Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

2016 International Conference on Computational Techniques in Information and Communication Technologies (ICCTICT)

A Comparative Study of Cooperative and Non-


Cooperative Game Theory in Network Selection
Meenakshi N. P. Singh
Dept. of Electronics & Communication Engineering Dept. of Electronics & Communication Engineering
NIT Kurukshetra NIT Kurukshetra
Haryana, India Haryana, India
Email Id: meenakshimunjal206@gmail.com Email Id: nirajatnitkkr@gmail.com

Abstract —Fifth generation wireless network is consisting cost, propagation delay, bandwidth etc [3]. Therefore, network
of heterogeneous networks to provide end to end mobility selection depends on many attributes and it is solved by
and service to mobile users. So in heterogeneous wireless different techniques.
network, selection of the optimal network becomes an Multi Attributes Decision making (MADM) algorithms
important task. In recent years, network selection has been that have been used for ranking of a network include Simple
mostly studied by using different mathematical methods. Additive Weighting (SAW), Grey Relational Analysis (GRA),
But in this paper, the network selection problem is Technique for Order Preference by Similarity to Ideal Solution
modelled as the evolutionary and bankruptcy game in (TOPSIS), Multiplicative Exponential Weighting (MEW),
heterogeneous wireless networks. Analytical Hierarchy Elimination and Choice Translating Reality (ELECTRE) are
Process (AHP) is used to decide the weights of different discussed in [4].SAW, GRA, MEW and TOPSIS algorithm are
attributes according to service requirement and used for ranking of the networks and ranking depend upon the
evolutionary and bankruptcy game theory is used for coefficients which is calculated by combining adjusted values
ranking of the networks. Evolutionary game is used as the of all criteria. And ELECTRE algorithm uses the pairwise
non-cooperative game for studying the behavior of selfish comparison of all the networks. The author in [5] used the
users and bankruptcy game is used as the cooperative game AHP and grey relational analysis (GRA) methods for optimal
for achieving mutual advantage. The comparison between access network selection. In [6], combined fuzzy AHP and
the cooperative game and non-cooperative game is studied ELECTRE algorithm are discussed where fuzzy AHP is used
in case of streaming, conversational and background traffic for assigning the weight to the criteria and ELECTRE is used
class. Simulation results show that different networks are for ranking the candidate networks.
selected for different traffic class according to user In [7], the basic concept of game theory and different types
preference. of games are described. The author describes the use of game
theory in network selection. Game Theory based network
Keywords – Always Best connected (ABC), Analytic Hierarchy selection concept in heterogeneous wireless network is
Process (AHP), Heterogeneous Wireless Network (HWN), Multiple described in [8].
Attribute Decision Making (MADM) and Radio Access Technology Detail about evolutionary game theory is described in
(RAT). [9].The author use the dynamic evolutionary game for sharing
of bandwidth in Heterogeneous wireless networks. Network
I. INTRODUCTION selection using bankruptcy game theory is described in
In the past few years, there is an impressive growth in [10].The use of utility function in evolutionary game theory is
wireless communication systems due to the popularity of smart described in [11]. In all the literature, network selection is
phones and other mobile devices. Nowadays, there are various done for two or three network while in practical life, there are
types of Radio Access Technologies (RATs) and each of them large number of networks available. Secondary, there is no
supports different data rates, mobility, coverage, power and difference is given between cooperative and non-cooperative
Quality of Service (QoS). So in the multi-access network game theory in terms of network selection.
scenario, the users is able to select the most appropriate In this paper, six networks are considered i.e. Wi-Fi,
network from various heterogeneous access networks.
WiMAX, WLAN, LTE and two different UMTS networks. In
Deciding to which one mobile connect is known as the network
this heterogonous network, selection of the optimal network is
selection problem [1].
done by using evolutionary and bankruptcy game theory in
Network selection is an important part from consumer- case of streaming, conversational and background traffic class.
centric and network-centric view for wireless access services After that, a comparative study of evolutionary and
[2]. Moreover, the selection of an optimal network depends on bankruptcy game is described. Evolutionary game is used as
various parameter having different importance such as service non-cooperative game and bankruptcy game is used as the

