You are on page 1of 5

Proceedings of 2015 RAECS UIET Panjab University Chandigarh 21-22nd December 2015

Effect of utility based functions on fuzzy-AHP


based network selection in heterogenous wireless
networks

Raman Kumar Goyal Sakshi Kaushal


Computer Science and Engineering, Computer Science and Engineering,
University Institute of Engineering and Technology, University Institute of Engineering and Technology,
Panjab University, Chandigarh, India. Panjab University, Chandigarh, India.
raman.uiet1985@gmail.com sakshi@pu.ac.in

Abstract - With the growth in the field of wireless technology, Energy efficiency has become a critical issue while
internet can be accessed by mobile devices through different accessing a specific network as the power requirements of the
wireless interfaces. These devices experience both horizontal and applications are more than the battery lifetime. MADM
vertical handovers in heterogeneous wireless environments. (Multiple Attribute Decision Making) methods can be applied
Multiple Attribute Decision Making (MADM) methods can be
for decision making in network selection for a vertical
used for network selection in heterogeneous wireless networks.
Traditional MADM methods calculate the final ratings by handover. The attributes for a MADM based network selection
considering the normalized values of the attributes. The main can be bandwidth, delay, price etc. Fig. 1 describes the
disadvantage with the normalized values is that they do not hierarchical structure for Analytic Hierarchy Process (AHP)
consider the actual usability of the attributes for a particular for the network selection.
application. In this paper, a utility based fuzzy-AHP method is
applied for selecting a network considering the user preferences,
network conditions and energy consumption in a real time
scenario like video conferencing. Four networks (WLAN1,
WLAN2, WiMax, and UMTS) are considered for network Goal
Network Selection
selection in vertical handover scenarios. Utility value of each
attribute is calculated from the utility functions. Simple Additive
Weighing (SAW) is also analyzed using the utility values. Results
show that with the utility based MADM methods, ratings of
Bandwidth Delay Energy
networks are more suitable with respect to the requirements of Attributes
the application and unnecessary handovers can be avoided in Consumption
comparison with traditional MADM methods.

Keywords— Vertical handovers; network selection; fuzzy-


WLAN1 WLAN2 WiMax UMTS
AHP, utility functions; energy efficiancy

Alternatives
I. INTRODUCTION
With upcoming all IP based networks, all the heterogeneous Fig 1: Hierarchical structure of AHP
network technologies can be integrated through a single IP-
based core. Therefore a freely moving mobile device can
undergo continuous horizontal and vertical handovers. The MADM selection methods select a network based upon their
switching of networks in same type of networks is called attribute values but the actual usage of the attribute is ignored.
horizontal handover while switching between different types To calculate the actual usage of attributes, utility values of
of networks is called vertical handover [1]. Different network attributes can be used. In this paper, utility based fuzzy-AHP
applications require different set of attribute values like VoIP is presented for energy efficient network selection. The rest of
(Voice over IP) application requires lesser bandwidth and this paper is organized as follows. Section II presents the
lesser delay while video conferencing requires higher related work on MADM based network selection. Section III
bandwidth and very less delay and simple browsing requires describes the proposed utility based fuzzy-AHP network
moderate bandwidth and delay attribute is not of much selection method. Results are discussed in Section IV. Finally,
importance. conclusion is presented in Section V.

978-1-4673-8253-3/15/$31.00 ©2015 IEEE

Authorized licensed use limited to: National Ilan Univ.. Downloaded on March 04,2021 at 15:14:54 UTC from IEEE Xplore. Restrictions apply.
Proceedings of 2015 RAECS UIET Panjab University Chandigarh 21-22nd December 2015

