, B., Liu, Y. K.: Expected value of fuzzy variable and fuzzy
expected value models. IEEE Transactions on Fuzzy Systems. 10
with finite capacity. Fuzzy simulation techniques were (2002) 445-450
designed to estimate the membership degree, the expected [13] Li, R. J., Lee, E. S.: Analysis of fuzzy queues. Computers and
value of the system length, and the credibility measure Mathematical Applications. 17 (1989) 1143-1147

The 9th International Conference on Electronic Business, Macau, November 30 - December 4, 2009
1102 Yufu Ning, Ruiqing Zhao

[14] Nahmias, S.: Fuzzy variable. Fuzzy Sets and Systems. 1 (1978) [17] Zadeh, L. A.: Fuzzy sets. Information and Control. 8 (1965) 338-
97-101 353
[15] Negi, D. S., Lee, E. S.: Analysis and simulation of fuzzy queues. [18] Zadeh, L. A.: Fuzzy sets as a basis for a theory of possibility. Fuzzy
Fuzzy Sets and Systems. 46 (1992) 321-330 Sets and Systems. 1 (1978) 3-28
[16] Wagner, D.: Waiting times of a finite-capacity multi-server model [19] Zadeh, L. A.: A theory of approximate reasoning. In: J. Hayes,
with non-preemptive priorities. European Journal of Operational D. Michie and R. M. Thrall, eds., Mathematical Frontiers of the
Research. 102 (1997) 227-241 Social and Policy Sciences, Westview Press, Boulder, Cororado.
(1979) 69-129

The 9th International Conference on Electronic Business, Macau, November 30 - December 4, 2009