978-1-5090-0082-1/16/$31.00 ©2016 IEEE


cooperative game. Next, the difference between two games is another player strategy then the solution of that game is called
described in terms of network selection. The weight of different Nash equilibrium. When the player cannot be further increased
attributes is measured by AHP method and then network its payoff with any strategy, then the game has its Pareto
selection analysis is done using evolutionary and bankruptcy optimal solution. If there are more than one equilibrium
game theory. candidate then the Pareto optimal solution are preferred [17].
The remainder of this paper is organized as follow: In It can be used for decision making between competitive
section II, basic introduction of network selection is given. In environment and cooperative environment. So depending on
the III section, the basic concept of game theory is given and this environment, there are two types of game i.e. cooperative
mapping of game theory to network selection problem is and non-cooperative game. In the cooperative game, players
described. Section IV describes the evolutionary game usage coordinate their strategies and share their payoff. In this type
for network selection and section V describes the bankruptcy of game, Shapley value gives the solution. Shapley value
game usage for network selection. In section VI, simulation provides the only one way to divide the total payoff in
results using evolutionary and bankruptcy game theory for between the different players for every coalition game [18]. In
different traffic classes are discussed. Finally, in section VII, the non-cooperative game, every player individually selects its
conclusion and future work directions are given. strategy. It is closely connected to minimax optimization and
the results is calculated by Nash equilibrium[19].
II. NETWORK SELECTION
Network selection is the critical issue in the vertical IV. EVOLUTIONARY GAME THEORY
handover process. It is a complex problem which involves Evolutionary game theory is the type of non-cooperative
several tasks so that the user is able to select the optimal game. In evolutionary game theory, players adopt their
network [12]. More specifically, the network selection problem strategies through a trial-and-error process through which they
is decomposed into four fundamental steps [13]. learn over time that some strategies work better than
In the first step, all the parameters are identified which is others[20]. It provides a simple framework for describing
considered during the network selection process. After strategic interactions among different players (base stations
selecting the parameters, value of parameters are collected and ,mobile terminals etc.)[21]. As the traditional game is
the network conditions are monitored. Most of the time, these described by set of players, payoff of each player, set of
values are normalized. Then weights are assigned according to strategies and solution of each player. But evolutionary game
their preference of different parameters. Finally, the ranking of includes the concept of population. The population can be
the alternatives is done. The outcome of this process is the finite or infinite number of players and the game is played
selection of the most suitable or optimal network according to between group of individual rather than individual [22].
the user requirements. [14] Nash equilibrium is the well-known solution in
traditionally game theory. But in this game, evolutionary
III. GAME THEORY equilibrium provides an optimal solution and it gives the
Game theory is a mathematical tool which is used to stability. None of the individual wants to change its strategy at
analysis the strategical interaction among multiple decision the evolutionary equilibrium because at this point, payoff of
makers [15]. Initially it was used in economics for individual is equal to average payoff of the population [8].
understanding the concept of economic behavior. But now it is In evolutionary game, initially random networks are assigned
used in various fields such as communication, biology, to all the users present in the scenario. Then payoff associated
psychology for modelling the decision making situation where with each user assigned any particular network in the scenario
the outcomes depend upon the interacting strategies of two or is calculated [23]. The payoff achieved by user choosing
more agents [16]. network is given by
Game theory is used to the study of decision-making = − (1)
process where different players make their choices. The agents
who are playing the game are called players. In every Where denotes the capacity associated with network, is
movement of the game, the player takes an action. Their plan the pricing function of the network and is the total number
of actions to be taken by the player is called strategy such as of users choosing network in that area which is given by
request for bandwidth, offered price, new call etc. [6]. After = ∗ (2)
taking the decision of all players, each player will get either a Where N and denotes the total number of users and
positive or a negative return. This return of every player is proportion of users choosing network respectively.
called its payoff. Each player will choose strategies which can The function in equation (1) is called utility function and
maximize their payoff. This will lead to the concept of defined by
equilibrium in a game, which is called the solution of a game.
Thus, equilibrium is defined as the combination of best = [ ( )] (3)
strategies for each player. When every player cannot increase
its payoff by changing its strategy while there is no change in