Table I: Attributes of various networks


Bandwidth Delay Power
II. RELATED WORK (kbps) (ms) (W)
WLAN1 11000 50 4.5
Rao [2] has listed various MADM methods for decision WLAN2 11000 95.90 4.0
making in manufacturing process. Among these methods AHP WiMax 2000 100 3.5
[3], TOPSIS [4], PROMETHEE [5] and ELECTRE [6] are UMTS 1000 200 1.2
most commonly used methods. The main advantages of the
AHP method is that it can convert complex decision problem
into the hierarchical structure and also decision makers find A. Utility functions
easy to give pairwise comparisons of attributes used in AHP
[7].AHP and GRA (Grey Rational Analysis) based technique The utility values of QoS parameters are generally obtained
where AHP is used for selection criteria and GRA evaluates from sigmoid function. For beneficial criteria like bandwidth
heterogeneous networks is presented in [8]. Lahby et al. [9] the utility function is ( ) = (1 ( )
) . The value of
have proposed an enhanced TOPSIS method by using the a, and b in case of bandwidth is 0.003 and 2000 respectively.
Analytic Network Process to weigh the criteria and then apply For downward criteria like delay is ( ) = 1 ( ). The
the TOPSIS method. Different fuzzy based MADM methods value of a, and b in case of delay is 0.1 and 112.5 respectively.
have also been used for vertical handovers [10]. Fuzzy logic The utility value of energy consumption is equal to the
and TOPSIS method is used to reduce the number of consumption of power for communications in the network to
handovers and increase user satisfaction [11]. Xinjun Liu et which mobile device is connected [16]. For energy
al. [12] have proposed fuzzy TOPSIS based network selection consumption criteria, a linearly decreasing utility function is
for heterogeneous networks based on connection number a+bi modeled as ( ) = 1 as shown in Fig 2 [16, 17]. The
where a denotes the certainty number and bi denotes the value of slope coefficient can be obtained from maximum
uncertainty number in the fuzzy language number. Falwo et power consumed by the specified network. From the literature
al. [13] have presented TOPSIS based network solution it is reasonable to assume the maximum power consumption
considering maximum data rate, security, delay, battery power equal to 5W [18- 20].
consumption and cost. This method selects the network based
on single application or multiple applications. Lahby et al. Table II: Utility values of attributes of alternative networks
[14] have presented the novel validation approach for network
selection by considering the group weighting techniques for Bandwidth Delay Power
MADM methods. AHP method is applied to determine the (Active
weight of interclass and intra class attributes. Goyal and State)
Kaushal [15] have applied AHP method to find the effect of
WLAN1 1 1 0.10
mobility of the mobile device on the handover decision
WLAN2 1 0.84 0.20
process in heterogeneous wireless networks.
WiMax 0.50 0.78 0.30
All the above mentioned MADM methods use actual
UMTS 0.05 0 0.76
attribute values without considering its usage. Chamodrkas
and Maratakos [16] have proposed a utility-based fuzzy
TOPSIS method for efficient network selection. This utility
based method has shown a new way for MADM selection 1
methods. In this paper, a fuzzy-AHP based network selection
method is presented for energy efficient network selection. 0.8

U (E) 0.6
III. PROPOSED TECHNIQUE
0.4
In this section, proposed techniques for network selection are
presented. Four networks (WLAN1, WLAN2, WiMax and
UMTS) are considered for network selection. Attributes of 0.2
these networks at a specific time are shown in Table 1.WLAN
networks have the higher bandwidths and lower delay while
0
UMTS networks are the most energy efficient but they provide 2 4 6 8 10
lower bandwidths and delay is also larger.
E(Watts)
Fig1: Linear utility function for energy consumption

978-1-4673-8253-3/15/$31.00 ©2015 IEEE

Authorized licensed use limited to: National Ilan Univ.. Downloaded on March 04,2021 at 15:14:54 UTC from IEEE Xplore. Restrictions apply.
Proceedings of 2015 RAECS UIET Panjab University Chandigarh 21-22nd December 2015

The utility values of these attributes for different networks are Table V: Comparative ratings of alternatives and weights of
illustrated in Table 2 [16]. Triangular fuzzy sets are used for the alternatives
representing fuzzy numbers ( , , ) as shown in equation 1.
Comparison Terms Fuzzy triangular matrix
0 , Equable (1,1,3)
, Slightly greater (1,3,5)
μ ( )= (1)
, Fairly greater (3,5,7)
0 Extremely greater (5,7,9)
Absolutely greater (7,9,9)

