BroadbandWireless Communication and Sensor Network Technology Key Laboratory of Ministry of Education Nanjing University ofPosts and Telecommunications Nanjing 210003 China
Correspondence should be addressed to Shengmei Liu smliunjupteducn
Received 17 July 2013 Revised 16 August 2013 Accepted 18 August 2013
Academic Editor Gelan Yang
Copyright copy 2013 Shengmei Liu et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
A novel preference ranking organization method by similarity to ideal solution (PROMSIS) vertical handoff algorithm is proposedfor heterogeneous wireless networks and its essential idea includes the preference structure of the PROMETHEE and the conceptof Euclid distance of the TOPSIS Four 3GPP defined traffic classes are considered in performance evaluation An attribute matrixis constructed considering some major attributes Handoff decision meeting multiattribute QoS requirement is made according tothe traffic features The weight relation of decision elements is determined with the least square (LS) approach The final decisionis made using the proposed PROMSIS algorithm based on the attribute matrix and weight vector The simulation results havemanifested that the proposed PROMSIS algorithm can provide satisfactory vertical handoff performance and the LS-PROMSISalgorithm can be fit to the characteristics of the traffic
1 Introduction
The architecture of the beyond 3rd generation (B3G) or 4thgeneration (4G)wireless networks aims at integrating variousheterogeneous wireless access networks over an IP basedbackbone To provide seamless mobility one of the design is-sues is the vertical handoff support [1 2] a multiple attributesdecision subject Since the handoff may happen in differ-ent RATs and management domains handoff decision willdepend on the combination of multiple attributes rather thana single parameter
In general the vertical handoff process can be dividedinto three main steps namely system discovery handoffdecision and handoff execution During the phase of systemdiscovery the networks may advertise the supported datarates and quality-of-service (QoS) parameters for differentservices Because the users are mobile the available collo-cated networks depend on the location of the userThe trafficload in each network may also change with time Thus thisphase may be periodically invoked
Various vertical handoff decision mechanisms have beenproposed In [3] a combining SINR based vertical handoff(CSVH) algorithm is proposed It studies the combined
effects of SINR in different access networks that is in thesource network and the equivalents in the target networkscompared with the RSS based vertical handoff algorithmFurther on a multidimensional adaptive SINR based ver-tical handoff (MASVH) algorithm is proposed in [4] Inaddition to the combined effects of SINR it also takesaccount of the user required bandwidth traffic cost andresource utilization in the participating access networks Aparameter 119896 is used in the MASVH algorithm to adjust theweight of multiple attributes Nevertheless no discussionelaborates on the determination of the optimal 119896 value underdifferent conditions the relation between multiple attributesthe relative importance of each attribute and the impactof system load There are also some researchers focusingon solving the ping-pong effect In [5] the user movementinformation is considered and the residence time in a basestation is estimated to avoid the unnecessary handoff
In this paper a novel algorithm namely the PROMSISalgorithm is proposed to be applied in the vertical handoffdecision technology based on the preference ranking orga-nization method for enrichment evaluation (PROMETHEE)[6 7] and the TOPSIS [8] In TOPSIS the level of thedecision makerrsquos participation is rather low in the process
2 International Journal of Distributed Sensor Networks
of decision making and the decision makerrsquos preferenceinformation is not integrated into the method So weintroduce the preference function associated with each cri-terion in PROMETHEE integrate the preference structureof PROMETHEE into TOPSIS and obtain PROMSIS as aresult The scenario analyzed is referred to in [4] It considersmultiple attributes concluding in the combined effects ofSINR in WLAN andWCDMA the required bandwidth ser-vice cost and available bandwidth of the participating accessnetworks to make handoff decisions meeting multiattributeQoS requirement An attributematrix of alternative networksis established An appropriate weight factor is assigned toeach criterion to account for its importance In the weightdetermining process four 3GPP defined traffic classes [8]are considered and the least square (LS) weighted approachmethod [9] is adopted Finally how the connections arecontained or rerouted is decided by the PROMSIS (or LS-PROMSIS) algorithm according to the attribute matrix andthe weight vector
2 PROMSIS and LS-PROMSIS VerticalHandoff Algorithm
The handoff metrics and QoS parameters are categorizedinto different groups (eg bandwidth latency power pricesecurity reliability availability etc) Some representativemetrics approaches are considered in this paper
Assuming that there are 119886BSs and 119887Aps all candidate BSsand APs for the user can be indexed by 1 to 119886 + 119887 in the set
P = [BS1BS2 BS
119886AP1AP2 AP
119887] (1)
For each handoff event the best BS or AP from thecandidate set P for each user will be determined by thehandoff algorithm considering the following criteria SINRthe required bandwidth traffic cost and network availablebandwidth
21 Attribute Matrix Let us presume 119877AP and 119877BS as themaximum achievable downlink data rates of WLAN andWCDMA According to Shannon capacity we have
119877AP = 119882AP log2 (1 +120574APΓAP
)
119877BS = 119882BS log2 (1 +120574BSΓBS)
(2)
where 120574AP and 120574BS are the receiving SINR values from thecoexisting networks When the networks offer the samedownlink data rate to the user that is 119877AP = 119877BS we cansolve the equation and get the relationship between 120574AP and120574BS as
120574BS = ΓBS ((1 +120574APΓAP
)
119882AP119882BS
minus 1) (3)
where the carrier bandwidth is 22MHz for WLAN119882AP and5MHz forWCDMA119882BS ΓAP is equal to 3 dB forWLAN andΓBS is equal to 16 dB for WCDMA
It is assumed that a BS is transmitted to merely one uservia the HSDPA channel at a time with the maximum powerto achieve the optimal physical rate The SINR 120574BS119895 119894 receivedby user 119894 fromWCDMA BS
where 119875AP119895 is the transmitting power of AP119895 119866AP119895 119894 is
the channel gain between user 119894 and AP119895 and 119875
119861is the
background noise power equal to minus86 dBmA macrocell propagation model for urban and suburban
areas [3] is adopted and for an antenna height of 15 metersthe path loss is
Path loss (dB) = 588 + 21log10(119891)
+ 376log10(119877) + log (119865)
(6)
where 119891 is the carrier frequency (2GHz for WCDMA and24GHz for WLAN) 119877 is the distance in meters betweenthe user and the BS or AP and log(119865) is the log-normaldistribution shadowing with standard deviation 120590 = 10 dB
Using (3) the SINR received fromAPs (SAP119894) is convertedto the equivalent SINR (S1015840AP119894) to achieve the same data rate viaBS
The set of the SINR value S119894of all BSs and APs in the
candidate set P for the user 119894 can be represented by
S119894= SBS119894 cup S1015840AP119894 (7)
For a required bandwidth 119877119894for a user 119894 the minimum
receiving SINR from BS (120574min119894) can be calculated in Shannonequation
Let us suppose C to be the system cost vector In order todirectly associate the cost value with the SINR value the costper bit is converted to cost per SINR (CSINR)
Let us suppose U as the network available bandwidthvector represented by the available capacity of each candidateBS and AP
International Journal of Distributed Sensor Networks 3
Then the attribute matrix is as follows
R119886=
[[[[[
[
S minus 120574min
1
CSINR
U
]]]]]
]
119879
=
11988311198832sdot sdot sdot 119883
119899
1198601
1198602
119860119898
[[[[[[[
[
11990911
11990912
