Professional Documents
Culture Documents
2, May 2010
Abstract — Exponential growth of user demands on a centric approach was introduced, leading to a vision of 4G and
single convergence platform has brought researchers to its key features and technological development. In addition,
explore various aspects/features of Fourth Generation (4G) 4G will be a convergence platform providing clear advantages
Mobile Communication System. Selection of application as in terms of bandwidth, coverage, power consumption and
per the user preference based on QoS (Quality of Service) is spectrum usage.
one salient feature of 4G. This user-centric system works on Further, it will be worth mentioning that the 4G systems are
the basis to provide best services of the network to the users. expected to co-exist and inter-work with existing 2G and 3G
In this paper, a selection algorithm has been proposed which mobile communication systems as well as satellite, wireless
provides a better way to implement user preferences as per LAN (WLAN), and IEEE 802.16e (also known as Wireless
the features of 4G. Rank based on distance function has been MAN (WMAN) or WiMAX [8, 9]) networks, all
computed for various available services/access technologies, interconnected through the service provider’s IP backbone
called networks. Weighted distance function is obtained based
networks and the Internet as shown in Figure 1. This inter-
on multiple QoS parameters as per user needs. The proposed
connected network has lead to the evolution of a new
algorithm shows better results compared to single parameter
paradigm for future communications, namely “Always Best
based system, under a heterogeneous network system.
Connected” (ABC), in comparison to typical 2G and 3G
Index Terms — 4G Mobile Network, Distance Function, QoS, concept of “Always Connected”. In new paradigm shift [10],
Rank. [11], [12] from “Always Connected” to “ABC”, the user will
always be connected through the best network, getting the
I. INTRODUCTION benefit of the best service and access technology.
In mobile communication systems, after 2G and 3G, the
Fourth Generation (4G) was originally expected as ultra-high
speed broadband wireless system [1]. Further, it was assumed
that the network will have a cellular structure which implies
that it will be built on existing architecture of the preceding
generations. A final agreement on what features characterize
4G mobile system, is yet to be reached. Sharing the 4G
objectives within research community is still open and lot of
features and applications have been suggested by the
researchers. Delivery of services to users in different location,
under different conditions with quality of service (QoS)
available in fixed environment, were some of the major issues
[2]. The Wireless World Research Forum (WWRF), aiming at
defining features of “beyond 3G” wireless communications
[3], the Japanese MIRAI project [4], the MobileVCE (Mobile
Virtual Centre of Excellence in personal and mobile
Figure 1: heterogeneous network (source: “4G: A user centric system”,
communications), and several European IST projects (such as
Mobile e-conference, 2004)
Daidalos, Magnet, Simplicity, Ambient Networks, etc.) are
only a few examples of international groups devoted to 4G
deployment. The European Commission (EC) foresees that
However, to realize the above benefits, each terminal device,
from the service point of view, 4G will be mainly a
used in 4G networks, need to support multi-mode, multi-
personalized services network [5]. As any
access and reconfiguration capabilities. That is, each terminal
service/application/system is to be used by the end user, the
can access more than one type of network and possibly access
system has to be designed taking into account the needs of the multiple networks simultaneously for different applications
user. A first attempt was done in [6] and [7] where user- (e.g., one for voice and another for receiving streaming
1
Amit Sehgal is with G. L. Bajaj Institute of Technology and Management. media).
(e-mail: amitsehgal09@rediffmail.com) In such an environment, a terminal must be able to discover
2
Rajeev Agrawal is with G. L. Bajaj Institute of Technology and Management all networks that are available for use. A more challenging
(e-mail: rajkecd@gmail.com) issue is to discover and select the best network whenever a
Contributed Paper
Manuscript received December 2, 2009
Current version published 06 29 2010;
Electronic version published 07 06 2010. 0098 3063/10/$20.00 © 2010 IEEE
Authorized licensed use limited to: Anna University Coimbatore. Downloaded on August 10,2010 at 07:20:58 UTC from IEEE Xplore. Restrictions apply.
A. Sehgal and R. Agrawal: QoS Based Network Selection Scheme for 4G Systems 561
Authorized licensed use limited to: Anna University Coimbatore. Downloaded on August 10,2010 at 07:20:58 UTC from IEEE Xplore. Restrictions apply.
