You are on page 1of 6

The use of MADM methods in the vertical handover

decision making context

MANSOURI Mouâd, LEGHRIS Cherkaoui


Department of computer sciences
FST, L@M Lab, RTM team
Hassan II university, Casablanca, Morocco
mansouri.mouad@yahoo.com, cleghris@yahoo.fr

Abstract—The vertical handover is an important issue in the good QoS), and fast enough, to not interrupt the active
world of internet. In fact, the use of internet is becoming wider sessions. This supposes to make very fast decisions, every
with the diversity of possible offered services, going from e- time, to decide which available connectivity is the best, and
mailing, Data and video transfers, to IoT. The number of internet then, ensure a fast communication transfers to a new wireless
users has reached multimillions of people these last two decades, technology, if this one is the same, the transfer of
in addition, they are willing to connect wirelessly anytime and communication is called horizontal handover (Ex: From two
anywhere, to the best available technology, under the Always base stations offering 3G connectivity), else, if the chosen
Best-Connected paradigm (ABC). The mobile terminals are technology type is different (From 3G connection to Wi-Fi
nowadays, equipped with multiple interfaces, allowing them, to
hotspot), it’s called vertical handover (VH).
connect to any available wireless access technology, and this
provokes vertical handovers to happen. The vertical handover is The VH process can be divided up into three major steps:
the changing of the wireless access technology (Ex: from Wi-Fi to
4G), unlike the horizontal handover, which means changing the  VH initiation, or discovery, where the MT discovers all
base station or access point, while staying within the same the available wireless networks in every decision point,
wireless technology. Thus, the vertical handover process needs to and collects data and metrics about them;
be improved, especially the decision phase, which is the most
crucial, while the sensors and antennas’ technologies are  VH decision making, The MT computes the global
developing very fast, to collect more efficient metrics, to make score for each available network, considering different
faster and more precise decisions anytime. attributes such as the delay, the bandwidth, the
Many theories have been used to deal with the vertical throughput, etc.
handover decision making, like utility function, fuzzy theory,
 VH execution: After the scores are computed and
neural networks, artificial intelligence, and Multi-Attributes
Decision Making (MADM). Naturally, MADM methods are the compared, and the best alternative is chosen, the MT
most suitable for this kind of decision making issues. Moreover, proceeds to the VH executing, transferring the
these algorithms are relatively simple to implement, and are communication to the new different wireless
executing fast enough, to ensure a “seamless” handover, without technology.
interrupting the services and sessions. The aim of this paper is to The main step of this process is the second one, because in
investigate which MADM combination is relatively the best, this phase, the MT computes the data gathered from the
amongst 120 combinations of MADM methods compared.
discovery phase by sensing different metrics about every
Keywords— MADM methods; Network selection; Vertical available wireless network. Then, it makes the decision,
handover; Wireless access networks; whether to perform a VH or not. Moreover, this process needs
to be fast, not to cause a session or service rupture, that’s why
we decided to research in improving the decision making
I. INTRODUCTION because it is the phase that takes the most of the global duration
The number of internet users has been exponentially of the VH, and we found that there is no best method ever to
increasing these last two decades; with the fast development in deal with this kind of problems, and that many theories have
the fields of electronic material, computation techniques, and been adapted to this context, such as the Fuzzy theory, gaming
networking technologies, mobile terminals (MTs) are now such theory, neural networks theory, and MADM theory methods,
sophisticated, that they are allowing users to connect wirelessly and those are perfectly adapted, by their nature to this kind of
to the best available wireless access technology anywhere and decision problems, and are fast enough to ensure the session
anytime, without interrupting the services either e-mailing, data and services continuity, which is called a “seamless VH”.
and voice transfers, or cloud computing and content streaming,
The VH decision was standardized in 2009 by the IEEE, as
which require a good quality of service (QoS), to be run.
“Media Independent Handover” MIH (IEEE 802.21), but no
To achieve this goal, networking and computation precise method is applied to make the decision, it only gives
techniques must be improved, to be efficient enough (offering a the general messages that can be sent within the MIHF (MIH

