Clustering over TV White Space in Dense Wireless Areas: Dynamic Hotspot Selection and Resource Allocation

Abstract

This paper, presents a novel technique to solve the problem of lack spectrum in dense areas using a clustering approach. The proposed clustering solution enhances the scalability and coverage of highly dense cellular networks.

Full text

Clustering over TV White Space in Dense Wireless Areas: Dynamic Hotspot
Selection and Resource Allocation
Ula’a Alhaddad 1, Ghadah Aldabbagh1 and Nikos Dimitriou2
1Department Computer Science, King Abdulaziz University, Jeddah, Saudi Arabia
2 Institute of Informatics and Telecommunications NCSR Demokritos, Athens, Greece
Abstract - The proliferation of smart-
phones has increased the demand for
ubiquitous access to wireless data in our
daily life. This problem of increasing
demand for mobile broadband with the lack
of spectrum, especially in densely populated
areas is a serious issue. This paper,
presents a novel technique to solve the above
mentioned problem using a clustering
approach. The proposed clustering solution
enhances the scalability and coverage of
highly dense cellular networks. The
proposed approach organizes the nodes into
groups; each group (cluster) has one master
node that communicates directly to the Base
Station (BS) while the other nodes (slaves)
communicate through their respective
master nodes. The proposed configuration
requires low power for transmitting and
receiving data. In addition, it enables
frequency reuse without causing
interference, thus it increases the scalability
of the network and reduces number of
blocked and outage users.
Keywords: Clustering, TV White-Space,
Wireless Networks, Cellular Networks, LTE
4G
1 Introduction
The rapidly increasing number of smart
devices and mobile users has established
widely the concepts of ubiquitous wireless
access to the internet. However, in crowded
events such as Hajj (annual pilgrimage
where, every year, millions of Muslims
from around the world make the journey to
Makkah, Saudi Arabia), a large number of
mobile devices contend for the limited
spectrum. Utilizing the unused television
bands "white space" is one of the proposed
solutions to solve this problem. The TV
white-space (TVWS) are present on the
50−700 MHz resulting in coverage ranges
of 10's of km due to the better propagation
characteristics, compared to the Wi-Fi ISM
bands . Long-Term Evolution (LTE) allows
operators to use new and wider spectrum
and complements 3G networks with higher
data rates and lower latency. Using a mix of
big cell and small cells enables flexible and
low-cost deployments and provides a
uniform broadband experience to users
anywhere in the network [1].
Within a wireless cellular infrastructure,
regions are divided into cells and each cell
has its own Base Station (BS) so as to
control it. Frequencies are divided into sets
and are reused in uniform models so that
nearby cells use different frequencies in
order to avoid co- channel interference.
One of the proposed solutions when one cell
becomes very densely populated is to divide
the cell into smaller cells such as Femtocell
and micro-cells with lower transmission
powers. This configuration will allow the
network capacity to be expanded and the

cell coverage to get enlarged by reducing
the path-loss while choosing the hotspots
closer to the their slaves than the Base
Station [2]. A study in [3] a comparison of
existing schemes that consider Femtocell in
case of dense area networks and the related
challenges in terms of interference,
offloading and mobility management is
presented. Also, this study points out the
main protocols and access type that should
be adopted in this case.
In [4], the authors present a method for
grouping the mobile nodes into clusters
using TV white space (TVWS) bands for
some users to relay their information
between the BS and other devices,
considering mobile users’ requirements by
studying the corresponding handover
scenarios and signaling schemes.
In [5], a hotspot-slave configuration was
proposed. A hotspot is defined as a mobile
device that connects to the BS directly and
aggregates its slaves’ (if any) uplink and
downlink data with its own data when
connecting with the BS. A slave is a mobile
device that communicates with the BS
through a hotspot. This study provides an
algorithm that attempts to organize all
active nodes into small clusters that allows
more efficient spectrum and power
resources distribution among users in dense
areas.
The concept of clustering provides several
advantages compared to the fixed small cell
deployment. Clustering can be used on an
ad-hoc fashion to enhance the capacity of a
macrocell and adapt to increasing traffic
requirements and variations that may be
observed on specific times within the day.
Since the clusters are formed by mobile
terminals that opportunistically play the role
of cluster-heads, there is no need for an
operator to deploy specific fixed base station
infrastructure. On the other hand, the
formation of clusters requires specific control
signaling and direct communication between
the involved terminals, which can be enabled
by proper modifications in the considered
Radio Access Network (RAN) standards.
This research paper aims to demonstrate the
benefits in terms of scalability and capacity
in cellular-based TVWS using clustering
approaches that clusters the nodes (mobile
users) into small groups. This architecture is
depicted in Fig 1. The research will also
consider suitable power control and resource
allocation schemes in order to provide a
sufficient spectrum usage. It will show that
the concept of clustering enables the reuse of
frequency channels among the clusters
without the need of additional infrastructure,
with less deployment costs, and lower
generated co-channel interference and also
higher energy efficiency.
The paper is organized as follows: In Section
II, related work is presented. The system
model and proposed algorithm are described
in Section III. Then, several use case
scenarios are assessed via simulation in
Section IV. Finally, the paper conclusions
are provided in Section V.
Figure 1: TVWS Clusters formed in a macro cell.
2 Related Work
Nowadays the growing demand for
mobile broadband and the limited portions
of the cellular spectrum particularly in