562 IEEE Transactions on Consumer Electronics, Vol. 56, No. 2, May 2010
same user ( Ts ,u ,( k −1) ) and effective duration of the previous parameters in set P as per the requirements of user u for the
session k. This ordered list is represented as Xp where p is a
session in similar situation ( Ts ,u ,( k −1) ) using equation: particular parameter in set P. Position of ith network in the
ordered set is given by its rank in the set represented as
Rank (i,p,u,k). For example, in a scenario of three networks
Ts ,u ,k = β (Ts ,u ,( k −1) ) + (1 − β )(Ts ,u ,( k −1) ) (3) (Net1, Net2 and Net3) and parameter set having two entities,
m=4, n=2 and P= {bandwidth, security}. For Net1, P1=
The distance function attribute is now calculated as: {free, low}. For Net2, P2= {medium, high}. For Net3, P3=
{high, medium}. Now, the ordered list Xbandwidth= {Net1,
Ts ,u ,k Net2, Net3} and Xsecurity= {Net1, Net3, Net2}. As shown in
Ps ,u ,k ,d = . The maximum admissible value of this
Tsd this example, an ith network may have different rank in
different ordered sets prepared corresponding to different
parameter, after normalization, is 1 which represents the case parameters. It, therefore, becomes necessary to have some
when Ts ,u , k > Tsd . optimal multi-parameter selection criterion. As mentioned
E. Other network related parameters: Apart from the above earlier in this section, we use Spearman footrule distance
parameters, the network selection depends on many other function which defines a distance between two ranks of a
network related parameters such as, security level offered and network as the sum, over all m networks in set N, of
area covered. A user interested in accessing a network to differences between the ranks.
m
2
perform e-transaction would be more concerned about the
level of security provided by the available networks.
DSF (τ x ,τ y ) = ∑
m(m − 1) j =1
{τ x Pj − τ y Pj } (4)
However, other user, moving frequently might give preference
to the network which offers uniform signal strength and/or This function generates an optimally ranked list of available
results in less number of handovers. networks from n number of lists obtained according to the n
different parameters, each list having m available networks.
III. PROPOSED NETWORK SELECTION The rank of a network in this optimal list is given as:
ALGORITHM n
τ *A = arg min ∑ DSF (τ P ,τ x ) (5)
Distance function: It is a common observation that, while τP x=1
accessing any network or service, a user demands low
billing rate, high bandwidth and data rate, adequate security, where τP is the value of the parameter in set P for which
low call drop etc. With the increase in number of quality ordered list X is obtained. τx is rank of the network in the
parameters required by the user, the complexity of the ordered list x.
network selection is likely to increase. Multi-parameter Weighted distance Function: Another level of customization
selection technique is required to meet user needs in terms required in 4G is the preference of the user in terms of giving
of automatic network selection during handover. Attempts priority to few parameters. For example a user performing
have been made in recent studies using static network electronic transaction will be more concerned about security
interface priority [14], received signal strength [15], cost rather than bandwidth or cost. This level of customization is
function [16]. Various algorithms have also been proposed implemented by weighted distance function. In this function
using different criterion for optimum network selection, extra weight is added to the parameters as per the level of user
such as, best network selection [17], score function [18], interest. Another vector L = {l1, l2, l3, ln,} where lz is the user
vertical handoff decision in [19] and terminal controlled interest level in the corresponding parameter p (z) in vector P.
n
mobility management [20]. In this paper, optimal network
selection is suggested using distance functions. Several The vector L is scaled such that ∑l
z =1
z = 1 . The weighted
distance functions have also been proposed earlier. We
focus on the Spearman footrule. ranking of the available networks can, now be written as:
To use the distance function, all the parameter values are Rank (i, u , p, l , k ) = τ *A Where the ordered list τ *A is
normalized so that they take on values in the range of zero defined as:
and one. These values are assigned such that the higher the
value, better it is for the user. For example a network with The equation pj (1) lj ≥ pj (2) lj holds if τ *A (P1) < τ *A (P2)
billing parameter of 0.9 is more economical than the one where pj represents jth parameter in set P.