sdot sdot sdot 1199091119899
11990921
11990922
sdot sdot sdot 1199092119899
sdot sdot sdot sdot sdot sdot
sdot sdot sdot sdot sdot sdot
sdot sdot sdot sdot sdot sdot
1199091198981
1199091198982
sdot sdot sdot 119909119898119899
]]]]]]]
]
= (119909119894119895)119898times119899
(8)
where1198601 1198602 119860
119898are feasible alternatives119898 = 119886+119887 and
1198831 1198832 119883
119899are evaluation attributes 119899 = 3 Here 119909
119894119895is the
performance rating for alternative 119860119894under attribute119883
119895
22 Handoff Decision The proposed PROMSIS is also amulticriteria analysis approach and its essence includes thepreference structure of the PROMETHEE and the concept ofEuclid distance of the TOPSIS In the first place it performsthe comparison between every pair of solutions (119886
119894 119886119903) using
a preference function 119901(119889) and 119889 = 119891119895(119886119894) minus 119891
119895(119886119903) is
the difference between the evaluations of two alternativesReference [7] contains many types of preference functionsThis function 119901(119889) reflects the preference level of 119886
119894over 119886
119903
in the interval [0 1] in such a way that if 119901(119889) = 0 then 119886119894is
indifferent to 119886119903 if 119901(119889) = 1 then 119886
119894is strictly preferred to 119886
119903
PROMSIS consists of the following steps
(1) Construct the decision matrix R119886and the weight
vector w
(2) Define the preference function for each attribute
(3) Define the preference index for each couple of alter-natives
119899 = 119881 (119886119894 119886119903) =
119899
sum
119895=1
119908119895sdot 119901119895(119891119895(119886119894) minus 119891119895(119886119903)) (9)
The preference index is given in the intensity of pref-erence of the decision maker for 119886
119894over 119886
119903 We have
0 le 119881(119886119894 119886119903) le 1 The matrixV = (V
119894119903)119898times119899
is obtainedto calculate the weighted Euclidean distances
(4) The preference concept from PROMETHEE is pre-sented as above and now we use the concept ofEuclid distance in TOPSIS to continue Define thepositive ideal point and the negative ideal point andcalculate the distance between each scheme and thepositivenegative ideal point Calculate the distance119878+
119894between each scheme and the positive ideal point
and the distance 119878minus119894between each scheme and the
negative ideal point
119878+
119894= radic
119899
sum
119895=1
(V119894119895minus V+119895)2
119894 isin 119898
119878minus
119894= radic
119899
sum
119895=1
(V119894119895minus Vminus119895)2
119894 isin 119898
(10)
(5) Calculate the relative approach degree 119862+119894of each
scheme to the ideal points
119862+
119894=
119878minus
119894
(119878+
119894+ 119878minus
119894) 0 lt 119862
+
119894lt 1 119894 isin 119898 (11)
(6) Ranke the schemes based on 119862+119894 The larger is the 119862+
119894
the better is the scheme
Above all the proposed PROMSIS combines the qualita-tive and quantitative analysis Preference comparison corre-sponds to the qualitative analysisThe Euclidean distance candescribe the degree of preference through the quantity Sothe utility of the network can be achieved in both qualitativeand quantitative aspects And the final decision will beappropriate based on subjective and objective factors
23 Weight Vector There are a variety of weight methodssuch as analytic hierarchy process (AHP) and the informationEntropy weight method some are subjective and others areobjective We can choose the appropriate weight methodaccording to actual conditions In the weight determiningprocess we apply the LS [9] to estimate the weights ofdecision elements introduced by Chu in 1979
Firstly the comparison matrix G119888= (119892119894119895)119899times119899
is definedaccording to the relative importance The judgments areranked on a 9-point scale [7] Numbers 1 to 9 are usedto represent equal weakly moderate moderate moderateplus strong strong plus very strong very very strong andextremely important to the objective respectively When anelement is less important than another the comparison resultequals the reciprocal of one of the numbers So for the matrixG119888to be the diagonal elements are observed 1 demonstrating
the elementsrsquo self-comparisons The other entries in thematrix are symmetric with respect to the diagonal as a resultof the inverted comparisons
Four traffic classes defined by 3GPP are taken into con-sideration namely the conversational streaming interactiveand background classes Based on the traffic requirementsthe comparison matrices for the four traffic classes accordingto the 9-point scale can be established
The element 119892119894119895of matrix G
119888shall be considered and
desired to determine the weights 119908119894 such that given 119892
119894119895
119892119894119895asymp 119908119894119908119895
4 International Journal of Distributed Sensor Networks
The weights can be obtained by solving the constrainedoptimization problem
Equations (15) and (13) form a set of 119899+1 inhomogeneouslinear equations with 119899 + 1 unknowns
By the way using the numerical method to solve mathe-matical problems due to almost inevitable rounding errorsthe results obtained are generally inaccurate Some othermeasures can be applied to estimate the error but we will notgo into detail here due to space limitations
3 Simulation Results
In this research we concentrate on the downlink traffic sinceit normally requires higher bandwidth than uplink especiallyfor multimedia services such as video streaming through theHSDPA channel while connected to WCDMA
The performance of different vertical handoff algorithmshas been evaluated with the scenario illustrated in Figure 1in which there are 7 BS and 12 AP placed at each WCDMAcell boundary The WCDMA cell radius is 1200 meter 200randomly generated UEs are used inside the simulation areawhose position changes in the time interval depending ontheir moving speed and directionThe direction is uniformlydistributed in the range of [0 2120587] and the speed change rateis 5 per 100 seconds The maximum userrsquos moving speedis 80 kmhour In the traffic generator module for a meansession duration of 60 seconds and a certain given meansession arrival rate user traffic is randomly generated witha Poisson arrival distribution As revealed in Figure 1 thedirection of the arrow represents the userrsquos moving directionThe length of the arrow corresponds to themoving distance ofthe user in 20 seconds (supposing that themoving direction isfixed in the 20 seconds) so the larger one indicates the fastermoving speed
The V-shape with indifference criterion type preferencefunction in PROMETHEE was adopted here In case of thistype the thresholds of indifference 119902 (119902 = 01) and strictpreference 119901 (119901 = 05) have to be separately selected
The system performance for different session arrivalrates is shown in Figure 2 The simulated algorithms include
the proposed LS-PROMSIS algorithm and the MASVH(119896 = 4) algorithm [6] The downlink system throughputof each algorithm is measured and shown in Figure 2(a) Itdemonstrates that the LS-PROMSIS algorithm for streamingtraffic class achieves the highest throughput performancebecause ldquothe available bandwidthrdquo attribute has the greatestweight in the handoff criterion so the network of the largestavailable bandwidth is selected considering the load balanceLikewise the system dropping probability of this algorithmis the lowest The dropping probability performance of eachalgorithm is exhibited in Figure 2(b) The average user trafficcost performance is presented in Figure 2(c) It is noted thatthe cost of the LS-PROMSIS algorithm for streaming trafficclass is quite high but the cost of the LS-PROMSIS algorithmfor conversational traffic is the lowest Since the attribute ofuser traffic cost covers the highest proportion in the handoffcriterion the network of the lowest cost is inclined to beselected Figure 2(d) indicates the number of vertical handoffIt is revealed that the number of vertical handoffs of the LS-PROMSIS algorithm for streaming traffic class is the highestand that of theMASVH algorithm finishes the second By theway if the unnecessary handoff needs to be further decreasedthe proposed MADM algorithm can be combined with themobile prediction technique to mitigate the impact of theping-pong effect