crowded venues, dictate the development of
more efficient approaches for distributing
resources. As a result, many solutions in the
literature were found to consider this
problem and suggest possible solutions.
Most of these approaches are based on
reusing the available frequencies and using
Heterogeneous Networks (HetNets). The
study in [6] proposed a new technique for
HetNet enhancement called the QoS-Aware
Tethering in a Heterogeneous Wireless
Network using LTE and TV White Spaces
to enhance QoS for Constant Bit Rate
(CBR) and Best Effort (BE) users in dense
wireless network. It aims to increase system
bandwidth and resolve the crowding
problems in a HetNet by dynamically
constructing new clusters. The authors in
[7] proposed Distributed Dynamic Load
Balancing (DDLB) using TVWS and LTE
technique. This technique enhanced
cellular-based devices by simply switching
the frequency of operation to activate both
TVWS and LTE bands.
2.1 Cellular and Heterogeneous
Networks
The concept of cellular systems is based on
re-using the available bandwidth as much as
possible within a coverage area. The
coverage area in a cellular system is divided
into cells and the frequency band allocated
to a cellular mobile radio system is
distributed over this group of cells. This
distribution is repeated within all the area
belonging to the operator domain. Thus,
frequencies used in a cell are reused into
several cells. The distance between the cells
using the same frequency must be sufficient
to avoid interference. The frequency reuse
can increase considerably the scalability of
the network and hence the number of served
users. However, in a densely populated cell,
the assigned frequency band will not be
sufficient to handle all the users. Therefore,
cellular systems have introduced smaller
cells, such as micro-cells and femto-cells to
further increase reuse and scalability [8].
Heterogeneous Networks (HetNets) involve
a combination of large cells (macro) and
small cells (Pico/Femto) with different radio
technologies (3G/LTE, LTE advanced). In
other words a HetNet introduces new small
cells inside the macro cell by adding new
Pico or Femto cells to provide the best
coverage and scalability possible in cellular
system [9]. In a comparison between
existing schemes of Femtocell that consider
the related challenges of offloading ,
interference and mobility management is
presented in addition to the main protocols
that should be adopted in case of dense
areas.
2.2 TV White- Space Band
Some solutions to this problem include
adding new infrastructure such as a study in
[10] suggested adding new infrastructure
such as Wi-Fi APs and Femtocells to enable
services to users. However, the lack of Wi-
Fi bands and the possibly uncontrollable
interference that may be generated in the
ISM band may not completely satisfy the
user demands and may not thus fulfill the
additional capacity objectives. This
approach restricts the scalability of wireless
networks even when there exist
opportunities to better use the spectrum.
In 2004 the Federal Communication
Commission (FCC) in the United States
showed that the unused T.V channels in
both very high frequencies (VHF) and ultra-
high frequency (UHF) bands could be
utilized for fixed broadband access [11].
Later in 2008, a study on the introduction of
open-spectrum policies has been published
showing that operating WRAN systems in
TV bands could bring hundreds times