978-1-5090-0082-1/16/$31.00 ©2016 IEEE


Where denotes the number of network attributes, denote ELSE (i.e., is the cost parameters)
the utility associated with each attribute, denotes the THEN = 1/
attributes vector and denotes the weight given to each END
attribute and it can be calculated by any method such as AHP 3) Column by column normalization :The column by column
method. normalization is done by this formula
After calculating individual payoff, the average payoff of the
user is calculated which is given by (5)
(4) =
= max ǀ1 ǀ1

This average payoff is used for selection of network. If the 4) Complete normalization: A matrix =[ ] ,
average payoff is equal to the individual payoff, then the 0 ( = 1,2 … ; = 1,2, . . . ) is finally given by
network which has the maximum payoff is selected by the user.
This selection technique is particularly used at the user side for (6)
= , = 1,2 … .
selection of network and in this no cooperation is involved as ∑
every time user select the network from the list of network
without consulting any other user present in the selection Matrix R is called the resource matrix. Each element of R
environment. So it is a user centric and non-cooperative game matrix represents the resource possessed by the candidate
theory approach. network.
5) Characteristics function: The coalition is taken as a
V. BANKRUPTCY GAME THEORY
subset of where the coalition of bankruptcy game is
denoted by( , ). Characteristics function is defined by
Bankruptcy game is the cooperative game theory which is
used to model distribution problems [6]. Bankruptcy game is ( : 2 → ) and it is given by
used in the case when the amount is insufficient to satisfy all
user condition. A coalition exists in this game as it is a
( ) = max 0, ∗
− , = 1,2, . . . (7)
cooperative game. For obtaining better or maximum payoff,
players cooperate with each other. After coalition forming, the , ∉

earned payoff is denoted by the characteristics function. After


that, the remaining payoff is perfectly divisible among a group Where
∗ (8)
of agents [24]. = max , = 1,2 … , = 1,2 …

Various network attributes such as available bandwidth, 6) Shapley value: After the calculation of characteristics
jitter, cost and loss has different characteristics. Some attributes function, the Shapley value is calculated. It used to find
should be small such as delay while some attribute should be the average payoff. Every players will get average payoff
larger such as available bandwidth. Different network have when he enter into the coalition [25].
different attributes value, so normalization is done to guarantee
that all the attributes are in the same order. (| | )!( | |)! (9)
∑ ⊂ , ∈ ( ( ) ( ))
!

A set of available candidate networks is given by = Where |S| is the count of elements in the set S.
, ….. where m is the total number of available
networks [9]. The network selection using bankruptcy game 7) Network ranking: Weight can be calculated by using AHP
theory is done by different steps. The first four steps are used method. For ranking the network using this game theory,
for normalization of different attributes matrix. After that Potential Contribution (PC) and Potential Contribution
characteristics function and Shapley value is calculated. And in Ratio (PCR) of networks are calculated as
the last step, potential contribution ratio is calculated for (10)
ranking of networks. The different steps are: = , = 1,2, …
1) Construct network attribute matrix
The matrix £ = [ ] ; 0 is constructed by
taking m network and n attribute corresponding to that (11)
= 100
network. ∑
2) Scale matrix according to attribute: The matrix is scaled
according to following rules Where wj is the weight assigned to different attributes.

IF is the benefit parameters In short, the network which has the highest value of PCR is
THEN = selected as the user network.

978-1-5090-0082-1/16/$31.00 ©2016 IEEE


This technique is usually used at the network operator side to Fig. 1 shows the different weights for different traffic class
provide resources to user. If one network is unable to provide which is calculated by AHP method. In case of streaming
resources to users then coalition provides the required traffic class, bandwidth is taken most important attributes so it
resources to the user. For distributing the coalition payoff, is given more weightage. In conversational traffic class,
Shapley value is used which provide a fair solution of packet delay is given more weightage and in case of
distribution as compared to other methods. background traffic class, cost is given more weightage.