4 Conclusions
In this paper a novel PROMSIS vertical handoff algorithmis proposed and compared with the existing CSVH andMASVH algorithms The vertical handoff of heterogeneousnetworks is amultiple attributes decision subjectWith regardto the relations between all the attributes the observedobjects are the four 3GPP defined traffic classes Accordingto the features of diverse traffic classes the weight of eachattribute in the handoff criterion is determined by LS The
International Journal of Distributed Sensor Networks 5
001 0015 002 0025 003 0035 004 0045 00510
15
20
25
30
35
Session arrival rate (sessionss)
Thro
ughp
ut (M
bps)
(a) System throughput
001 0015 002 0025 003 0035 004 0045 0050
005
01
015
02
025
03
035
Session arrival rate (sessionss)
Syste
m d
ropp
ing
prob
abili
ty(b) System dropping probability
001 0015 002 0025 003 0035 004 0045 005042
0425
043
0435
044
0445
045
0455
046
Session arrival rate (sessionss)
Aver
age u
ser c
ost (
cost
SIN
R)
MASVH k = 4
PROMSIS conversationalPROMSIS streaming
PROMSIS interactivePROMSIS background
(c) Average cost to user traffic of the algorithms
001 0015 002 0025 003 0035 004 0045 00520
40
60
80
100
120
140
160
180
Session arrival rate (sessionss)
The n
umbe
r of h
ando
ff (n
umbe
rsim
ulat
ion)
MASVH k = 4
PROMSIS conversationalPROMSIS streaming
PROMSIS interactivePROMSIS background
(d) The number of vertical handoffs
Figure 2 Performance of each algorithm
simulation results display that the performance of the algo-rithm is affected by the allocated weight vector Consequentlyin practice we should consider both the characteristics of thetraffic and the preference of the user andweigh the advantagesand disadvantages before making the decision According tothe analysis and simulation results the PROMSIS algorithm
can achieve the satisfactory performance for the network andthe user
For future work more comparisons with other verti-cal handoff methods will be further discussed and othertechniques to solve the decision problem such as the gametheory will also be taken into account
6 International Journal of Distributed Sensor Networks
Acknowledgments
This work was supported by the National Natural ScienceFoundation of China under Contract no 61271235 and Nat-ural Science Foundation of Education Committee of JiangsuProvince (no 11KJB510014)
References
[1] A Singhrova and N Prakash ldquoVertical handoff decision algo-rithm for improved quality of service in heterogeneous wirelessnetworksrdquo IET Communications vol 6 no 2 pp 211ndash223 2012
[2] F Kaleem and A Mehbodniya ldquoDynamic target wireless net-work selection technique using fuzzy linguistic variablesrdquoChinaCommunications vol 10 no 1 pp 1ndash16 2013
[3] K Yang I Gondal B Qiu and L S Dooley ldquoCombined SINRbased vertical handoff algorithm for next generation hetero-geneous wireless networksrdquo in Proceedings of the 50th AnnualIEEEGlobal Telecommunications Conference (GLOBECOM rsquo07)pp 4483ndash4487 Washington DC USA November 2007
[4] K Yang I Gondal and B Qiu ldquoMulti-dimensional adaptiveSINR based vertical handoff for heterogeneous wireless net-worksrdquo IEEE Communications Letters vol 12 no 6 pp 438ndash440 2008
[5] L Bin and L Shengmei ldquoVertical handoff algorithm based onmobility predictionrdquoApplication of Electronic Technique vol 39no 1 pp 93ndash95 2013
[6] R O Parreiras J H R D Maciel and J A Vasconcelos ldquoThea posteriori decision in multiobjective optimization problemswith smarts promethee II and a fuzzy algorithmrdquo IEEETransactions on Magnetics vol 42 no 4 pp 1139ndash1142 2006
[7] J P Brans P Vincke and B Mareschal ldquoHow to select and howto rank projects the Promethee methodrdquo European Journal ofOperational Research vol 24 no 2 pp 228ndash238 1986
[8] E Stevens-Navarro and V W S Wong ldquoComparison betweenvertical handoff decision algorithms for heterogeneous wirelessnetworksrdquo in Proceedings of the IEEE 63rd Vehicular TechnologyConference (VTC-Sping rsquo06) pp 947ndash951Melbourne AustraliaMay 2006
[9] A T W Chu R E Kalaba and K Spingarn ldquoA comparison oftwo methods for determining the weights of belonging to fuzzysetsrdquo Journal of Optimization Theory and Applications vol 27no 4 pp 531ndash538 1979
2 International Journal of Distributed Sensor Networks
of decision making and the decision makerrsquos preferenceinformation is not integrated into the method So weintroduce the preference function associated with each cri-terion in PROMETHEE integrate the preference structureof PROMETHEE into TOPSIS and obtain PROMSIS as aresult The scenario analyzed is referred to in [4] It considersmultiple attributes concluding in the combined effects ofSINR in WLAN andWCDMA the required bandwidth ser-vice cost and available bandwidth of the participating accessnetworks to make handoff decisions meeting multiattributeQoS requirement An attributematrix of alternative networksis established An appropriate weight factor is assigned toeach criterion to account for its importance In the weightdetermining process four 3GPP defined traffic classes [8]are considered and the least square (LS) weighted approachmethod [9] is adopted Finally how the connections arecontained or rerouted is decided by the PROMSIS (or LS-PROMSIS) algorithm according to the attribute matrix andthe weight vector
2 PROMSIS and LS-PROMSIS VerticalHandoff Algorithm
The handoff metrics and QoS parameters are categorizedinto different groups (eg bandwidth latency power pricesecurity reliability availability etc) Some representativemetrics approaches are considered in this paper
Assuming that there are 119886BSs and 119887Aps all candidate BSsand APs for the user can be indexed by 1 to 119886 + 119887 in the set
P = [BS1BS2 BS
119886AP1AP2 AP
119887] (1)
For each handoff event the best BS or AP from thecandidate set P for each user will be determined by thehandoff algorithm considering the following criteria SINRthe required bandwidth traffic cost and network availablebandwidth
21 Attribute Matrix Let us presume 119877AP and 119877BS as themaximum achievable downlink data rates of WLAN andWCDMA According to Shannon capacity we have
119877AP = 119882AP log2 (1 +120574APΓAP
)
119877BS = 119882BS log2 (1 +120574BSΓBS)
(2)
where 120574AP and 120574BS are the receiving SINR values from thecoexisting networks When the networks offer the samedownlink data rate to the user that is 119877AP = 119877BS we cansolve the equation and get the relationship between 120574AP and120574BS as
120574BS = ΓBS ((1 +120574APΓAP
)
119882AP119882BS
minus 1) (3)
where the carrier bandwidth is 22MHz for WLAN119882AP and5MHz forWCDMA119882BS ΓAP is equal to 3 dB forWLAN andΓBS is equal to 16 dB for WCDMA
It is assumed that a BS is transmitted to merely one uservia the HSDPA channel at a time with the maximum powerto achieve the optimal physical rate The SINR 120574BS119895 119894 receivedby user 119894 fromWCDMA BS
where 119875AP119895 is the transmitting power of AP119895 119866AP119895 119894 is
the channel gain between user 119894 and AP119895 and 119875
119861is the
background noise power equal to minus86 dBmA macrocell propagation model for urban and suburban
areas [3] is adopted and for an antenna height of 15 metersthe path loss is
Path loss (dB) = 588 + 21log10(119891)
+ 376log10(119877) + log (119865)
(6)
where 119891 is the carrier frequency (2GHz for WCDMA and24GHz for WLAN) 119877 is the distance in meters betweenthe user and the BS or AP and log(119865) is the log-normaldistribution shadowing with standard deviation 120590 = 10 dB
Using (3) the SINR received fromAPs (SAP119894) is convertedto the equivalent SINR (S1015840AP119894) to achieve the same data rate viaBS
The set of the SINR value S119894of all BSs and APs in the
candidate set P for the user 119894 can be represented by