higher spectral efficiency, and more than
two times higher coverage efficiency [12].
From this point, IEEE has developed the
first cognitive radio wireless regional area
network (WRAN) called the IEEE 802.22.
This standard operates on TV white spaces
and aims at providing broadband access in
rural areas by utilizing the unused TV
channels (white-space) [13]. This unused
broadcast television spectrum as indicated
as TV white-space (TVWS) are existing on
the 50−700 MHz bands and have better
propagation characteristics compared with
WiFi ISM bands. Combine a TVWS overlay
to the LTE network has been investigated
in[14].
In [15] the design and implementation of the
first Wi-Fi system constructed on top of
UHF white spaces (WhiteFi) is presented. In
this system, choosing the best transmission
channel is done through an adaptive
spectrum assignment algorithm which deals
with many challenges including spatial and
temporal variations of available spectrum
and fragmentation.
The algorithm periodically examines the
assignment of channels based on white
space availability at the access point (AP)
and clients, as well as for handling
disconnections due to interference or station
mobility. This technique improves wireless
network efficiency and provides coverage to
several mobile users; hence it provides
scalability via opportunistic wireless
network architectures.
2.3 Clustering Techniques
Clustering is a wide methodology defined as
the "division of data into groups of similar
objects. Each group is called a cluster and
consists of objects that are similar between
themselves and different to objects of other
groups" [20].
Mobile relaying mainly depends on the
clustering approach. Since large-scale
mobile ad-hoc networks (MANET) cannot
guarantee performance within a flat node
structure, many hierarchical clustering
algorithms have been proposed to solve the
related scalability issues. In a MANET that
employs a clustering scheme, the mobile
nodes are divided into different virtual
groups, and they are allocated to a cluster
that is ‘adjacent’ to them according to
specific criteria (e.g. geolocation) and also
according to the role each node is assumed
to possess, such as cluster-head, cluster-
gateway, or cluster member [21]. It has also
been shown that a clustered architecture
enables and facilitates frequency reuse
which would increase the system’s
scalability. Therefore, available techniques
in MANETs can be adopted to increase
scalability in LTE and TVWS networks as
well.
The proposed clustering schemes in
MANETs can be classified based on their
objectives, for instance, the Ryu's
algorithm[22] proposed a clustering scheme
for energy conservation. This algorithm is
somehow related to this research project
since it is focused on minimizing the
transmission power in each cluster in order
to achieve a better performance with longer
battery lifetime. Furthermore, each node is
either considered as a master (cluster-head)
or as a slave. A slave node can connect to
only one master node and a direct link
between slaves is not allowed. In addition,
an on demand weighted clustering algorithm
(WCA) is found in [18, 19]. Combined
metrics are taken into account including
node degree, distance, speed, and the battery
energy of cluster-heads. It can flexibly
adjust the weighting factors for each metric
depending on how important this metric is
in a particular situation. One of its key
benefits is that it aims at electing the most
suitable cluster-head in a local area.

Table 1: Technique’s Attributes
We can find that many studies have
suggested frequencies to be reused among
small cell configurations and HetNets to
enhance network scalability. These cells can
be independent small cells or clusters based
on relaying slave node signals to the BS.
Those that use relays could have either fixed
or mobile relay stations due to the flexibility
and low cost of mobile relaying. This
research will be based on mobile relaying.
Thus, D2D communication in LTE between
the base station (BS) and cluster-heads, as
well as, communication in TVWS between
cluster- heads and their slaves must be taken
into account. Table 1 shows different
technique’s attributes involved in this study.
3 PROPOSED ALGORITHM
The proposed clustering algorithm consists
of three main steps as summarize below:
Step-1: Clustering Nodes into Hotspots and
Slaves
K-means clustering is a technique of cluster
analysis that aims to partition n observations
into k clusters in which each observation
belongs to the cluster with the nearest mean.
The proposed algorithm is based on K-
means algorithm and considers as inputs the
Euclidian distances between the nodes.
Moreover, the cluster head or hotspot choice
in each of the formed clusters depends on
the node’s distance from the base station. In
each round of the K-means clustering
algorithm, k nodes are chosen as cluster
heads according t the abovementioned logic.
Figure 2 below illustrates K-means
algorithm by flow chart and Figure 3
explains the algorithm.
Figure 2: Flowchart of K-mean algorithm
Research no.
[16]
[5]
[9]
[10]
[15]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
Technique’s Attributes
Use HetNet
√
√
√
√
√
√
Use Micro cell- Femtocell
√
√
√
√
√
√
√
√
√
√
Infrastructure AP
√
√
√
Wihte Fi
√
Mobile relay sys.
√
√
√
√
√
√
D2D
√
√
√
√
√
√
√
√
√
Fixed Relay / Mobile Relay
√
√
√
Clustering schemes
√
√
√
√
√
√
√
Initial value to cluster the nodes into “K” clusters
Compute each nodes Euclidean distance to the centroid
nearest
Recompute centroid in each cluster
If centroids change
End Clustering
 