978-1-5386-2123-3/17/$31.00 ©2017 IEEE

Authorized licensed use limited to: National Ilan Univ.. Downloaded on March 04,2021 at 15:51:02 UTC from IEEE Xplore. Restrictions apply.
Function), then the transfers of communication after the exponent weighting (MEW), simple additive weighting
decision is made, is ensured by Mobile IPv6 (MIPv6), Fast (SAW), technique for order preferences by similarity to ideal
Handover for Mobile IPv6 (FHMIPv6), etc. The VH decision Solution (TOPSIS), and grey relational analysis (GRA) are
is made either from the network side “Network Controlled compared using different normalization methods and weighting
Handover” (NCH), or the mobile side (MCHO), which can be algorithms. In [7], propose a novel network selection algorithm
Network Assisted (NAHO); Network stations communicate for a heterogeneous environment consisting of WiMAX and
information about networks and link states, and the decision is LTE technologies, based on the scores obtained for each
made in the mobile terminal (MT) side. In our case, we treat alternative using a cost function of weighted criteria.
the Mobile Controlled Network Assisted Handover
(MCNAHO). Different works have been published, using a modified
MADM method. In [8], M. Lahby et al. provide an optimized
The second section of this paper will present the network selection decision that allow mobile users to choose
background of the VH decision problem, the third section will the best available network, in a seamless manner and to exploit
be about our approach to handle this problem with brief a minimum of criteria for all traffic classes using ANP and
mathematic description of some MADM methods used in our MAHALANOBIS distance to compute the distance between
simulation, the fourth section will give illustrations of the the alternatives. M. M. Alkhawlani and K. A. Alsalem in [9],
obtained results and discuss them, and finally, we will present a multi criteria radio network selection that considers
conclude about the best relative MADM method, or an environment with a co-existed WWAN, WMAN, and
combination of methods, that fits the best, with the VH context. WLAN. The developed solution contains two modules. The
first one contains a network-assisted terminal-controlled
II. BACKGROUND algorithm to reflect the user’s point of view in the selection
process. The second module contains a terminal assisted
We found in the literature many works dealing with the VH network-controlled algorithm to reflect the operator point of
decision making (Network selection), using different kinds of view of the selection decision, using MADM.
algorithms and applying many theories, going from basic
methods based on one single criterion (RSS, Bandwidth, etc.), There exist also hybrid methods, combining two or more
cost function methods, and multi-attribute utility functions, to different technologies (MADM methods and other theory), like
the fuzzy logic, neural networks, artificial intelligence, gaming in [10], where K. Radhika, and A. V. Reddy propose a network
theory, and the MADM methods, which can take into account selection algorithm based on Fuzzy Multiple Attribute
many criteria at the same time. Decision Making, that takes into account different criteria of
selection, and computes a network selection function that
MADM methods are widely used in this context of VH, but calculates the efficiency and utility in handing off to a new
they are also applied in many other decision-making situations wireless technology. In [11], R. K. GOYAL et al. used extent
going from leader selection problems and choosing the best car analysis on fuzzy analytic hierarchy process (FAHP) is applied
to buy, to complicated economic and R&D problems. These to select the Radio access technology (RAT), among multiple
methods are by nature, the best adapted to the VH decision RATs by calculating and comparing the values of the fuzzy
context, because it is a multi-criteria decision-making problem, synthetic extent of each available wireless network. In [12], M.
where we have to determine the best alternative, from a set of El Helou’s hybrid approach involves two inter-dependent
distinct alternatives, taking in account many criteria together. decision-making processes. The first one is on the network
MADM methods are used in different decision-making side, it consists in exploring network information so as to guide
contexts, by computing the values gathered from any system, user decisions in a way to meet operator objectives. The second
and giving scores to each alternative (seen all the criteria), they one, consists in selecting the RAT to be associated with in a
all follow the same process whose phases are: way to maximize user utility by combining their needs and
 Normalization: Data gathered are normalized to be preferences. Moreover, this proposal uses introduce two
computable and aggregated together. heuristic methods, namely the staircase and the slope tuning
policies, to dynamically collect network information, to
 Weighting: This phase can be treated independently, in increase the efficiency.
which different weights are assigned to each criterion,
reflecting its relative importance towards the other III. OUR PROPOSAL
criteria.
We extended our previous work in [6], by adding new
 Ranking: Alternatives are ranked following the score combinations based on Distance to Ideal Alternative (DIA)
that obtained each one, after applying the ranking method, and using different normalization and weighting
algorithm on the normalized and weighted data; the methods, namely Euclidian, Sum, Product, Max, and Min-Max
alternative with the best score is chosen. with two variants as normalization methods, and AHP, ANP,
In [5], Viji R. and Rathipriya R. used MADM methods to FAHP, and fuzzy ANP, as weighting methods. These new
compute the best score alternative using multi-criteria vertical combinations are compared with the results obtained using the
handover decision metrics, the proposed solution’s aim is to combinations based on TOPSIS, GRA, MEW, SAW, and DIA.
choose the best network from the available networks and to The steps of the VH decision making using different MADM
reduce the number of unnecessary VHs. In [6] different VH methods combinations are illustrated in figure 1, to explain
decision algorithms are compared, namely multiplicative when are used every type of MADM methods.