B. Fuzzy- AHP based network selection


Step5: Compute the value of the fuzzy synthetic extent with
Extent analysis method [21] is used to determine crisp weights respect to the object (attribute) according to the equation 2.
from the fuzzy comparison matrix in fuzzy AHP. The various = ∑ ∑ ∑ (2)
steps involved in fuzzy-AHP method for network selection are
as follows: where all are triangular fuzzy numbers. Synthetic extent
fuzzy values of attributes are shown in Table 6.
Step1: Objective of the problem is to be formulized. In our
case, Network selection from available networks is the Table VI: Synthetic extent values of attributes
objective of the problem. Attribute Synthetic extent fuzzy
values
Step2: Obtain the matrix having the information of the Bandwidth (0.07, 0.14,0.45)
alternatives according to the different criteria. Attribute values Delay (0.16,0.43,1.34)
of different networks are shown in Table 1. Power (Active State) (0.12,0.43,1.04)

Step3: Calculate the utility value of each attribute from


the respective utility function as discussed above. Utility Step6: Compute the degree of possibility of and as
values are represented in Table 2. In case of traditional fuzzy- shown in equation 3. and are calculated from Step4.
AHP, normalized weights are used as shown in Table 3. = ( , , ) and =( , , )
V( ≥ )= min( μ ( ), μ ( ) (3)
Table III: Normalized values of attributes of alternative which can be represented as equation 4 and equation 5.
networks V ( ≥ ) = hgt ( )= μ ( ) (4)
Values of degree of possibility of different attributes are
Bandwidth Delay Power shown in Table 7.
1 ≥
WLAN1 1 1 0.27 0 ≥
WLAN2 1 0.52 0.30 μ ( )= (5)
WiMax 0.18 0.50 0.34 ( ) ( )
UMTS 0.11 0.25 1
Table VII: Degree of Possibility between attributes
Step4: Prepare the relative comparison matrix of attributes as Degree of Possibility Value
shown in Table 4. Fuzzy triangular numbers values based on V (Bandwidth>Delay) 0.5
Table 5 are used for comparison. V (Bandwidth> Energy) 0.46
V (Delay> Bandwidth) 1
V (Delay> Energy) 1
Table IV: Fuzzy Comparison matrix for attributes V (Energy>Bandwidth) 1
Bandwidth Delay Power V (Energy>Delay) 1
(Active
State)
Bandwidth (1,1,1) (0.22,0.33,1) (0.22,0.33,1) Step6: Compute the degree of possibility for a convex fuzzy
Delay (1,3,5) (1,1,1) (1,1,3) number to be greater than k convex fuzzy numbers . This is
Power (1,3,5) (0.33,1,1) (1,1,1) to be calculated as shown in equation 6.
(Active V( , … . . ) = min V ( ), = 1,2. . . . . . (6)
State)

978-1-4673-8253-3/15/$31.00 ©2015 IEEE

Authorized licensed use limited to: National Ilan Univ.. Downloaded on March 04,2021 at 15:14:54 UTC from IEEE Xplore. Restrictions apply.
ECS UIET Panjab University Chandigarh 21-22nd Deccember 2015
Proceedings of 2015 RAE

Step7: Calculate the weight vector for each comparison requirement of 150ms. Theerefore, the rating of UMTS
matrix as shown in equation (7). WA AC = (min V( ≥ network is reduced considerabbly. For the reasons discussed
), min V( 2≥ )... min V( ≥ )) , = 1,2,..... , above, a mobile device will not undergo the unnecessary
= 1,2. . . . . , . handovers with utility based fuuzzy-AHP method as compared
(7)
Then the weights are normalized. Weightss of attributes are to the traditional AHP method.
shown in Table 8.

Table VIII: Normalized weights of attributtes in Fuzzy-AHP


method 0.8
Attribute Weight 0.7
Bandwidth 0.2 0.6
Rating 0.5
Delay 0.4 0.4
Power (Active State) 0.4 0.3
0.2
0.1
Step8: Total Score of each alternative iss calculated as in 0
equation 8. WLAN
N WLAN
Wimax UMTS
( ) = ∑ , = 1,2. . , 1 2
where is equal to number of attributes. (8) Fuzzy AHP 0.71 0.53 0.37 0.52
Step9: Sort the alternatives in descending order
o according to
Utility based fuzzy
the total scores calculated in Step8. 4
0.64 0.62 0.53 0.31
AHP

Fuzzy AHP Utility based fuzzy AHP


C. SAW based network selection

Also, we have analyzed the Simple Additive Weighing (SAW)


method for network selection. The crisp weights
w for fuzzy Fig 3: Comparisons of ratings of networks with fuzzy-AHP
triangular numbers are derived from [16]. Normalized methods
weights ( ) for the attributes are shown in Table 9.