S119894= SBS119894 cup S1015840AP119894 (7)
For a required bandwidth 119877119894for a user 119894 the minimum
receiving SINR from BS (120574min119894) can be calculated in Shannonequation
Let us suppose C to be the system cost vector In order todirectly associate the cost value with the SINR value the costper bit is converted to cost per SINR (CSINR)
Let us suppose U as the network available bandwidthvector represented by the available capacity of each candidateBS and AP
International Journal of Distributed Sensor Networks 3
Then the attribute matrix is as follows
R119886=
[[[[[
[
S minus 120574min
1
CSINR
U
]]]]]
]
119879
=
11988311198832sdot sdot sdot 119883
119899
1198601
1198602
119860119898
[[[[[[[
[
11990911
11990912
sdot sdot sdot 1199091119899
11990921
11990922
sdot sdot sdot 1199092119899
sdot sdot sdot sdot sdot sdot
sdot sdot sdot sdot sdot sdot
sdot sdot sdot sdot sdot sdot
1199091198981
1199091198982
sdot sdot sdot 119909119898119899
]]]]]]]
]
= (119909119894119895)119898times119899
(8)
where1198601 1198602 119860
119898are feasible alternatives119898 = 119886+119887 and
1198831 1198832 119883
119899are evaluation attributes 119899 = 3 Here 119909
119894119895is the
performance rating for alternative 119860119894under attribute119883
119895
22 Handoff Decision The proposed PROMSIS is also amulticriteria analysis approach and its essence includes thepreference structure of the PROMETHEE and the concept ofEuclid distance of the TOPSIS In the first place it performsthe comparison between every pair of solutions (119886
119894 119886119903) using
a preference function 119901(119889) and 119889 = 119891119895(119886119894) minus 119891
119895(119886119903) is
the difference between the evaluations of two alternativesReference [7] contains many types of preference functionsThis function 119901(119889) reflects the preference level of 119886
119894over 119886
119903
in the interval [0 1] in such a way that if 119901(119889) = 0 then 119886119894is
indifferent to 119886119903 if 119901(119889) = 1 then 119886
119894is strictly preferred to 119886
119903
PROMSIS consists of the following steps
(1) Construct the decision matrix R119886and the weight
vector w
(2) Define the preference function for each attribute
(3) Define the preference index for each couple of alter-natives
119899 = 119881 (119886119894 119886119903) =
119899
sum
119895=1
119908119895sdot 119901119895(119891119895(119886119894) minus 119891119895(119886119903)) (9)
The preference index is given in the intensity of pref-erence of the decision maker for 119886
119894over 119886
119903 We have
0 le 119881(119886119894 119886119903) le 1 The matrixV = (V
119894119903)119898times119899
is obtainedto calculate the weighted Euclidean distances
(4) The preference concept from PROMETHEE is pre-sented as above and now we use the concept ofEuclid distance in TOPSIS to continue Define thepositive ideal point and the negative ideal point andcalculate the distance between each scheme and thepositivenegative ideal point Calculate the distance119878+
119894between each scheme and the positive ideal point
and the distance 119878minus119894between each scheme and the
negative ideal point
119878+
119894= radic
119899
sum
119895=1
(V119894119895minus V+119895)2
119894 isin 119898
119878minus
119894= radic
119899
sum
119895=1
(V119894119895minus Vminus119895)2
119894 isin 119898
(10)
(5) Calculate the relative approach degree 119862+119894of each
scheme to the ideal points
119862+
119894=
119878minus
119894
(119878+
119894+ 119878minus
119894) 0 lt 119862
+
119894lt 1 119894 isin 119898 (11)
(6) Ranke the schemes based on 119862+119894 The larger is the 119862+
119894
the better is the scheme
Above all the proposed PROMSIS combines the qualita-tive and quantitative analysis Preference comparison corre-sponds to the qualitative analysisThe Euclidean distance candescribe the degree of preference through the quantity Sothe utility of the network can be achieved in both qualitativeand quantitative aspects And the final decision will beappropriate based on subjective and objective factors
23 Weight Vector There are a variety of weight methodssuch as analytic hierarchy process (AHP) and the informationEntropy weight method some are subjective and others areobjective We can choose the appropriate weight methodaccording to actual conditions In the weight determiningprocess we apply the LS [9] to estimate the weights ofdecision elements introduced by Chu in 1979
Firstly the comparison matrix G119888= (119892119894119895)119899times119899
is definedaccording to the relative importance The judgments areranked on a 9-point scale [7] Numbers 1 to 9 are usedto represent equal weakly moderate moderate moderateplus strong strong plus very strong very very strong andextremely important to the objective respectively When anelement is less important than another the comparison resultequals the reciprocal of one of the numbers So for the matrixG119888to be the diagonal elements are observed 1 demonstrating
the elementsrsquo self-comparisons The other entries in thematrix are symmetric with respect to the diagonal as a resultof the inverted comparisons
Four traffic classes defined by 3GPP are taken into con-sideration namely the conversational streaming interactiveand background classes Based on the traffic requirementsthe comparison matrices for the four traffic classes accordingto the 9-point scale can be established
The element 119892119894119895of matrix G
119888shall be considered and
desired to determine the weights 119908119894 such that given 119892
119894119895
119892119894119895asymp 119908119894119908119895
4 International Journal of Distributed Sensor Networks
The weights can be obtained by solving the constrainedoptimization problem
Equations (15) and (13) form a set of 119899+1 inhomogeneouslinear equations with 119899 + 1 unknowns
By the way using the numerical method to solve mathe-matical problems due to almost inevitable rounding errorsthe results obtained are generally inaccurate Some othermeasures can be applied to estimate the error but we will notgo into detail here due to space limitations
3 Simulation Results
In this research we concentrate on the downlink traffic sinceit normally requires higher bandwidth than uplink especiallyfor multimedia services such as video streaming through theHSDPA channel while connected to WCDMA
The performance of different vertical handoff algorithmshas been evaluated with the scenario illustrated in Figure 1in which there are 7 BS and 12 AP placed at each WCDMAcell boundary The WCDMA cell radius is 1200 meter 200randomly generated UEs are used inside the simulation areawhose position changes in the time interval depending ontheir moving speed and directionThe direction is uniformlydistributed in the range of [0 2120587] and the speed change rateis 5 per 100 seconds The maximum userrsquos moving speedis 80 kmhour In the traffic generator module for a meansession duration of 60 seconds and a certain given meansession arrival rate user traffic is randomly generated witha Poisson arrival distribution As revealed in Figure 1 thedirection of the arrow represents the userrsquos moving directionThe length of the arrow corresponds to themoving distance ofthe user in 20 seconds (supposing that themoving direction isfixed in the 20 seconds) so the larger one indicates the fastermoving speed
The V-shape with indifference criterion type preferencefunction in PROMETHEE was adopted here In case of thistype the thresholds of indifference 119902 (119902 = 01) and strictpreference 119901 (119901 = 05) have to be separately selected
The system performance for different session arrivalrates is shown in Figure 2 The simulated algorithms include