having a value of 0.5. Proposed Algorithm: As this is a multi-parameter ranking
Let N= {net (1), net (2), net (3), net (m)} be the set of problem, a modified version of Borda’s method of rank
various network choices available to the user. The n aggregation [21] has been used as positional ranking
numbers of user customization parameters are represented technique. Ordered lists of available networks obtained earlier
by set: P= {p (1), p (2), p (3) and p (n)}. Thus jth network in in this section have been used to assign a score to each
the set N is characterized by a set of parameter values network corresponding to its position in the ordered lists over
written as Pj= {pj (1), pj (2), pj (3), pj (n)}. The networks in each parameter in set P. Each network is, thereby, having a set
set N are arranged in an ordered sequence with respect to the or n scores, each corresponding to a parameter.
Authorized licensed use limited to: Anna University Coimbatore. Downloaded on August 10,2010 at 07:20:58 UTC from IEEE Xplore. Restrictions apply.
A. Sehgal and R. Agrawal: QoS Based Network Selection Scheme for 4G Systems 563
Input: Rank lists α1, α2, α3, αn of m networks from the set N
corresponding to each of the n parameters.
Begin
Output: an optimally ordered list N ().
Begin initialize
1: Sort respective lists α1, α2, α3, αn Input: Rank lists α1, α2, α3, αn of
m networks from the set N
2: Assign a position score for each network in each of the n corresponding to each of the n
lists. parameters.
3: For j = 1 to m
4: For x = 1 to n
5: Sx (j) = the number of networks whose score
of parameter x is ranked below j in αx Sort
6: end for
n Respective lists α1,
N ( j) = ∑ Sx ( j)
α2, α3, αn
7:
x =1
8: end for
9: sort N (j) in decreasing order where j = 1 to m.
Assign a position score
for each network in each
IV. TESTING OF PROPOSED ALGORITHM of the n lists.
TEST AREA
R1
Sort N (j) in
Net1 decreasing order
U1 where j = 1 to m.
U2
Net4
Net3 Output: an optimally
ordered list N ().
R2 R3
Net2
Net1 U3
End
Figure 2: The Test Area with inter-working networks and three test
Figure 3: Flowchart of proposed network selection scheme.
users
Authorized licensed use limited to: Anna University Coimbatore. Downloaded on August 10,2010 at 07:20:58 UTC from IEEE Xplore. Restrictions apply.
564 IEEE Transactions on Consumer Electronics, Vol. 56, No. 2, May 2010
TABLE 1
FEATURES OF THE TEST NETWORKS
Net1 Net2 Net3 Net4
Cost of Fixed Fixed Data Based Free
Net4 4
Service Access
Network selected
Net3 3 cost
BU factor Medium Low High Low
Call Drop Low Medium Medium High
Prob. Net2 2 bandwidth
Security Medium High Low Low
Level Net1 1 distance function
TABLE 2 0
USER PREFERENCES WITH LEVEL OF INTEREST (IN SCALE OF 0-1]) R3 R2
User 1 User 2 User 3
Regions
Application of Live Video e-Transaction Multimedia
Interest Messaging Figure 6 Comparison of network selection profile for user 3
Cost of Service Medium(0.6) High (0.2) Low (0.9)
Bandwidth High (1) Medium(0.5) Medium(0.4)
Results
Utilization
Call Drop Medium(0.5) Low (1) Medium(0.7) Figure 4 shows that when the cost of accessing service is the only
Probability criterion i.e. user is looking for the low cost network, the user 1
Security Level Low (0.3) High (1) Low(0.3) gets connected to net4 in R1 and remains connected to net4 in R2.