Total Score (Rating) = ∑ 0.8


0.7
Table IX: Normalized weights of attributes in
i SAW method 0.6
Attribute Weight 0.5
Bandwidth 0.30 0.4
Delay 0.35 0.3
Power (Active State) 0.35 0.2
0.1
0
WLAN1
1 WLAN2 Wimax UMTS
IV RESULTS AND DISCUSS
SION SAW 0.7445
5 0.594 0.348 0.4705
Utility based
Results are obtained by calculating thee final rating as SAW
0.69 0.72 0.53 0.28
mentioned in step 8. The results obtained for fuzzy-AHP
methods for network selection are shown inn Figure 3. In the SAW Utility based SAW
traditional fuzzy-AHP the rating of WLAN 1 is very dominant
in comparison to the other networks as itt provides highest
bandwidth and least delay. WiMax has the leeast rating because
it consumes more energy in comparison withh UMTS network. Fig 4: Comparisons of ratings of
o networks with SAW methods
While in case of utility based fuzzy-AHP method,
m the actual
usage value of attribute is considered. Therefore, the The ratings with the utility based network selection are more
difference between the ratings of differrent networks is appropriate also as UMTS network cannot provide the
reduced. As the video conferencing applicattion requires 5000 required attribute values for viideo application. But in case of
kbps of bandwidth and 75ms of the delay [116], the difference traditional fuzzy-AHP method ratings of UMTS is more than
between WLAN1, WLAN2 and WiMax iss less. The utility the WiMax which is inapproprriate. For the utility based SAW
value for bandwidth is almost negligible inn UMTS network method also the ratings of WLAN1,
W WLAN2 and WiMax
and for delay its value is zero as it does not meet
m the minimum

978-1-4673-8253-3/15/$31.00 ©2015 IEEE

Authorized licensed use limited to: National Ilan Univ.. Downloaded on March 04,2021 at 15:14:54 UTC from IEEE Xplore. Restrictions apply.
Proceedings of 2015 RAECS UIET Panjab University Chandigarh 21-22nd December 2015