Authorized licensed use limited to: National Ilan Univ.. Downloaded on March 04,2021 at 15:51:02 UTC from IEEE Xplore. Restrictions apply.
C. The third step is to weight the criteria using equation (7),
to have the V matrix:
vij  rij *W j (7)

D. The final step is to compute the score for each alternative,


and to rank them by score, we give two examples in the
following equations (8) to (11) of TOPSIS and GRA
methods:
1) TOPSIS:
Basically, TOPSIS uses the Euclidian normalization
method, and its process begins after constructing the V matrix
as follows:
 Defining the best and worst alternative:
V+= max {Vij}; for benefice criteria, and min {Vij}; for
Fig. 1. An organigram showing MADM methods proccessing steps in the cost criteria.
VH context.
V-= min {Vij}; for benefice criteria, and max {Vij}; for
We give in the following some mathematic equations of cost criteria.
MADM methods, as treated in this paper:
 Compute the distances between every vij and the best
and the worst alternative respectively:
A. The first step is to construct a decision matrix D, with dij
is the value of attribute ith for the jth technology, gathered n
in the discovery phase, the D matrix is represented in (1): S i   (V
i 1

 vij ) 2 (8)

 d11  d1n 
 
Dmn      (1)
n

d 
S i   (v ij  V  )2 (9)

 m1  d mn 
i 1

 Compute the final score for each alternative:


B. The second thing to do is to normalize the D matrix, using
one of the methods (2) to (6): Si
Ci  (10)
dij Si  Si
 Euclidian method: rij  (2)
m
 Finally, the scores are ranked in a decreasing order.
d
i 1
2
ij
2) GRA:
Basically, GRA uses the min max normalization method,
dij its process after creating the V matrix:
 Sum method: rij  m
(3)
 Defining the best alternative: Same as TOPSIS.
 dij
i 1  Compute the GRA coefficient:

dij 1
 Product method: rij  (4) GRCi  n
(11)
m

d 2
ij v
i 1
ij

 Vi  1
i 1

 Min-Max method:  Rank the GRCi in decreasing order.