, 1,...,k
j
cj
 
, 1,...,
i
x i n
 
 
21
1min ,d x i
n





YES
NO

/10
PLTotal = 10* log 10 (PL*10 )
:
:
:4
:
S
where
PL PathLoss
ple pathLossExponant
S Shadowing

Algorithm 1: K-means Clustering
Input:
= {1,2,…..,} (set of nodes to be clustered)
 (number of clusters)
Output:
= {1,2,…..,} (set of cluster centroids)
Initialization
for each   
←   (random selection)

Repeat
for each    
 ←(,){1…};
(based on Euclidean distance
For i = 1 to n, )
UpdateCluster( )

 Convergence
Fig 3: K-means Clustering Algorithm
By following this selection method, i.e. by
selecting the node that is closest, on average,
to both the center of the cluster and the BS
provides better communication links for
both layer-1 LTE and layer-2 TVWS
connections[25].
Fig 4: Flow chart of proposed clustering Algorithm
This proposed algorithm as shown above in
figure 4 is a dynamic hotspots selection and
resource allocation. It starts by initializing
the number of clusters to two and then by
assigning user coordinates using Euclidean
distance between the node and BS such that
they are not greater than the radius of the
macro cell R:
(2)
Where:
is path loss , : Euclidean distance
between the nodes and is path loss
exponent.
(3)
Then, the users with the least path loss PL
(i.e. those closer to the macro BS) are
assumed to be connected to the macro base
station. Figure 6 shows the association of
users to the macro BS.
1
n




ED
ple
 
min 2,
ii
j d x c




Assign users randomly
Start
K-mean clustering
Grouping based minimum distance
List
Initial no. of clusters
Clusters=2
Calculate the Centroid (HS)
Rate requirement met ?
End
Clustes= clusters + 1
NY
Resource Allocation among users
ple
PL ED
PL

h
C
R
f
sh
x
 
 
,1
1
min
_:
log 1 ,
1,
0, ,
0,1 ,
h
Ee
P X hh
h H e
ee
Eehh hh
hh h s
e
e s C
hh
e
hh
hH
e
hh
e
hh
imize p
subject to
pg
x b R R h H
x
xe
p e h H
x e h H




    



   
   



Fig 6: User Association with macro BS User Association
with macro BS
Step-2: Resource Allocation Among users
This step of the algorithm which electing
hotspots according to k-mean clustering and
power victor includes sub-steps.
(i) Resource Allocation among Slaves:
Each hotspot allocates resource block
groups RBGs to its slaves based on an
OFDMA-like approach with rate and power
constraints [18]. In this approach, each slave
can use only one RBG.
(ii) Allocating RBGs and Transmission
Powers: After determining each user
transmission power, and having knowledge
of the rate requirement of each slave, each
hotspot solves the following optimization
problem:
(4)
(5)
(6)
(7)
(8)
Where h corresponds to the transmitter (
hotspot), f corresponds to the operating
frequency band (set of channels) for
example, LTE or TVWS band, g is the
average channel gain to noise power ratio,
is the set of slaves associated with
hotspot h, to the total transmission
power and fixed rate to the
destination node. In this approach, each
RBG can be used by only one slave. The
purpose of this algorithm is to minimize the
total transmission power over all the
available RBGs with minimum rate
requirements Constraints (5) determine the
rate constraint. Constraint (7) forces the
power to be positive and constraint (8)
enforces the channel selection variables
to get on binary values.
(iii) Resource Allocation among Hotspots: The
BS allocates its resources along with the selected
hotspots h  H.
(9)
(10)
(11)
(12)
(13)
, where H is set of Hotspots and E
corresponds to the different spectrum band
(set of channels). Constraint (10) indicates
that the rate hotspot h obtains has to be
greater than or equal the sum of its slaves’
rates or its own rate requirement
 