VI. SIMULATION SCENARIO AND CONFIGURATIONS

A. Simulation Scenario

The six networks are considered for network selection i.e.


Wi-Fi, WiMAX, WLAN, LTE, UMTS_1, and UMTS_2. Three
different applications streaming, conversational, background
are taken to investigate the different game theory.
In case of streaming traffic class, Bandwidth is considered as
the extremely important attribute because it is used for
multimedia and video transfer. As video application required
more bandwidth so in this case most preference is given to
bandwidth. For conversational traffic class, packet delay is
taken as the extremely important attributes because it is used
for voice transfer. A voice application require continuous flow
of information so packet delay is taken as most important
factor. For background traffic class, cost is taken as the most
important attribute because it is used for downloading of a file. Fig. 1 Weight of Different Attributes for Different Traffic Class

B. Simulation Configuration The selection of network is done using evolutionary and


bankruptcy game theory. The score calculation and ranking of
There are different networks available i.e. Wi-Fi, WiMAX, the networks is done using evolutionary game theory in case
WLAN, LTE, UMTS_1, UMTS_2. The selection of network of streaming, conversational, background traffic class is given
is depending upon the applications. The weight calculation is
in Fig. 2. LTE network is selected for streaming traffic class as
done using AHP method [26].The different attributes taken
LTE and WiMAX both networks provides the maximum
are Available Bandwidth(AB) in Mbps, Propagation
Delay(PD) in millisecond, Packet Jitter(PJ) in millisecond, bandwidth. In case of conversational traffic class, UMTS_1 is
Cost per Byte(CB) in percent, Packet Loss(PL) per each 106 selected because UMTS network provides the less packet
packets. For simulation, maximum value of attributes are delay. And in case of background traffic class, WLAN
taken. The different networks with their attributes are given in network is selected as WLAN network provide least cost.
table I.
3.5
TABLE I: DIFFERENT NETWORK ATTRIBUTES
Streaming Traffic Class
3 Conversational Traffic Class
Network AB PD PJ CB PL
Background Traffic Class
(Mbps) (ms) (ms) (%) (per 106)
2.5

Wi-Fi 10 115 15 7 20-80 2


Network Scores

WiMAX 1-60 60-100 3-10 40 20-80 1.5

1
WLAN 1-11 100-150 10-20 10 20-80
0.5
LTE 1-60 60-100 3-10 50 20-80
0

UMTS_1 0.1-2 25-50 5-10 60 20-80


-0.5

UMTS_2 0.1-2 25-50 5-10 80 20-80 -1


WiFi WiMAX WLAN LTE UMTS_1 UMTS_2
Available Networks

Fig. 2 Network Scores using Evolutionary Game Theory

978-1-5090-0082-1/16/$31.00 ©2016 IEEE


The score calculation and ranking of the networks is done for different traffic class is calculated by the AHP method. In
using bankruptcy game theory in case streaming, case of streaming traffic class, bandwidth is given more
conversational, background traffic class is given in Fig. 3. preference. In case of conversational traffic class, delay is
WiMAX and LTE networks are selected for streaming traffic taken as most important factor and in case of background
class as it provides the maximum bandwidth. In case of traffic class, cost is taken as most important factor. Simulation
conversational Traffic class, UMTS network is selected as it results show that in case of streaming traffic class, WiMAX
provides the less packet delay. And In case of background and LTE is selected because both provide maximum
traffic class, WLAN network is selected because it provides bandwidth. In case of conversational traffic class, UMTS is
less cost. selected and in case of background traffic class, WLAN
network is selected.
0.35 In the evolutionary game theory, initial network selection
Streaming Traffic Class is based on the random behavior and payoff function of each
0.3 Conversational Traffic Class network is calculated. Due to random and dynamic nature of
Background Traffic Class
evolutionary game, new network is selected every time but in
0.25 case of bankruptcy game, fixed network is selected every time.
In the evolutionary game theory, the user behavior is selfish
Network S c ores