We designed a dataset containing a simulation of the
dij gathered data, every time along the trajectory of a designed
For benefit criteria: rij  (5)
max  d j  movement scenario of a MT, described in [13], and then we
applied the different combinations of MADM algorithms on
the same dataset, to compute every time, the returned scores.
min  d j  Finally, we compare the results obtained by using different
For cost criteria: rij  (6) MADM combinations during the whole movement, in terms of
dij the offered quality of service (QoS), the number of VHs

Authorized licensed use limited to: National Ilan Univ.. Downloaded on March 04,2021 at 15:51:02 UTC from IEEE Xplore. Restrictions apply.
occurred, and the number of rank reversals. The number of the simulations with eight networks, and we compared the
VHs is increased with every change of the wireless technology, results). Finally, we evaluate the results obtained in terms of
and the number of rank reversals is calculated by redoing the delivered QoS, the number of VHs, and the rank reversals.
simulation while taking off the worst alternative, and seeing if
the best wireless network chosen is still the same.
For our simulation, we considered nine wireless network
technologies: GPRS, EDGE, UMTS, HSxPA, LTE, Wi-Fi
a/b/g, Wi-Fi n, Wi-Fi ac, and WiMAX, because these are the
most used and omnipresent technologies. These wireless
networks are judged, taking into account eight different
criteria: The throughput (T), the network availability (Av), the
security (S), the delay (D), the losses (L), the cost (C), the
energy consumption (EC), and the jitter (J). We will add some
other criteria in the future, like the MT velocity or speed, or the
battery level every time. The values used in the simulation are
taken uniformly from the table. 1 below. Fig. 2. Record of decisions made by applying FAHP/FANP with the
Euclidian normalization, and TOPSIS.
TABLE I. VALUES OF ATTRIBUTES USED TO SIMULATE THE DATA
GATHERED DURING THE MOVEMENT

NETWORK F (KBPS) AV(%) S (%) D (MS) L EC* C* J (MS)


GPRS 21.4 - 171.2 50-100 50 50-70 50-80 2 1 3 – 20
EDGE 43.2 - 345.6 40-100 50 20-60 25-70 2 2 3 – 20
UMTS 144 - 2000 40-100 60 20-40 15-65 4 4 3 – 20
HSDPA 14 MBPS 50-100 60 10-50 10-80 5 5 3 – 20
LTE 10 - 300 MBPS 40-100 65 10-30 10-40 7 7 3 – 20
WIFI A,B,G 8 - 54 MBPS 40-100 60 130–200 30-70 3 1 3 – 20
WIFI N 72 - 450 MBPS 30-100 65 100–140 20-60 4 1 3 – 20
WIFI AC 433 - 1300 MBPS 50-100 70 90–110 10-40 5 2 3 – 20
WIMAX 70 MBPS 40-100 60 60-100 10-70 7 5 3 – 20
a.
Attributes in green are of benefit, those in red are of cost.
We distinguish between eight different zones of decision Fig. 3. Record of decisions made by applying FAHP/FANP with the Sum
normalization, and GRA.
during the MT movement, every decision zone is represented
by 22 decision points; so as to simulate the changes in the
attributes values in the same heterogeneous zone. We made
different functions for the different methods, in order to
compare 6 normalization methods, 4 weighting methods, and 5
ranking methods, what makes: 6
C 1 * C 41 * C 51  120 possible
combinations of MADM methods. An interesting way of
further researches is to compare more combinations in this
same context.

IV. RESULTS AND DISCUSSION

A. Results: Fig. 4. Record of decisions made by applying AHP/ANP with the Sum
After the simulations were run we recorded the chosen normalization, and GRA.
wireless network at every decision point during the movement
scenario. The results were captured for every combination of
MADM methods. The figures 2 and 3 show the records of VH
decisions made by the best combinations, compared to some
other less efficient combinations shown in figures 4, 5 and 6, in
order to have an idea about the QoS offered in every decision
zone and the stability of decisions made by these different
methods. These results are obtained using the same dataset, so
the more stable results are the best in this case.
The figures 8 and 9 show respectively the total number of
VHs and the number of rank reversals occurred using the
different combinations, which means the change of the VH
decisions when changing the number of alternatives (We rerun Fig. 5. Record of decisions made by applying FAHP/FANP with the Max
normalization, and TOPSIS.