 
,1
1
min
_:
log 1 ,
1,
0, ,
0,1 ,
h
h
Ff
P X sh
s c f
ff
Ffsh sh
sh s h
f
fsh
f
sh
sc
f
sh h
f
sh h
imize p
subject to
pg
x b R s C
x
xf
p f s C
x f s C




   



   
   



f
sh
p

Table 1: Number of iterations, clusters and restricted users
Step-3: BS Rate and Power Constraint
Check
In this final step the feasibility of the
resource allocation performed in previous
step is checked to make sure that all Nodes
have been successfully granted their
requested resources to satisfy their cluster
rate requirements and whether we have
reached the maximum number of iterations.
If it is not feasible, this means that there are
not enough resources available and this step
is yet not completed successfully. The BS
re-clusters the nodes by incrementing the
clusters number and calculating the
restricted users.
4 SIMULATION RESULTS
4.1 Simulation Setup
A crowded wireless cell is modeled as a
100×100 meters square area full with
hundreds of mobile users with the BS at the
center of the circular cell. The LTE BS is
assumed to operate on the 20 MHz band and
we are concentrated on the uplink. Consider
a similar environment as in downtown San
Francisco where the BS can operate on
channel 26 (frequency bands 542 -
547MHz). In addition, the simplified path-
loss model [26] with path-loss exponent 4 is
assumed between the transmitters and
receivers.
4.2 Discussion and Results setup
For simulation purposes and to illustrate the
flow of the proposed algorithm, it is
assumed that the three TVWS channels are
available everywhere within the 100 × 100
area of the cell. Figure 7 (a, b) shows the
result of the clustering steps during the
resource allocation step with 100 users
during the resource allocation step. When
cluster’s hotspot cannot satisfy the per user
constraints the BS re-clusters the nodes
according to K-mean algorithm. This
proposed algorithm iteratively clusters the
nodes into hotspots and slaves, then
allocates resources with the purpose of
maximize network spectrum reuse.
Fig 7 a: iteration number 1 Fig 7 b: iteration number 8
The results in Figure 8 (a , b) show the
clustering method with 100 users, when the
number of iterations increases from 1 to 8.
The cluster number also increases and this
results in less blocked users. Blocked users
are those users who could not be allocated
resources. In DM, the BS implements
resource allocation in an OFDMA-like
manner over LTE bands. Table 2 shows that
the number of blocked users decreases as
the number of TVWS clusters (and
iterations) increases until it reaches zero
which means all the users are covered. Re-
clustering is chosen if the current resource
allocation to all users does not satisfy their
rate requests. . Table 3 shows the power in
each modes within 100 users.
Users No.
100 users
Iteration #
1
2
3
4
5
6
7
8
No. of Clusters
2
3
4
5
6
7
8
9
Restricted users
50
35
20
6
4
4
3
0