0.2 and always tries to gain maximum benefit. While in the


bankruptcy game, users behave in a manner to achieve mutual
0.15
advantage. In the evolutionary game theory, load balancing
can be achieved to the highest degree as compared to the
bankruptcy game theory. Load balancing is still a research
0.1
topic in the bankruptcy game theory.
0.05
REFERENCES

0
[1] D. E. Charilas and A. D. Panagopoulous, “Multiaccess
WiFi WiMAX WLAN LTE UMTS_1 UMTS_2 Radio Network Enviroments,” IEEE Veh. Technol. Mag.,
Available Networks vol. 5, no. 4, pp. 40–49, 2010.
[2] M. O’Droma and I. Ganchev, “Toward a ubiquitous
Fig. 3 Network Scores using Bankruptcy Game Theory consumer wireless world,” IEEE Wirel. Commun., vol. 14,
no. 1, pp. 52–63, 2007.
In the evolutionary game theory, initial network selection is [3] F. Bari and V. Leung, “Automated network selection in a
based on the random behavior and payoff function of each heterogeneous wireless network environment,” IEEE Netw.,
network is calculated. Due to random and dynamic nature of vol. 21, no. 1, pp. 34–40, 2007.
evolutionary game, new network is selected every time but in [4] L. Wang and D. Binet, “MADM-based network selection in
case of bankruptcy game, fixed network is selected every time. heterogeneous wireless networks: A simulation study,” in
1st International Conference on Wireless Communication,
In the evolutionary game theory, load balancing can be
Vehicular Technology, Information Theory and Aerospace
achieved to the highest degree as compared to the bankruptcy & Electronic Systems Technology, 2009.
game theory. [5] R. Verma and N. P. Singh, “GRA Based Network Selection
Most commonly, game theory is used for distributive in Heterogeneous Wireless Networks,” Wirel. Pers.
approaches with a lower communication overhead and self- Commun., vol. 72, no. 2, pp. 1437–1452, 2013.
configuration features. The aim of these two game theory [6] D. E. Charilas, O. I. Markaki, J. Psarras, and P.
approach in network selection is to improve the overall system Constantinou, “Application of Fuzzy AHP and ELECTRE to
performance (e.g. QoS guarantee, throughput maximization, Network Selection,” Mob. Light. Wirel. Syst., vol. 13, pp.
efficient resource utilization) but in MADM algorithm, only 63–73, 2009.
[7] R. Trestian, O. Ormond, and G. Muntean, “Game Theory-
most weight attributes is considered i.e. no impact of another
Based Network Selection : Solutions and Challenges,” IEEE
attributes. This simulation always provide Always Best Commun. Surv. Tutorials, vol. 14, no. 4, pp. 1212–1231,
Connected (ABC) because every time, the network is selected 2012.
according to application e.g. if higher bandwidth is required [8] H. Pervaiz and J. Bigham, “Game Theoretical Formulation
then WiMAX and LTE is always selected because it provide of Network Selection in Competing Wireless Networks: An
maximum bandwidth and so on. Analytic Hierarchy Process Model,” Third Int. Conf. Next
Gener. Mob. Appl. Serv. Technol., pp. 292–297, 2009.
VII. CONCLUSION AND FUTURE SCOPE [9] D. Niyato and E. Hossain, “Dynamics of Network Selection
in Heterogeneous Wireless Networks: An Evolutionary
In this paper, the network selection problem in Game Approach,” IEEE Trans. Veh. Technol., vol. 58, no. 4,
heterogeneous network is modelled as the evolutionary and pp. 2008–2017, 2009.
bankruptcy game. Decision factors are taken in case of [10] B. Liu, H. Tian, B. Wang, and B. Fan, “AHP and Game
streaming, conversational, background traffic class. The weight Theory based Approach for Network Selection in