Authorized licensed use limited to: National Ilan Univ.. Downloaded on March 04,2021 at 15:51:02 UTC from IEEE Xplore. Restrictions apply.
GRA and Sum produces more rank reversals (6). The other
combinations are less satisfying either in terms of the delivered
QoS, or they made a large number of VHs and rank reversals.
We still have to optimize more the VH decision making by
combining MADM methods with fuzzy logic, as an example,
the fuzzy logic is designed to deal with the uncertainties in the
metrics and judgements. The decision should take into account
more important criteria, such as the battery level or the MT
velocity.

V. CONCLUSION
Fig. 6. Record of decisions made by applying FAHP/FANP with the In this paper, we presented a comparison of 120 MADM
Euclidian normalization, and DIA..
methods in the context of the VH decision making. We
designed a simulation scenario containing different decision
zones, containing every time, different wireless technologies to
be compared. We generated data to simulate the data gathered
in the discovery phase of the VH, and we made a separated
module for each MADM method, so as to try all the possible
combinations. We recorded the results obtained by applying
every MADM method combination, and compared them in
terms of the number of VHs, the rank reversals, and the QoS
delivered in every decision zone. We concluded that FANP-
TOPSIS-Euclidian normalization is relatively the best
combination of MADM methods, followed by FAHP-TOPSIS-
Euclidian normalization, FANP-GRA-Sum normalization, and
FAHP-GRA-Sum normalization as the four relatively best
MADM combinations in the VH context.
There is a lot of other MADM methods to be compared also
Fig. 7. Number of VHs occurred using different MADM combination.
with these methods, to see if there is a better efficient
combination of MADM methods. Other combinations between
different technologies are to be tried in this same context, to try
to find a hybrid decision making method, in the context of VH
decision.

REFERENCES

[1] Tarek Bouali and Sidi-Mohammed Senouci. “A Fuzzy Logic-Based


Communication Medium Selection for QoS Preservation in Vehicular
Networks”. IEEE ICC 2016 - Next-Generation Networking and Internet
Symposium.
[2] J. Marquez-Barja, C. T. Calafate, J-C. Cano, and P. Manzoni. MACHU.
“A novel vertical handover algorithm for vehicular environments”.
Wireless Telecommunications Symposium, April 2012..
Fig. 8. Number of rank reversals occurred using different MADM [3] A. De La Oliva, A. Banchs, I. Soto, T. Melia, and A. Vidal. “An
combination. overview of ieee 802.21: media-independent handover services.
Wireless Communications, IEEE, Aug 2008.
[4] Amali Chinnappan, Ramachandran Balasubramanian. “Complexity–
B. Discussion: consistency trade-off in multi-attribute decision making for vertical
handover in heterogeneous wireless networks”. IET Networks, 2016, Vol. 
As shown in the records, we observed that FANP with 5, Iss. 1, pp. 13‒21.
TOPSIS using the Euclidian normalization is relatively the best [5] Viji R, Rathipriya R. The Network Selection Using for Seamless
amongst the compared ones, taking into consideration three Connectivity of Heterogeneous Networks Based on MADM Methods.
decision evaluation parameters, namely the number of VHs, the Indian Journal of Engineering, 2015, 12(30), 380-387.
rank reversals and the QoS offered during the movement. with [6] M. Mansouri, C. Leghris, “Towards a better combination of the MADM
a good delivered QoS seen all the trajectory, with only 5 VHs, algorithms for the Vertical Handover optimization in a mobile network
and 4 rank abnormalities. Followed by FAHP with the same multi-access environment”. Intelligent Systems: Theories and
Applications (SITA), 2015 10th.
TOPSIS with a little bit less QoS delivered in the 8th decision
[7] Kiran Ahuja et al, “Optimal Network Selection in Heterogeneous
zone. FANP with GRA using the sum normalization comes in Wireless Environment for Multimedia Services”, Wireless Personal
the third place with the same number of VHs (5) and rank Communications. Feb. 2015.
reversals (4), but a little bit less efficient QoS, FAHP with