Table 3: calculating all DM and RM modes total power in
mW
As shown previously, when we have 100
users we can form 9 clusters supported by 9
hotspots. When a node is chosen to be a
hotspot, it has to satisfy its own rate
requirement in addition to its slaves’ rate
requirements. Table 4 shows the total
required power in each cluster .The results
show that the HS power depends upon the
number of slaves it is serving and its
distance from BS.
Table 4: Power in each cluster made by a hotspot
Figure 8 shows a comparison between the
three modes for different numbers of users
ranging between 100 and 120 users. The
results show the power savings achieved in
Relay Mode operation compared to the
Direct Mode when increasing the number
of users.
Fig 8: Transmission power vs. Number of Users
Table 5: the four modes transmission power vs. 100-120
Users.
Table 5 shows the results in numbers, that
by keeping the amount of resources fixed
and increasing the number of users from
100-120, the power will be saved by
employing the (LTE+TVWS) Direct Mode
and Relay Mode. By increasing users
number and when the cell becomes dense
the average percentage between the Direct
Mode and Rely Mode will increase as well.
Total Direct Mode Power (
DM: LTE )
24047.7017
Total Direct Mode Power (DM:
LTE +TVWS)
18808.4352
Total RM Mode Power
11401.258
Pct. DM & RMII difference
71.3476%
difference
Hot-Spot #
Power of each HS mW
HS1
0.028083
HS2
0.4361
HS3
0.027021
HS4
1.2208
HS5
0.49723
HS6
1.0493
HS7
0.033473
HS8
0.47883
HS9
3.7646
Total power of Clusters
7.535437
No. of
Users
100
110
120
DM:
LTE
24047.702
30047.933
36045.748
DM:
LTE+TVWS
18808.435
26303.581
33808.303
RM:
11401.258
22236.925
25955.836
Total No.
HS
9
9
10
Pct. DM &
RMII
difference
71.34%
29.8787%
32.5473
%

5 CONCLUSION
Due to the limited spectral resources and the
large signaling overhead in dense areas, one
Base Station cannot accommodate these
large numbers of users simultaneously. In
this paper, we have proposed an algorithm
for clustering mobile nodes and organizing
them into hotspots and slaves and then
allocate LTE and TVWS resources
dynamically with the objective of reducing
total number of restricted users and network
transmission power. Simulation results show
that the total users within the coverage area
can appreciably increase in the proposed
algorithm using hotspot-slave configuration
(Rely Mode) compared to conventional
(LTE) direct mode communication.
Acknowledgement
This paper was funded by King Abdulaziz
City for Science and Technology (KACST),
titled: "High Density Wireless Access-
Coverage for Crowded Events and
Locations" (under grant numbered: 12-
INF2743 03) the author, therefore,
acknowledges with gratitude KACST's
technical and financial support. Also, the
author would like to thank Dr. Haleh Tabrizi
and Dr. Tahir Bakhsh for their support and
guidance.
References
[1] A. Khandekar, N. Bhushan, J. Tingfang,
and V. Vanghi, "LTE-advanced:
heterogeneous networks," in Wireless
Conference (EW), 2010 European, pp.
978-982.
[2] N. Akkari, G. Aldabbagh, M. Nahas,
and J. Cioffi, "Dynamic Clustering
Protocol for coordinated tethering over
cellular networks," Journal of Network
and Computer Applications, vol. 42, pp.
92-101, 2014.
[3] A. Khalifah, N. Akkari, and G.
Aldabbagh, "Dense areas femtocell
deployment: Access types and
challenges," in e-Technologies and
Networks for Development (ICeND),
2014 Third International Conference on,
pp. 64-69.
[4] N. Akkari, G. Aldabbagh, M. Nahas, B.
Bawazeer, J. Cioffi, and H. Tabrizi,
"Coordinated tethering over cellular
networks: Handover scenarios and
signaling," in Personal Indoor and
Mobile Radio Communications
(PIMRC), 2013 IEEE 24th International
Symposium on, pp. 2170-2174.
[5] H. Tabrizi, G. Farhadi, J. Cioffi, and G.
Aldabagh, "Coordinated Tethering over
White-Spaces," 2014.
[6] G. Aldabbagh, S. T. Bakhsh, N. Akkari,
S. Tahir, H. Tabrizi, and J. Cioffi, "QoS-
Aware Tethering in a Heterogeneous
Wireless Network using LTE and TV
White Spaces," Computer Networks,
vol. 81, pp. 136-146, 2015.
[7] G. Aldabbagh, S. T. Bakhsh, N. Akkari,
S. Tahir, S. Khan, and J. Cioffi,
"Distributed dynamic load balancing in
a heterogeneous network using LTE and
TV white spaces," Wireless Networks,
pp. 1-12, 2015.
[8] W. Stallings, Wireless communications
& networks: Pearson Education India,
2009.
[9] http://www.thinkFemtocell.com