networks are much closer than the traditional SAW method as [11] M.M Alkhawini., K.A. Alslam, and A.A. Hussein, “Multi Criteria
Vertical Handover by TOPSIS and Fuzzy Logic”, IEEE International
shown in Fig 4.
Conference on Communications & Information Technology, pp. 96-102,
2011.
[12] X. Liu , and C. Chang, “The TOPSIS algorithm based on a + bi type
V CONCLUSION connection numbers for decision-making in the convergence of heterogeneous
networks”, IEEE 3rd International Conference on Advanced Computer
In this paper, a fuzzy-AHP based method for network Theory and Engineering (ICACTE), 323-327, 2010.
selection is presented for heterogeneous wireless networks. [13] O. E. Falowo, and H. A. Chan, “Dynamic RAT selection for multiple
Two WLAN networks, a WiMax and a UMTS network are calls in heterogeneous wireless networks using group decision-making
considered for network selection for a mobile device. Pairwise technique”, Computer Networks, vol. 56, 1390-401, 2012.
[14] M. Lahby, L. Cherkaoui, and A. Adib,“Novel validation approach for
comparisons are represented in the form of fuzzy sets. Extent network selection algorithm by applying the group MADM”, IEEE ACS
analysis method is used to derive crisp weights from this International Conference on Computer Systems and Applications (AICCSA),
comparison matrix. In the decision table, instead of pp. 1-4, 2013.
normalized values of attributes, utility values are used which [15] R.K. Goyal and S. Kaushal “Network Selection Using AHP for Fast
Moving Vehicles in Heterogeneous Networks” in Advs in Intelligent Syst.,
are calculated from utility functions. SAW method is also Comput., vol. 395, pp. 235-243, 2016.
implemented using utility values. The rating of each network [16] I. Chamodrakas, D. Martakos, “A utility-based fuzzy TOPSIS method for
is obtained for interactive video conferencing application. energy efficient network selection in heterogeneous wireless networks”, Appl.
Results show that with the proposed methods the ratings of Soft Comput., vol. 11, pp. 3734-3743, 2011.
[17] M. Xiao, N.B. Shroff and E.K.P. Chong, “A utility-based power-control
different types of networks are more suitable with respect to scheme in wireless cellular systems”, IEEE/ACM Trans. Netw, vol. 11, pp.
the requirements of the application these methods consider the 210–221, 2003.
utility values of with respect to the application. The ratings of [18] Q.-T. Nguyen-Vuong, N. Agoulmine, Y. Ghamri-Doudane, A user-
WLAN2 and WiMax networks are comparable to WLAN1 centric and context-aware solution to interface management and access
network selection in heterogeneous wireless environments, Comput. Netw,
and the rating of UMTS network is decreased as its bandwidth vol. 52, pp. 3358–3372, 2008.
utility value is very less and utility value of delay is zero in [19] I. Joe, W.-T. Kim, S. Hong, A network selection algorithm considering
case of video conferencing. power consumption in hybrid wireless networks, in: Proceedings of 16th
International Conference on Computer Communications and Networks,
With the proposed methods the mobile device can avoid Honolulu, Hawaii, USA, 2007, pp. 1240–1243.
[20] P.N. Tran, N. Boukhatem, A utility-based interface selection scheme for
unnecessary handovers as the ratings of networks that satisfy multi-homed mobile terminals, in: Proceedings of IEEE 20th International
the QoS criteria for video conferencing are very close to each Symposium on Personal, Indoor and Mobile Radio Communications, Tokyo,
other. Handover triggering factors like received signal strength pp. 767–772, 2009.
and velocity of the mobile terminal’s user is not considered in [21] D.Y. Chang, “Applications of the extent analysis method on fuzzy AHP”,
Eur. J. of Oper. Res., vol. 95, pp. 649-655, 1996.
this study. Future directions will lead to incorporate these
factors with the proposed network selection method.

References

[1] L.Budisz, R. Ferrus, A. Brunstorm, K. Grinnemo, R. Fracchia, G. Galante


and F. Casadevall, “Towards Transport-Layer Mobility: Evolution of SCTP
Multihoming”. Comput. Commun, vol. 31, pp. 980-998, 2008.
[2] R.V. Rao, Decision Making in Manufacturing Environment Using Graph
Theory and Fuzzy Multiple Attribute Decision Making Methods, Springer,
London, 2007.
[3] T.L.Saaty, The Analytic Hierarchy Process, McGraw-Hill, New York,
1980.
[4] C. L. Hwang and K. Yoon. Multiple attribute decision making: Methods
and applications, Springer-Verlag ,Berlin, 1981.
[5] J.P. Brans and P.H. Vincke, “A preference ranking organization method
(The PROMETHEE Method for MCDM). Manag. Sci, vol. 31, pp. 647–656.
[6] Roy B (1991) The outranking approach and the foundations of ELECTRE
methods. Theory Decis, vol. 31, pp. 49–73.
[7] E. Triantaphyllou and C. T. Lin (1996) Development and evaluation of
five fuzzy multiattribute decision-making methods. Int J Approx Reason, vol.
14, pp. 281–310.
[8] Q. Song, and A. Jamalipour, “An adaptive quality-of-service network
selection mechanism for heterogeneous mobile networks”, Wirel. Commun.
Mob. Com, vol. 5, pp.697-708, 2005.
[9] M. Lahby, L. Cherkaoui, and A. Adib, “An Enhanced-TOPSIS Based
Network Selection Technique for Next Generation Wireless Networks.” IEEE
20th International Conference on Telecommunications”, pp. 1-5, 2013.
[10] A. Ismail,and B. Roh, “Adaptive Handovers in heterogeneous networks
using fuzzy MADM”, IEEE International Conference on Mobile IT
convergence, pp. 99-104, 2011.

978-1-4673-8253-3/15/$31.00 ©2015 IEEE

Authorized licensed use limited to: National Ilan Univ.. Downloaded on March 04,2021 at 15:14:54 UTC from IEEE Xplore. Restrictions apply.

You might also like