the proposed LS-PROMSIS algorithm and the MASVH(119896 = 4) algorithm [6] The downlink system throughputof each algorithm is measured and shown in Figure 2(a) Itdemonstrates that the LS-PROMSIS algorithm for streamingtraffic class achieves the highest throughput performancebecause ldquothe available bandwidthrdquo attribute has the greatestweight in the handoff criterion so the network of the largestavailable bandwidth is selected considering the load balanceLikewise the system dropping probability of this algorithmis the lowest The dropping probability performance of eachalgorithm is exhibited in Figure 2(b) The average user trafficcost performance is presented in Figure 2(c) It is noted thatthe cost of the LS-PROMSIS algorithm for streaming trafficclass is quite high but the cost of the LS-PROMSIS algorithmfor conversational traffic is the lowest Since the attribute ofuser traffic cost covers the highest proportion in the handoffcriterion the network of the lowest cost is inclined to beselected Figure 2(d) indicates the number of vertical handoffIt is revealed that the number of vertical handoffs of the LS-PROMSIS algorithm for streaming traffic class is the highestand that of theMASVH algorithm finishes the second By theway if the unnecessary handoff needs to be further decreasedthe proposed MADM algorithm can be combined with themobile prediction technique to mitigate the impact of theping-pong effect
4 Conclusions
In this paper a novel PROMSIS vertical handoff algorithmis proposed and compared with the existing CSVH andMASVH algorithms The vertical handoff of heterogeneousnetworks is amultiple attributes decision subjectWith regardto the relations between all the attributes the observedobjects are the four 3GPP defined traffic classes Accordingto the features of diverse traffic classes the weight of eachattribute in the handoff criterion is determined by LS The
International Journal of Distributed Sensor Networks 5
001 0015 002 0025 003 0035 004 0045 00510
15
20
25
30
35
Session arrival rate (sessionss)
Thro
ughp
ut (M
bps)
(a) System throughput
001 0015 002 0025 003 0035 004 0045 0050
005
01
015
02
025
03
035
Session arrival rate (sessionss)
Syste
m d
ropp
ing
prob
abili
ty(b) System dropping probability
001 0015 002 0025 003 0035 004 0045 005042
0425
043
0435
044
0445
045
0455
046
Session arrival rate (sessionss)
Aver
age u
ser c
ost (
cost
SIN
R)
MASVH k = 4
PROMSIS conversationalPROMSIS streaming
PROMSIS interactivePROMSIS background
(c) Average cost to user traffic of the algorithms
001 0015 002 0025 003 0035 004 0045 00520
40
60
80
100
120
140
160
180
Session arrival rate (sessionss)
The n
umbe
r of h
ando
ff (n
umbe
rsim
ulat
ion)
MASVH k = 4
PROMSIS conversationalPROMSIS streaming
PROMSIS interactivePROMSIS background
(d) The number of vertical handoffs
Figure 2 Performance of each algorithm
simulation results display that the performance of the algo-rithm is affected by the allocated weight vector Consequentlyin practice we should consider both the characteristics of thetraffic and the preference of the user andweigh the advantagesand disadvantages before making the decision According tothe analysis and simulation results the PROMSIS algorithm
can achieve the satisfactory performance for the network andthe user
For future work more comparisons with other verti-cal handoff methods will be further discussed and othertechniques to solve the decision problem such as the gametheory will also be taken into account
6 International Journal of Distributed Sensor Networks
Acknowledgments
This work was supported by the National Natural ScienceFoundation of China under Contract no 61271235 and Nat-ural Science Foundation of Education Committee of JiangsuProvince (no 11KJB510014)
References
[1] A Singhrova and N Prakash ldquoVertical handoff decision algo-rithm for improved quality of service in heterogeneous wirelessnetworksrdquo IET Communications vol 6 no 2 pp 211ndash223 2012
[2] F Kaleem and A Mehbodniya ldquoDynamic target wireless net-work selection technique using fuzzy linguistic variablesrdquoChinaCommunications vol 10 no 1 pp 1ndash16 2013
[3] K Yang I Gondal B Qiu and L S Dooley ldquoCombined SINRbased vertical handoff algorithm for next generation hetero-geneous wireless networksrdquo in Proceedings of the 50th AnnualIEEEGlobal Telecommunications Conference (GLOBECOM rsquo07)pp 4483ndash4487 Washington DC USA November 2007
[4] K Yang I Gondal and B Qiu ldquoMulti-dimensional adaptiveSINR based vertical handoff for heterogeneous wireless net-worksrdquo IEEE Communications Letters vol 12 no 6 pp 438ndash440 2008
[5] L Bin and L Shengmei ldquoVertical handoff algorithm based onmobility predictionrdquoApplication of Electronic Technique vol 39no 1 pp 93ndash95 2013
[6] R O Parreiras J H R D Maciel and J A Vasconcelos ldquoThea posteriori decision in multiobjective optimization problemswith smarts promethee II and a fuzzy algorithmrdquo IEEETransactions on Magnetics vol 42 no 4 pp 1139ndash1142 2006
[7] J P Brans P Vincke and B Mareschal ldquoHow to select and howto rank projects the Promethee methodrdquo European Journal ofOperational Research vol 24 no 2 pp 228ndash238 1986
[8] E Stevens-Navarro and V W S Wong ldquoComparison betweenvertical handoff decision algorithms for heterogeneous wirelessnetworksrdquo in Proceedings of the IEEE 63rd Vehicular TechnologyConference (VTC-Sping rsquo06) pp 947ndash951Melbourne AustraliaMay 2006
[9] A T W Chu R E Kalaba and K Spingarn ldquoA comparison oftwo methods for determining the weights of belonging to fuzzysetsrdquo Journal of Optimization Theory and Applications vol 27no 4 pp 531ndash538 1979
International Journal of Distributed Sensor Networks 3
Then the attribute matrix is as follows
R119886=
[[[[[
[
S minus 120574min
1
CSINR
U
]]]]]
]
119879
=
11988311198832sdot sdot sdot 119883
119899
1198601
1198602
119860119898
[[[[[[[
[
11990911
11990912
sdot sdot sdot 1199091119899
11990921
11990922
sdot sdot sdot 1199092119899
sdot sdot sdot sdot sdot sdot
sdot sdot sdot sdot sdot sdot
sdot sdot sdot sdot sdot sdot
1199091198981
1199091198982
sdot sdot sdot 119909119898119899
]]]]]]]
]
= (119909119894119895)119898times119899
(8)
where1198601 1198602 119860
119898are feasible alternatives119898 = 119886+119887 and
1198831 1198832 119883
119899are evaluation attributes 119899 = 3 Here 119909
119894119895is the
performance rating for alternative 119860119894under attribute119883
119895
22 Handoff Decision The proposed PROMSIS is also amulticriteria analysis approach and its essence includes thepreference structure of the PROMETHEE and the concept ofEuclid distance of the TOPSIS In the first place it performsthe comparison between every pair of solutions (119886
119894 119886119903) using
a preference function 119901(119889) and 119889 = 119891119895(119886119894) minus 119891
119895(119886119903) is
the difference between the evaluations of two alternativesReference [7] contains many types of preference functionsThis function 119901(119889) reflects the preference level of 119886
119894over 119886
119903
in the interval [0 1] in such a way that if 119901(119889) = 0 then 119886119894is
indifferent to 119886119903 if 119901(119889) = 1 then 119886
119894is strictly preferred to 119886
119903
PROMSIS consists of the following steps
(1) Construct the decision matrix R119886and the weight
vector w
(2) Define the preference function for each attribute
(3) Define the preference index for each couple of alter-natives
119899 = 119881 (119886119894 119886119903) =
119899
sum
119895=1
119908119895sdot 119901119895(119891119895(119886119894) minus 119891119895(119886119903)) (9)
The preference index is given in the intensity of pref-erence of the decision maker for 119886
119894over 119886
119903 We have
0 le 119881(119886119894 119886119903) le 1 The matrixV = (V
119894119903)119898times119899
is obtainedto calculate the weighted Euclidean distances
(4) The preference concept from PROMETHEE is pre-sented as above and now we use the concept ofEuclid distance in TOPSIS to continue Define thepositive ideal point and the negative ideal point andcalculate the distance between each scheme and thepositivenegative ideal point Calculate the distance119878+
119894between each scheme and the positive ideal point
and the distance 119878minus119894between each scheme and the
negative ideal point
119878+
119894= radic
119899
sum
119895=1
(V119894119895minus V+119895)2