Based on the best bandwidth utilization, net3 in R1 and net1 in R2
Various available access technologies in different regions, as given
are the selected networks to access the desired service. Since user1
in table 1, are arranged in ordered sequence according to various
is interested in accessing a live video he/she requires high
user preference parameters mentioned in table 2. Using distance
bandwidth but, he/she is willing to pay moderately. Using proposed
function (see section III) the most suitable network is selected using
distance function based network selection algorithm, user1 gets
proposed algorithm presented in section III. As the user moves
connected to net1 in R1. This is because the user1 preferences of
from one region to another, the networks available in the new
bandwidth utilization and call drop probability can not be
region are again searched for optimal solution as per the user needs
supported in the cheapest network net4. For reasons of bandwidth
and preferences and, if required, the user is connected to a most
requirements, the user1 remains connected to net1 in R2 also.
suitable network. Three network selection schemes have been
User2 wants to execute some e-transactions, so security is
compared namely, cost based, bandwidth based and proposed
of highest concern. As shown in figure 5, in cost based network
distance function based. Figures 4, 5 and 6 show the comparison of
selection, net4, being a free service, in R1 and net2 in R3 is
these schemes for three mobile test users: U1, U2 and U3 as per the
selected assuming that network with fixed cost of access is more
test environment described earlier in this section.
economical than data rate based costing scheme. When bandwidth
Net4 4 utilization is the only criterion for network selection user2 is
connected to net3 in R1 and remain connected to same network in
Network selected
cost
Net3 3 R3 also since net3 has highest bandwidth utilization amongst all the
competing networks. In distance function based network selection
Net2 2 bandwidth scheme, net1 in R1 and net2 in R3 is the most suitable network
since the user is equally concerned about the low call drop
Net1 1 distance function probability.
0 Multimedia messaging chosen as the application of
R1 R2 interest for user3 is a representative of data transfer for
entertainment where the user is more interested in cost economical
Regions
network than bandwidth utilization or security. Low call drop
Figure 4: Comparison of networks selection profile for user 1 probability is also desired with slightly lower level of interest since
a call drop may force the user to retransfer the entire multimedia
file. Figure 6 shows the network selection profile of user3 for
Net4 4 different selection criterion. Cost based selection results in
Network selected
Authorized licensed use limited to: Anna University Coimbatore. Downloaded on August 10,2010 at 07:20:58 UTC from IEEE Xplore. Restrictions apply.
A. Sehgal and R. Agrawal: QoS Based Network Selection Scheme for 4G Systems 565
and reconfigurable devices to support inter-working of [15] M. Buddhikot. et al., Integration of 802.11 and Third Generation
Wireless Data Networks, IEEE INFOCOM 2003.
heterogeneous networks. The algorithm selects appropriate
[16] H. Wang, R.H. Katz, J. Giese, Policy-Enabled Handoffs across
network during handoff based on user preferences and Heterogeneous Wireless Networks, Proc. Of ACM WMCSA, 1999.
interests. The user can opt for multiple QoS parameters like [17] L. Suciu, J.M. Bonnin, A survey of multi-criteria network selection
bandwidth, cost of service, security level, call drop probability algorithms, Global Mobile Congress, 2004 Shanghais.
etc. to select appropriate networks. [18] L. Chen, T. Sun, B. Chen, V. Rajendran, M. Gerla, A smart decision
Model for Vertical Handoff, Proceedings of 4th ANWIRE International
The proposed algorithm uses a distance function to generate Workshop on Wireless Internet and Re-configurability (ANWIRE 2004),
an ordered list of various available access networks in a Greece, 2004.
particular region according to the multiple user preferences [19] Lee SK, Sriram K, Kim K, Lee JH, Kim YH, Golmie N (2007) Vertical
and level of interest. In case of handoff due to the movement handoff decision algorithms for providing optimized performance in
heterogeneous wireless networks. In: IEEE GLOBECOM 2007.
of the user from one region to another region, the process is Washington DC, November.
repeated to find again an appropriate network from all the [20] Nguyen-Vuong Q-T, Agoulmine N, Ghamri-Doudane Y (2007)
available networks in the new region. The results clearly show Terminal-controlled mobility management in heterogeneous wireless
that the proposed algorithm always best connect the user, as networks. In: IEEE_M_COM, April.
[21] C. Dwork, R. Kumar, M. Naor and D. Sivakumar, “Rank Aggregation
per his preferences of QoS parameters in a 4G System. Methods for the Web”, In proceeding sof the 10th International World
wide Web Conference, pages 613 – 622, 2001.