Authorized licensed use limited to: National Ilan Univ.. Downloaded on March 04,2021 at 15:51:02 UTC from IEEE Xplore. Restrictions apply.
[8] Lahby, M., Leghris, C., & Adib, A. “New Optimized Network Selection Networks”. 2016 UKSim-AMSS 18th International Conference on
Decision in Heterogeneous Wireless Networks”. International Journal of Computer Modelling and Simulation.
Computer Applications (0975 - 8887), Volume 54 - No. 16, September [15] Loveneet Kaur Johal and Amandeep Singh Sandhu,“ An Overview of
2012. Vertical Handover Process and Techniques”. Indian Journal of Science
[9] Mohammed M. Alkhawlani and Kasim A. Alsalem. “Radio Network and Technology, Vol 9(14), April 2016.
Selection for Tight-Coupled Wireless Networks”. Informatics and [16] Enoruwa Obayiuwana and Olabisi Emmanuel Falowo, “Network
Systems (INFOS), 2010 The 7th International Conference on. selection in heterogeneous wireless networks using multi-criteria
[10] K. Radhika, and Dr. A. Venugopal Reddy, “Network Selection in decision-making algorithms: a review”. Wireless Networks, pp 1-33,
Heterogeneous Wireless Networks based on Fuzzy Multiple Criteria 2016.
Decision Making”International Journal of Computer Applications (0975 [17] Dhanaraj Cheelu et al. “A Study of Vertical Handoff Decision Strategies
– 8887) Volume 22– No.1, May 2011. in Heterogeneous Wireless Networks”. International Journal of
[11] Raman Kumar Goyal, Sakshi Kaushal, and Sundarapandian Engineering and Technology (IJET). Vol 5 No 3 (Jun-Jul 2013).
Vaidyanathan. “Fuzzy AHP for Control of Data Transmission by [18] K.Savitha & C.Chandrasekar “Trusted Network Selection using SAW
Network Selection in Heterogeneous Wireless Networks”. I J C T A, and TOPSIS Algorithms for Heterogeneous Wireless Networks”.
9(1), 2016, pp. 133-140. International Journal of Computer Applications (0975 – 8887) Volume
[12] Melhem El Helou, Samer Lahoud, Marc Ibrahim, Kinda Khawam, 26– No.8, July 2011.
Bernard Cousin, et al.. “A Hybrid Approach for Radio Access [19] Liu, H.-T., & Wang, C.-H. (2010). An advanced quality function
Technology Selection in Heterogeneous Wireless Networks”. Wireless deployment model using fuzzy analytic network process. Applied
Personal Communications, Springer Verlag, 2015, 86 (2), pp.1-46. Mathematical Modelling, 34(11), 3333–3351.
[13] IEEE 802.21 MEDIA INDEPENDENT HANDOVER, IEEE P802.21 [20] Quang Hung Do1, Jeng-Fung Chen “Establishing the Index System for
Tutorial, July 17, (2006). Sustainable Urban Transport Project Selection: An Application of Group
[14] Ali F. Almutairi, Mohamed A. Landolsi, Aliaa O. Al-Hawaj. “Weighting MCDM based on the Fuzzy AHP Approach”, International Journal of
Selection in GRA-based MADM for Vertical Handover in Wireless Business and Management Invention. Volume 2 Issue 6, PP.47-57, June.
2013.

Authorized licensed use limited to: National Ilan Univ.. Downloaded on March 04,2021 at 15:51:02 UTC from IEEE Xplore. Restrictions apply.

You might also like