119894 isin 119898
119878minus
119894= radic
119899
sum
119895=1
(V119894119895minus Vminus119895)2
119894 isin 119898
(10)
(5) Calculate the relative approach degree 119862+119894of each
scheme to the ideal points
119862+
119894=
119878minus
119894
(119878+
119894+ 119878minus
119894) 0 lt 119862
+
119894lt 1 119894 isin 119898 (11)
(6) Ranke the schemes based on 119862+119894 The larger is the 119862+
119894
the better is the scheme
Above all the proposed PROMSIS combines the qualita-tive and quantitative analysis Preference comparison corre-sponds to the qualitative analysisThe Euclidean distance candescribe the degree of preference through the quantity Sothe utility of the network can be achieved in both qualitativeand quantitative aspects And the final decision will beappropriate based on subjective and objective factors
23 Weight Vector There are a variety of weight methodssuch as analytic hierarchy process (AHP) and the informationEntropy weight method some are subjective and others areobjective We can choose the appropriate weight methodaccording to actual conditions In the weight determiningprocess we apply the LS [9] to estimate the weights ofdecision elements introduced by Chu in 1979
Firstly the comparison matrix G119888= (119892119894119895)119899times119899
is definedaccording to the relative importance The judgments areranked on a 9-point scale [7] Numbers 1 to 9 are usedto represent equal weakly moderate moderate moderateplus strong strong plus very strong very very strong andextremely important to the objective respectively When anelement is less important than another the comparison resultequals the reciprocal of one of the numbers So for the matrixG119888to be the diagonal elements are observed 1 demonstrating
the elementsrsquo self-comparisons The other entries in thematrix are symmetric with respect to the diagonal as a resultof the inverted comparisons
Four traffic classes defined by 3GPP are taken into con-sideration namely the conversational streaming interactiveand background classes Based on the traffic requirementsthe comparison matrices for the four traffic classes accordingto the 9-point scale can be established
The element 119892119894119895of matrix G
119888shall be considered and
desired to determine the weights 119908119894 such that given 119892
119894119895
119892119894119895asymp 119908119894119908119895
4 International Journal of Distributed Sensor Networks
The weights can be obtained by solving the constrainedoptimization problem
Equations (15) and (13) form a set of 119899+1 inhomogeneouslinear equations with 119899 + 1 unknowns
By the way using the numerical method to solve mathe-matical problems due to almost inevitable rounding errorsthe results obtained are generally inaccurate Some othermeasures can be applied to estimate the error but we will notgo into detail here due to space limitations
3 Simulation Results
In this research we concentrate on the downlink traffic sinceit normally requires higher bandwidth than uplink especiallyfor multimedia services such as video streaming through theHSDPA channel while connected to WCDMA
The performance of different vertical handoff algorithmshas been evaluated with the scenario illustrated in Figure 1in which there are 7 BS and 12 AP placed at each WCDMAcell boundary The WCDMA cell radius is 1200 meter 200randomly generated UEs are used inside the simulation areawhose position changes in the time interval depending ontheir moving speed and directionThe direction is uniformlydistributed in the range of [0 2120587] and the speed change rateis 5 per 100 seconds The maximum userrsquos moving speedis 80 kmhour In the traffic generator module for a meansession duration of 60 seconds and a certain given meansession arrival rate user traffic is randomly generated witha Poisson arrival distribution As revealed in Figure 1 thedirection of the arrow represents the userrsquos moving directionThe length of the arrow corresponds to themoving distance ofthe user in 20 seconds (supposing that themoving direction isfixed in the 20 seconds) so the larger one indicates the fastermoving speed
The V-shape with indifference criterion type preferencefunction in PROMETHEE was adopted here In case of thistype the thresholds of indifference 119902 (119902 = 01) and strictpreference 119901 (119901 = 05) have to be separately selected
The system performance for different session arrivalrates is shown in Figure 2 The simulated algorithms include
the proposed LS-PROMSIS algorithm and the MASVH(119896 = 4) algorithm [6] The downlink system throughputof each algorithm is measured and shown in Figure 2(a) Itdemonstrates that the LS-PROMSIS algorithm for streamingtraffic class achieves the highest throughput performancebecause ldquothe available bandwidthrdquo attribute has the greatestweight in the handoff criterion so the network of the largestavailable bandwidth is selected considering the load balanceLikewise the system dropping probability of this algorithmis the lowest The dropping probability performance of eachalgorithm is exhibited in Figure 2(b) The average user trafficcost performance is presented in Figure 2(c) It is noted thatthe cost of the LS-PROMSIS algorithm for streaming trafficclass is quite high but the cost of the LS-PROMSIS algorithmfor conversational traffic is the lowest Since the attribute ofuser traffic cost covers the highest proportion in the handoffcriterion the network of the lowest cost is inclined to beselected Figure 2(d) indicates the number of vertical handoffIt is revealed that the number of vertical handoffs of the LS-PROMSIS algorithm for streaming traffic class is the highestand that of theMASVH algorithm finishes the second By theway if the unnecessary handoff needs to be further decreasedthe proposed MADM algorithm can be combined with themobile prediction technique to mitigate the impact of theping-pong effect
4 Conclusions
In this paper a novel PROMSIS vertical handoff algorithmis proposed and compared with the existing CSVH andMASVH algorithms The vertical handoff of heterogeneousnetworks is amultiple attributes decision subjectWith regardto the relations between all the attributes the observedobjects are the four 3GPP defined traffic classes Accordingto the features of diverse traffic classes the weight of eachattribute in the handoff criterion is determined by LS The
International Journal of Distributed Sensor Networks 5
001 0015 002 0025 003 0035 004 0045 00510
15
20
25
30
35
Session arrival rate (sessionss)
Thro
ughp
ut (M
bps)
(a) System throughput
001 0015 002 0025 003 0035 004 0045 0050
005
01
015
02
025
03
035
Session arrival rate (sessionss)
Syste
m d
ropp
ing
prob
abili
ty(b) System dropping probability
001 0015 002 0025 003 0035 004 0045 005042
0425
043
0435
044
0445
045
0455
046
Session arrival rate (sessionss)
Aver
age u
ser c
ost (
cost
SIN
R)
MASVH k = 4
PROMSIS conversationalPROMSIS streaming
PROMSIS interactivePROMSIS background
(c) Average cost to user traffic of the algorithms
001 0015 002 0025 003 0035 004 0045 00520
40
60
80
100
120
140
160
180
Session arrival rate (sessionss)
The n
umbe
r of h
ando
ff (n
umbe
rsim
ulat
ion)
MASVH k = 4
PROMSIS conversationalPROMSIS streaming
PROMSIS interactivePROMSIS background
(d) The number of vertical handoffs
Figure 2 Performance of each algorithm
simulation results display that the performance of the algo-rithm is affected by the allocated weight vector Consequentlyin practice we should consider both the characteristics of thetraffic and the preference of the user andweigh the advantagesand disadvantages before making the decision According tothe analysis and simulation results the PROMSIS algorithm
can achieve the satisfactory performance for the network andthe user
For future work more comparisons with other verti-cal handoff methods will be further discussed and othertechniques to solve the decision problem such as the gametheory will also be taken into account
6 International Journal of Distributed Sensor Networks
Acknowledgments
