17 15969 31364-1-sm(edit)n

Indonesian Journal of Electrical Engineering and Computer Science
Vol. 7, No. 1, July 2017, pp. 138 ~ 149
DOI: 10.11591/ijeecs.v7.i1.pp138-149  138
Received March 29, 2017; Revised June 1, 2017; Accepted June 17, 2017
Research on Resource Allocation based on
Clustering in Femtocell Networks
Ding Kai*
1
, Liu Zhanjun
2
, Zhou Shiyan
3
, Ran Weiyi
4
, Li Wei
5
1,2,3
Chongqing Key Lab of Mobile Communications, Chongqing University of Posts and
Telecommunications, Chongqing 400065, P.R. China
4,5
Bureau of Public Security of Chongqing Municipality Nanan District Branch, Chongqing 401336, P.R.
China
*Corresponding author, e-mail: 805440713@qq.com
Abstract
Aiming at the problem that existing femtocell base station searching the optimal clustering scheme
based on the clustering resource allocation algorithm is complex. We propose that building an conflict
graph and adjacency matrix before clustering to calculate the number of clusters needed for FBS group by
using the adaptive clustering heuristic algorithm. We follow maximization the sum of the FBS distances in
the cluster and group the femtocell base stations to narrow the search range to reduce the computational
complexity. In order to achieve different business types service, based on the above clustering algorithm,
this paper proposes a new method that using the weighted energy efficiency, which including the user
interruption and the network spectral efficiency as a fitness function of the power control scheme to solve
the problem. The simulation results show that the same rate requirement reduces the complexity, while the
same complexity increases the user's average rate.
Key words: femtocell, clustering, power control, resource allocation
Copyright © 2017 Institute of Advanced Engineering and Science. All rights reserved.
1. Introduction
The coverage to indoor users have a vital role on the performance of wireless
communication because of the fact that the 70% demand of users in the wireless
communication focus on indoor area. To cover indoor area by macro station, the
electromagnetic wave needs to penetrate walls, causing serious attenuation, so the indoor
communication needs to improve the covering effect. Femtocell technology employing indoor
access points to cover, received extensive attention in the industry [1-3] due to the excellent
indoor coverage effect and the characteristics of low power consumption and low cost.
However, because of sharing wireless channel, wireless resource in wireless
communication is precious. The resource allocation has become a very important research
direction. Currently existing femtocell resource allocation technology is mainly divided into three
categories: pure spectrum allocation, power control and the co-allocation of the spectrum and
power. The characteristic of pure spectrum allocation technology is the efficiency of spectrum
utilization can be improved by inhibiting interference between the femtocell. The basic idea of
power control is considering the tradeoff between network coverage and interference. The co-
allocation of spectrum and power can improve the spectrum efficiency and energy of the
network at the same time. The spectrum allocation of the fixed transmission power does not
implement the energy saving of the network. The network spectrum efficiency of the power
allocation without spectrum reuse is low. Due to the fact that network spectrum efficiency can be
improved under the condition of network energy efficiency by co-allocation of spectrum and
power. Therefore an effective spectrum and power allocation mechanism is required to promote
the network performance. But under the condition of satisfying user’s quality of service, the
oversize transmission power of FBS (Femtocell Base Station, FBS) will cause energy waste.
While reducing the transmission power can reduce energy consumption, but it is likely to reduce
the QoS of users and improve network outage probability. Obviously, the suitable downlink
transmitted power configuration of FBS is the key technology to improve the network efficiency
under the condition of satisfying the QoS of users.

IJEECS ISSN: 2502-4752 
Research on Resource Allocation based on Clustering in Femtocell… (Ding Kai)
139
The clustering is used by existing technology to resource allocation. Clustering is an
effective management method of interference, it can reduce the femtocell interference in the
same layer, and separate the whole network resource allocation problem into several clusters in
resource allocation problems, to reduce the difficulty of spectrum allocation and power control.
Through the clustering to reuse frequency in femtocell network, increase downlink transmission
power of femtocell does not necessarily cause interference to the adjacent femtocell, because
the orthogonal spectrum is used by large interference of femtocell.
But the scope of the search for the optimal clustering scheme of existing femtocell
clustering algorithms is extensive, which lead to high computational complexity, due to the fact
that the grouping is not executed before clustering femtocell. In the genetic algorithm based
femtocell power control scheme, which consider the energy efficiency as the objective function,
increasing the outage probability is the expense to improve the energy efficiency of network,
and only for single user business.
Therefore, in this paper femtocell clustering algorithm and power allocation scheme in
the network is researched. In view of the large computational complexity in existing clustering
algorithm of the femtocell, we propose the grouping femtocell according to conflict graph and
adjacency matrix in front of the clustering to shrink search scope of the optimal clustering
scheme to reduce the computational complexity of clustering algorithm. Aim at the problem that
in the existing femtocell power control scheme, increasing the outage probability is the expense
to improve the energy efficiency of network, and only for single user business, we
comprehensive consider the user interrupt problem, different business requirements and the
network spectrum efficiency weighted energy efficiency as the fitness function, to ensure power
adjustment without disrupting users communication, adapting different business needs, while
improving energy efficiency both users QoS of power control. Therefore, the reasonable
mathematical model for the rational design of clustering mechanism of resource allocation is
needed.
2. Mathematical Model
Define optimize network efficiency as the objective functions of the resource allocation:
 