/System/hetnetsheterogeneous-
networks-and-Femtocells.html.
[10] D. Goldman, "Why no one got a wi-fi
connection at mobile world congress ".
vol. 2011, 2011.
[11] C. R. Stevenson, G. Chouinard, Z. Lei,
W. Hu, S. J. Shellhammer, and W.
Caldwell, "IEEE 802.22: The first
cognitive radio wireless regional area
network standard," IEEE
Communications Magazine, vol. 47, pp.
130-138, 2009.
[12] C.-s. Leem, J. Lee, H. Kang, C.-J. Kim,
M. Lee, and S.-c. Kang, "Making the
Best out of Spectral Efficiency; Studies
on The Introduction of Open-Spectrum
Policy," in Cognitive Radio Oriented
Wireless Networks and
Communications, 2008. CrownCom
2008. 3rd International Conference on,
2008, pp. 1-4.
[13] U.S. FCC and ET Docket 02-380,
"Notice of Inquiry, in the matter of
Additional Spectrum for Unlicensed
Devices Below 900MHz and in the
3GHz Band," Dec. 20, 2002.
[14] J. D. Deaton, M. Benonis, L. DaSilva,
and R. E. Irwin, "Supporting dynamic
spectrum access in heterogeneous lte+
networks," in Dynamic Spectrum Access
Networks (DYSPAN), 2012 IEEE
International Symposium on, pp. 305-
316.
[15] P. Bahl, R. Chandra, T. Moscibroda, R.
Murty, and M. Welsh, "White space
networking with wi-fi like
connectivity," ACM SIGCOMM
Computer Communication Review, vol.
39, pp. 27-38, 2009.
[16] H. Tabrizi, G. Farhadi, and J. M. Cioffi,
"Tethering Over TV White-space:
Dynamic Hotspot Selection and
Resource Allocation," in Vehicular
Technology Conference (VTC Fall),
2013 IEEE 78th, pp. 1-6.
[17] L. Lei, Z. Zhong, C. Lin, and X. Shen,
"Operator controlled device-to-device
communications in LTE-advanced
networks," IEEE Wireless
Communications, vol. 19, p. 96, 2012.
[18] H. Xing and S. Hakola, "The
investigation of power control schemes
for a device-to-device communication
integrated into OFDMA cellular
system," in Personal Indoor and Mobile
Radio Communications (PIMRC), 2010
IEEE 21st International Symposium on,
pp. 1775-1780.
[19] H. Nourizadeh, S. Nourizadeh, and R.
Tafazolli, "Performance evaluation of
cellular networks with mobile and fixed
relay station," in Vehicular Technology
Conference, 2006. VTC-2006 Fall. 2006
IEEE 64th, 2006, pp. 1-5.
[20] O. A. Abbas, "Comparisons Between
Data Clustering Algorithms," Int. Arab
J. Inf. Technol., vol. 5, pp. 320-325,
2008.
[21] J. Y. Yu and P. H. J. Chong, "A survey
of clustering schemes for mobile ad hoc
networks," IEEE Communications
Surveys and Tutorials, vol. 7, pp. 32-48,
2005.
[22] J.-h. Ryu, S. Song, and D.-H. Cho,
"New clustering schemes for energy
conservationin two-tiered mobile ad hoc
networks," Vehicular Technology, IEEE
Transactions on, vol. 51, pp. 1661-1668,
2002.

[23] M. Chatterjee, S. K. Das, and D. Turgut,
"An on-demand weighted clustering
algorithm (WCA) for ad hoc networks,"
in Global Telecommunications
Conference, 2000. GLOBECOM'00.
IEEE, 2000, pp. 1697-1701.
[24] T. D. Todd and D. Zhao, "Cellular
CDMA capacity in hotspots with limited
ad hoc relaying," in Personal, Indoor
and Mobile Radio Communications,
2003. PIMRC 2003. 14th IEEE
Proceedings on, 2003, pp. 2828-2832.
[25] H. Tabrizi, G. Farhadi, and J. M. Cioffi,
"CaSRA: An algorithm for cognitive
tethering in dense wireless areas," in
Global Communications Conference
(GLOBECOM), 2013 IEEE, 2013 pp.
3855-3860.
[26] A. Goldsmith, Wireless
communications: Cambridge university
press, 2005.