This work was supported by the National Natural ScienceFoundation of China under Contract no 61271235 and Nat-ural Science Foundation of Education Committee of JiangsuProvince (no 11KJB510014)
References
[1] A Singhrova and N Prakash ldquoVertical handoff decision algo-rithm for improved quality of service in heterogeneous wirelessnetworksrdquo IET Communications vol 6 no 2 pp 211ndash223 2012
[2] F Kaleem and A Mehbodniya ldquoDynamic target wireless net-work selection technique using fuzzy linguistic variablesrdquoChinaCommunications vol 10 no 1 pp 1ndash16 2013
[3] K Yang I Gondal B Qiu and L S Dooley ldquoCombined SINRbased vertical handoff algorithm for next generation hetero-geneous wireless networksrdquo in Proceedings of the 50th AnnualIEEEGlobal Telecommunications Conference (GLOBECOM rsquo07)pp 4483ndash4487 Washington DC USA November 2007
[4] K Yang I Gondal and B Qiu ldquoMulti-dimensional adaptiveSINR based vertical handoff for heterogeneous wireless net-worksrdquo IEEE Communications Letters vol 12 no 6 pp 438ndash440 2008
[5] L Bin and L Shengmei ldquoVertical handoff algorithm based onmobility predictionrdquoApplication of Electronic Technique vol 39no 1 pp 93ndash95 2013
[6] R O Parreiras J H R D Maciel and J A Vasconcelos ldquoThea posteriori decision in multiobjective optimization problemswith smarts promethee II and a fuzzy algorithmrdquo IEEETransactions on Magnetics vol 42 no 4 pp 1139ndash1142 2006
[7] J P Brans P Vincke and B Mareschal ldquoHow to select and howto rank projects the Promethee methodrdquo European Journal ofOperational Research vol 24 no 2 pp 228ndash238 1986
[8] E Stevens-Navarro and V W S Wong ldquoComparison betweenvertical handoff decision algorithms for heterogeneous wirelessnetworksrdquo in Proceedings of the IEEE 63rd Vehicular TechnologyConference (VTC-Sping rsquo06) pp 947ndash951Melbourne AustraliaMay 2006
[9] A T W Chu R E Kalaba and K Spingarn ldquoA comparison oftwo methods for determining the weights of belonging to fuzzysetsrdquo Journal of Optimization Theory and Applications vol 27no 4 pp 531ndash538 1979
Equations (15) and (13) form a set of 119899+1 inhomogeneouslinear equations with 119899 + 1 unknowns
By the way using the numerical method to solve mathe-matical problems due to almost inevitable rounding errorsthe results obtained are generally inaccurate Some othermeasures can be applied to estimate the error but we will notgo into detail here due to space limitations
3 Simulation Results
In this research we concentrate on the downlink traffic sinceit normally requires higher bandwidth than uplink especiallyfor multimedia services such as video streaming through theHSDPA channel while connected to WCDMA
The performance of different vertical handoff algorithmshas been evaluated with the scenario illustrated in Figure 1in which there are 7 BS and 12 AP placed at each WCDMAcell boundary The WCDMA cell radius is 1200 meter 200randomly generated UEs are used inside the simulation areawhose position changes in the time interval depending ontheir moving speed and directionThe direction is uniformlydistributed in the range of [0 2120587] and the speed change rateis 5 per 100 seconds The maximum userrsquos moving speedis 80 kmhour In the traffic generator module for a meansession duration of 60 seconds and a certain given meansession arrival rate user traffic is randomly generated witha Poisson arrival distribution As revealed in Figure 1 thedirection of the arrow represents the userrsquos moving directionThe length of the arrow corresponds to themoving distance ofthe user in 20 seconds (supposing that themoving direction isfixed in the 20 seconds) so the larger one indicates the fastermoving speed
The V-shape with indifference criterion type preferencefunction in PROMETHEE was adopted here In case of thistype the thresholds of indifference 119902 (119902 = 01) and strictpreference 119901 (119901 = 05) have to be separately selected
The system performance for different session arrivalrates is shown in Figure 2 The simulated algorithms include
the proposed LS-PROMSIS algorithm and the MASVH(119896 = 4) algorithm [6] The downlink system throughputof each algorithm is measured and shown in Figure 2(a) Itdemonstrates that the LS-PROMSIS algorithm for streamingtraffic class achieves the highest throughput performancebecause ldquothe available bandwidthrdquo attribute has the greatestweight in the handoff criterion so the network of the largestavailable bandwidth is selected considering the load balanceLikewise the system dropping probability of this algorithmis the lowest The dropping probability performance of eachalgorithm is exhibited in Figure 2(b) The average user trafficcost performance is presented in Figure 2(c) It is noted thatthe cost of the LS-PROMSIS algorithm for streaming trafficclass is quite high but the cost of the LS-PROMSIS algorithmfor conversational traffic is the lowest Since the attribute ofuser traffic cost covers the highest proportion in the handoffcriterion the network of the lowest cost is inclined to beselected Figure 2(d) indicates the number of vertical handoffIt is revealed that the number of vertical handoffs of the LS-PROMSIS algorithm for streaming traffic class is the highestand that of theMASVH algorithm finishes the second By theway if the unnecessary handoff needs to be further decreasedthe proposed MADM algorithm can be combined with themobile prediction technique to mitigate the impact of theping-pong effect
4 Conclusions
In this paper a novel PROMSIS vertical handoff algorithmis proposed and compared with the existing CSVH andMASVH algorithms The vertical handoff of heterogeneousnetworks is amultiple attributes decision subjectWith regardto the relations between all the attributes the observedobjects are the four 3GPP defined traffic classes Accordingto the features of diverse traffic classes the weight of eachattribute in the handoff criterion is determined by LS The
International Journal of Distributed Sensor Networks 5
001 0015 002 0025 003 0035 004 0045 00510
15
20
25
30
35
Session arrival rate (sessionss)
Thro
ughp
ut (M
bps)
(a) System throughput
001 0015 002 0025 003 0035 004 0045 0050
005
01
015
02
025
03
035
Session arrival rate (sessionss)
Syste
m d
ropp
ing
prob
abili
ty(b) System dropping probability
001 0015 002 0025 003 0035 004 0045 005042
0425
043
0435
044
0445
045
0455
046
Session arrival rate (sessionss)
Aver
age u
ser c
ost (
cost
SIN
R)
MASVH k = 4
PROMSIS conversationalPROMSIS streaming
PROMSIS interactivePROMSIS background
(c) Average cost to user traffic of the algorithms
001 0015 002 0025 003 0035 004 0045 00520
40
60
80
100
120
140
160
180
Session arrival rate (sessionss)
The n
umbe
r of h
ando
ff (n
umbe
rsim
ulat
ion)
MASVH k = 4
PROMSIS conversationalPROMSIS streaming
PROMSIS interactivePROMSIS background
(d) The number of vertical handoffs
Figure 2 Performance of each algorithm
simulation results display that the performance of the algo-rithm is affected by the allocated weight vector Consequentlyin practice we should consider both the characteristics of thetraffic and the preference of the user andweigh the advantagesand disadvantages before making the decision According tothe analysis and simulation results the PROMSIS algorithm
can achieve the satisfactory performance for the network andthe user
For future work more comparisons with other verti-cal handoff methods will be further discussed and othertechniques to solve the decision problem such as the gametheory will also be taken into account
6 International Journal of Distributed Sensor Networks
Acknowledgments
This work was supported by the National Natural ScienceFoundation of China under Contract no 61271235 and Nat-ural Science Foundation of Education Committee of JiangsuProvince (no 11KJB510014)
References
[1] A Singhrova and N Prakash ldquoVertical handoff decision algo-rithm for improved quality of service in heterogeneous wirelessnetworksrdquo IET Communications vol 6 no 2 pp 211ndash223 2012
[2] F Kaleem and A Mehbodniya ldquoDynamic target wireless net-work selection technique using fuzzy linguistic variablesrdquoChinaCommunications vol 10 no 1 pp 1ndash16 2013