REFERENCES [22] D. Charilas, O. Markaki, D. Nikitopoulos, M. Theologou, “Packet
Switched Network Selection with Highest QoS in 4G Networks”,
[1] E. Bohlin, S. Lindmark, J. Bjrkdahl, A. Weber, B. Wingert, P. Ballon, Computer Networks: The International Journal of Computer and
”The Future of Mobile Communications in the EU: Assessing the Telecommunications Networking, vol 52, issue 1, pages 248 – 258,
Potential of 4G”, ESTO Publications, February, 2004. 2008.
[2] S. Frattasi, F. Fitzek,”Preliminary Recommendation for Overall [23] J. Govil, “An Empirical Feasibility Study of 4G’s Key Technology”,
Scenarios and Requirements”, JADE Deliverable 1.1.1, June, 2004. Electro Information Technologies, pages 267 – 270, 2008.
[3] WWRF, http:// wireless-world-research.org/, Book of vision 2001www
[4] M. Inoue, K. Mahmud, M. Hasegawa, H. Muratami, G. Wu, MIRAI: A
Solution to Seamless Integration of Heterogeneous Wireless Networks,
ICT +2002
[5] J. M. Pereira, ”Fourth Generation: Now, it is Personal”, Proceedings of BIOGRAPHIES
the 11th IEEE International Symposium on Personal, Indoor and Mobile
Radio Communications (PIMRC), London, UK, September 2000. Amit Sehgal graduate and Post Graduate both in
[6] S. Frattasi, H. Fathi, F. Fitzek, K. Chung, R. Prasad,”4G: The User- Electronics and Communication Engineering. He is
Centric System”, Mobile e-Conference (Me), Electronic Conference, pursuing Ph.D. in Wireless Neworks. His current
August, 2004. areas of research are performance evaluation of
[7] S. Frattasi, H. Fathi, F. Fitzek, M. Katz, R. Prasad,”A Pragmatic wireless networks and 4G systems. He has 8 papers to
Methodology to Design 4G: From the User to the Techn01ology”, in his credit in various national level conferences and
Proceedings of the 5th International Conference on Networking (ICN), international journal.
IEEE, Reunion Island, France, April 17-21, 2005.
[8] IEEE 802 standards, http:// www.ieee802.org
[9] WiMAX, http://www.wimaxforum.org.
[10] V. Gazis, N. Alonistioti, L. Merakos, Toward a Generic “Always Best
Connected” Capability in Integrated WLAN/UMTS Cellular Mobile Rajeev Agrawal Graduated in Electronics
Networks (and Beyond), IEEE Wireless Communications, June 2005 Engineering Post graduation in Systems Sciences.
[11] E. Gustafsson, A. Jonsson, Always Best Connected, IEEE Wireless He has done his Ph.D. in the area of Wireless
Communications, February 2003. communication channels His current areas of
[12] O’ Droma, Wireless, Mobile and Always Best Connected, Proc. of the 1st research are performance evaluation of wireless
International ANWIRE Workshop, 2003. networks and Ultrasound Medical Imaging. He has
[13] Bader, F.; Pinart, C.; Christophi, C.; Tsiakkouri, E.; Ganchev, I.; more than 16 publications in reputed International
Friderikos, V.; Bohoris, C.; Correia, L.; Ferreira, L., “User-centric Journals & conferences along with a number of
analysis of perceived QoS in 4G IP mobile/wireless networks Personal, national conferences papers to his credit. He is also a
Indoor and Mobile Radio Communications”, 2003. PIMRC 2003. 14th member of the IEEE, EU-IndiaGrid community, ISTE and world academy of
IEEE Proceedings on Volume 3, Issue , 7-10 Sept. 2003 Page(s): 2047 - science, engineering & technology (WASET)
2053 vol.3 Digital Object Identifier 10.1109/PIMRC.2003.1259074
[14] D. Anderson, E. Otchet, F. Christensen, Web Sphere Everyplace
Connection Manager: Increasing mobile security, reducing wireless
costs, IBM Pervasive Computing, July 2004.
Authorized licensed use limited to: Anna University Coimbatore. Downloaded on August 10,2010 at 07:20:58 UTC from IEEE Xplore. Restrictions apply.