978-1-5090-0082-1/16/$31.00 ©2016 IEEE


Heterogeneous Wireless Networks,” Consum. Commun.
Netw. Conf., pp. 973–978, 2014.
[11] M. El Helou, M. Ibrahim, S. Lahoud, K. Khawam, D.
Mezher, and B. Cousin, “A Network-Assisted Approach for
RAT Selection in Heterogeneous Cellular Networks,” IEEE
J. Sel. Areas Commun., vol. 33, no. 6, pp. 1055–1067, 2015.
[12] G. A. F. Mohamed Khalaf and H. Z. Badr, “A comprehensive
approach to vertical handoff in heterogeneous wireless
networks,” J. King Saud Univ. - Comput. Inf. Sci., vol. 25, no.
2, pp. 197–205, 2013.
[13] C.Makaya and S.Pierre, Emerging Wireless
Networks:Concepts, Techniques, and Applications. CRC
Press,New York, 2012.
[14] E. Aryafar, A. Keshavarz-Haddad, M. Wang, and M. Chiang,
“RAT selection games in HetNets,” Proc. - IEEE INFOCOM,
pp. 998–1006, 2013.
[15] B. Wang, Y. Wu, and K. J. R. Liu, “Game theory for
cognitive radio networks: An overview,” Comput. Networks,
vol. 54, no. 14, pp. 2537–2561, 2010.
[16] W. Liu, J. Liu, M. Cui, and M. He, “An introductory review
on quantum game theory,” Proc. - 4th Int. Conf. Genet. Evol.
Comput. ICGEC 2010, pp. 386–389, 2010.
[17] X. Liang and Y. Xiao, “Game Theory for Network Security,”
IEEE Commun. Surv. Tutorials, vol. 15, no. 1, pp. 472–486,
2013.
[18] G. Chalkiadakis, E. Elkind, and M. Wooldridge,
“Cooperative game theory: Basic concepts and computational
challenges,” IEEE Intell. Syst., vol. 27, no. 3, pp. 86–90,
2012.
[19] J. Sun and Q. H. Wu, “A Non-Cooperative Power Control
Game via New Pricing in Cognitive Radio,” 2009 Int. Conf.
Wirel. Commun. Signal Process. (Wcsp 2009), pp. 1295–
1299, 2009.
[20] B. Benmammar and F. Krief, “Game theory applications in
wireless networks: A survey,” 13th Int. Conf. Softw. Eng.
Parallel Distrib. Syst. (SEPADS ’14), Gdansk, Poland;
05/2014.
[21] M. Khan, H. Tembine, and A. Vasilakos, “Evolutionary
coalitional games: design and challenges in wireless
networks,” IEEE Wirel. Commun., vol. 19, no. 2, pp. 50–56,
2012.
[22] J. Chen, L. C. Jiao, J. Wu, and X. Wang, “Fast efficient
spectrum allocation and heterogeneous network selection
based on modified dynamic evolutionary game,” Phys.
Commun., vol. 13, pp. 53–60, 2014.
[23] N. Sui, D. Zhang, W. Zhong, L. Wu, and Z. Zhang,
“Evolutionary Game Theory Based Network Selection for
Constrained Heterogeneous Networks,” 2015 2nd Int. Conf.
Inf. Sci. Control Eng., pp. 738–742, 2015.
[24] M.J.Osborne and A. Rbinstein, A course in game theory. The
MIT Press,London, 1995.
[25] D. Niyato and E. Hossain, “A Cooperative Game Framework
for Bandwidth Allocation in 4G Heterogeneous Wireless
Networks,” vol. 9, no. 1, pp. 4357–4362, 2006.
[26] T. L. Saaty, “The Analytic Hierarchy and Analytic Network
Measurement Processes: Applications to Decisions under
Risk,” Eur. J. Pure Appl. Math., vol. 1, no. 1, pp. 122–196,
2008.

978-1-5090-0082-1/16/$31.00 ©2016 IEEE

You might also like