[3] K Yang I Gondal B Qiu and L S Dooley ldquoCombined SINRbased vertical handoff algorithm for next generation hetero-geneous wireless networksrdquo in Proceedings of the 50th AnnualIEEEGlobal Telecommunications Conference (GLOBECOM rsquo07)pp 4483ndash4487 Washington DC USA November 2007
[4] K Yang I Gondal and B Qiu ldquoMulti-dimensional adaptiveSINR based vertical handoff for heterogeneous wireless net-worksrdquo IEEE Communications Letters vol 12 no 6 pp 438ndash440 2008
[5] L Bin and L Shengmei ldquoVertical handoff algorithm based onmobility predictionrdquoApplication of Electronic Technique vol 39no 1 pp 93ndash95 2013
[6] R O Parreiras J H R D Maciel and J A Vasconcelos ldquoThea posteriori decision in multiobjective optimization problemswith smarts promethee II and a fuzzy algorithmrdquo IEEETransactions on Magnetics vol 42 no 4 pp 1139ndash1142 2006
[7] J P Brans P Vincke and B Mareschal ldquoHow to select and howto rank projects the Promethee methodrdquo European Journal ofOperational Research vol 24 no 2 pp 228ndash238 1986
[8] E Stevens-Navarro and V W S Wong ldquoComparison betweenvertical handoff decision algorithms for heterogeneous wirelessnetworksrdquo in Proceedings of the IEEE 63rd Vehicular TechnologyConference (VTC-Sping rsquo06) pp 947ndash951Melbourne AustraliaMay 2006
[9] A T W Chu R E Kalaba and K Spingarn ldquoA comparison oftwo methods for determining the weights of belonging to fuzzysetsrdquo Journal of Optimization Theory and Applications vol 27no 4 pp 531ndash538 1979
International Journal of Distributed Sensor Networks 5
001 0015 002 0025 003 0035 004 0045 00510
15
20
25
30
35
Session arrival rate (sessionss)
Thro
ughp
ut (M
bps)
(a) System throughput
001 0015 002 0025 003 0035 004 0045 0050
005
01
015
02
025
03
035
Session arrival rate (sessionss)
Syste
m d
ropp
ing
prob
abili
ty(b) System dropping probability
001 0015 002 0025 003 0035 004 0045 005042
0425
043
0435
044
0445
045
0455
046
Session arrival rate (sessionss)
Aver
age u
ser c
ost (
cost
SIN
R)
MASVH k = 4
PROMSIS conversationalPROMSIS streaming
PROMSIS interactivePROMSIS background
(c) Average cost to user traffic of the algorithms
001 0015 002 0025 003 0035 004 0045 00520
40
60
80
100
120
140
160
180
Session arrival rate (sessionss)
The n
umbe
r of h
ando
ff (n
umbe
rsim
ulat
ion)
MASVH k = 4
PROMSIS conversationalPROMSIS streaming
PROMSIS interactivePROMSIS background
(d) The number of vertical handoffs
Figure 2 Performance of each algorithm
simulation results display that the performance of the algo-rithm is affected by the allocated weight vector Consequentlyin practice we should consider both the characteristics of thetraffic and the preference of the user andweigh the advantagesand disadvantages before making the decision According tothe analysis and simulation results the PROMSIS algorithm
can achieve the satisfactory performance for the network andthe user
For future work more comparisons with other verti-cal handoff methods will be further discussed and othertechniques to solve the decision problem such as the gametheory will also be taken into account
6 International Journal of Distributed Sensor Networks
Acknowledgments
This work was supported by the National Natural ScienceFoundation of China under Contract no 61271235 and Nat-ural Science Foundation of Education Committee of JiangsuProvince (no 11KJB510014)
References
[1] A Singhrova and N Prakash ldquoVertical handoff decision algo-rithm for improved quality of service in heterogeneous wirelessnetworksrdquo IET Communications vol 6 no 2 pp 211ndash223 2012
[2] F Kaleem and A Mehbodniya ldquoDynamic target wireless net-work selection technique using fuzzy linguistic variablesrdquoChinaCommunications vol 10 no 1 pp 1ndash16 2013
[3] K Yang I Gondal B Qiu and L S Dooley ldquoCombined SINRbased vertical handoff algorithm for next generation hetero-geneous wireless networksrdquo in Proceedings of the 50th AnnualIEEEGlobal Telecommunications Conference (GLOBECOM rsquo07)pp 4483ndash4487 Washington DC USA November 2007
[4] K Yang I Gondal and B Qiu ldquoMulti-dimensional adaptiveSINR based vertical handoff for heterogeneous wireless net-worksrdquo IEEE Communications Letters vol 12 no 6 pp 438ndash440 2008
[5] L Bin and L Shengmei ldquoVertical handoff algorithm based onmobility predictionrdquoApplication of Electronic Technique vol 39no 1 pp 93ndash95 2013
[6] R O Parreiras J H R D Maciel and J A Vasconcelos ldquoThea posteriori decision in multiobjective optimization problemswith smarts promethee II and a fuzzy algorithmrdquo IEEETransactions on Magnetics vol 42 no 4 pp 1139ndash1142 2006
[7] J P Brans P Vincke and B Mareschal ldquoHow to select and howto rank projects the Promethee methodrdquo European Journal ofOperational Research vol 24 no 2 pp 228ndash238 1986
[8] E Stevens-Navarro and V W S Wong ldquoComparison betweenvertical handoff decision algorithms for heterogeneous wirelessnetworksrdquo in Proceedings of the IEEE 63rd Vehicular TechnologyConference (VTC-Sping rsquo06) pp 947ndash951Melbourne AustraliaMay 2006
[9] A T W Chu R E Kalaba and K Spingarn ldquoA comparison oftwo methods for determining the weights of belonging to fuzzysetsrdquo Journal of Optimization Theory and Applications vol 27no 4 pp 531ndash538 1979
6 International Journal of Distributed Sensor Networks
Acknowledgments
This work was supported by the National Natural ScienceFoundation of China under Contract no 61271235 and Nat-ural Science Foundation of Education Committee of JiangsuProvince (no 11KJB510014)
References
[1] A Singhrova and N Prakash ldquoVertical handoff decision algo-rithm for improved quality of service in heterogeneous wirelessnetworksrdquo IET Communications vol 6 no 2 pp 211ndash223 2012
[2] F Kaleem and A Mehbodniya ldquoDynamic target wireless net-work selection technique using fuzzy linguistic variablesrdquoChinaCommunications vol 10 no 1 pp 1ndash16 2013
[3] K Yang I Gondal B Qiu and L S Dooley ldquoCombined SINRbased vertical handoff algorithm for next generation hetero-geneous wireless networksrdquo in Proceedings of the 50th AnnualIEEEGlobal Telecommunications Conference (GLOBECOM rsquo07)pp 4483ndash4487 Washington DC USA November 2007
[4] K Yang I Gondal and B Qiu ldquoMulti-dimensional adaptiveSINR based vertical handoff for heterogeneous wireless net-worksrdquo IEEE Communications Letters vol 12 no 6 pp 438ndash440 2008
[5] L Bin and L Shengmei ldquoVertical handoff algorithm based onmobility predictionrdquoApplication of Electronic Technique vol 39no 1 pp 93ndash95 2013
[6] R O Parreiras J H R D Maciel and J A Vasconcelos ldquoThea posteriori decision in multiobjective optimization problemswith smarts promethee II and a fuzzy algorithmrdquo IEEETransactions on Magnetics vol 42 no 4 pp 1139ndash1142 2006
[7] J P Brans P Vincke and B Mareschal ldquoHow to select and howto rank projects the Promethee methodrdquo European Journal ofOperational Research vol 24 no 2 pp 228ndash238 1986
[8] E Stevens-Navarro and V W S Wong ldquoComparison betweenvertical handoff decision algorithms for heterogeneous wirelessnetworksrdquo in Proceedings of the IEEE 63rd Vehicular TechnologyConference (VTC-Sping rsquo06) pp 947ndash951Melbourne AustraliaMay 2006
[9] A T W Chu R E Kalaba and K Spingarn ldquoA comparison oftwo methods for determining the weights of belonging to fuzzysetsrdquo Journal of Optimization Theory and Applications vol 27no 4 pp 531ndash538 1979
![QoS Aware Fuzzy Rule Based Vertical Handoff Decision Algorithm for Wireless ... considered for making a decision in handoff [4]. A multi-criteria based handoff has a greater advantage](https://static.documents.pub/doc/80x56/5fd948332f0ef86c037208bb/qos-aware-fuzzy-rule-based-vertical-handoff-decision-algorithm-for-wireless-.jpg)
![Evolutive Algorithm for Spectral Handoff Prediction in ... · Evolutive algorithm for spectral handoff prediction 675 feedback analytic hierarchy process algorithm (FFAHP) [14]. The](https://static.documents.pub/doc/80x56/5b36da147f8b9ab9068b9541/evolutive-algorithm-for-spectral-handoff-prediction-in-evolutive-algorithm.jpg)