,
KN
, 2 ,
1 1
,P 1 C _
N
, 2 , min
1
,
K
,
1
1 1
min max
log (1 P )
max
P
. . C1: log (1 P ) R ,
C2: 0,1 , , ,
C3: 1, , ,
C4: C ; C5: C
C6:P P P ,
f
n
k f f
l
f
n n
k f f k f
n k
l f f in
n n
k f f k f k
n
n
k f
n
k f
k
l l
l l
f
f
st f k
n k f
n k f
f
 
  


 
   
     
  
  
 
  




C
C C
F
(1)
Where,
f is the bandwidth of the channel;
Pf
is transmitted power of FBS
f ; ,
n
k f
is
user kth of FBS
th
f on the
th
n sub-channels SINR ; If the n
th
sub-channel is assigned to the k
th
users in the
th
f FBS, , 1n
k f 
, otherwise, , 0n
k f 
; when the transmitted power is
Pf
, _Pf in
is FBS actual need to consume power, we have:
_ 0 PP P Pf in f  (2)
Where, 0P is 10.1W, P is 15.

 ISSN: 2502-4752
IJEECS Vol. 7, No. 1, July 2017 : 138 – 149
140
C1 means each user’s data rate shall not be less than demand rate; C2 means value of
,
n
k f is 0 or 1; C3 means for any femtocell, each channel can only be assigned to a user; C4
means the whole cluster constitute a femtocell set F; C5 means femtocell is not repeated in
different clusters; C6 means the scope of FBS transmission power.
Both continuous variables and discrete variables are included in above optimization
problem, which is a mixed integer nonlinear programming. The computational complexity of this
problem is large and exponential increase with the number of FBS increasing. In order to
simplify and solve the above problems, the resource allocation problem is divided into two
steps, to avoid mixed integer nonlinear programming. To solve this problem by these two steps
into two sub-problems: the clustering problem to reduce the complexity and sub-channels
allocation and power allocation problem.
2.1. Clustering
For femtocell base station network, the goal of the clustering is to assign FBS which is
close and larger interference to different clusters to use different spectrum so as to improve the
spectrum utilization of the network, and makes the sum of the largest cluster of FBS in distance
in order to reduce interference so as to improve the spectrum utilization efficiency of resources
of the network. Therefore, clustering objective function can be expressed as Equation (3).
0
0 0
0
,
1 , C ,
,
1
1
max
. . C1: R 0
C2: C
C3: C
l
f f
l f f f f
f f th
l
l
l
l
d
st d
  


 

 
 
C
C
C
F
(3)
Where , 0,f fd is the distance between the femtocell f located in ( , )f fx y and
femtocell 0f located in 0 0
( , )f fx y ,we have
0 0 0
2 2
, ( ) ( )f f f f f fd x x y y    . In order to avoid
collision, any two femtocell with of the distance less than Rth must be assigned to different
clusters.
In order to solve the above optimization problem, the clustering algorithm is proposed
according to the interference figure and adjacency matrix for grouping femtocell in front of the
clustering.
(1) The construction of conflict graph
According to the definition of the graph theory [4], femtocell interference figure can be
represented by G=(V, E), among them, V is vertices set , E is edge set. The femtocell f in F
one to one correspondence with the vertices in the graph v ,f f F. E is edge between
vertices fv and 0fv with 0 0e(v ,v ), ,f f f f F, reflects the femtocell collision relations
between the interference. Specifically, if the distance between the femtocell f and femtocell
0 0( )f f f is less than Rth , that means the interference between them is larger, there is
boundary 0
e(v ,v )f f between corresponding vertex fv and 0fv .
Assume that there are six FBS in the network of femtocell, and 12 23 13 45, , , Rthd d d d  .
When the femtocell distribution as shown in figure 1, the corresponding conflict graph is shown
in Figure 2.

141
1
2
3
4
5
6
1
2
3
4
5
6
Figure 1. The distribution of femtocell Figure 2. The conflict graph of femtocell
Because of there is no interference between FBS and its attribution FUE, there is no
loop in the conflict graph. G( )fd v is the degree of vertex fv . V in the vertex has the following
properties.
Property 1 If G ( ) 0fd v  , vertex fv is isolated point, which means that there is no
collisions interference between corresponding femtocell and any other femtocells.
Property 2 If G ( ) 0fd v  , that is collision interference between femtocell f and
femtocell corresponding to the vertexes connected with vertex fv . The bigger G( )fd v is, the
bigger the femtocell f collision distractions.
(2) The construction of adjacency matrix
To express quantitatively the conflict graph G, construct adjacency matrix
. Specially, if 0, Rf f thd  , 0
( , ) 1f fa v v  , or 0
( , ) 0f fa v v  .
For conflict graph shown in Figure 2, the adjacency matrix (G)A is:
0 1 1 0 0 0
1 0 1 0 0 0
1 1 0 0 0 0
0 0 0 0 1 0
0 0 0 1 0 0
0 0 0 0 0 0
 
 
 
 
 
 
 
 
 
(3) Clustering
Clustering [5] goal is to divide the vertices connected in interfere figure into different
clusters, makes femtocells have interfere collision relationship use different sub-channels.
2.2. Sub-channel and Power Allocation Sub-problem
(1) Sub-channel allocation [6]
First of all, according to the number of clusters to differentiate frequency band, distribute
orthogonal spectrum for different clusters. Due to there is no interference between femtocell
within the cluster, the cluster is spectrum reuse unit, namely all femtocell within clusters use the
same spectrum. It is beneficial to reduce the difficulty of the spectrum allocation, improve
spectrum efficiency. It units of the In the communication system using resource block as unit,
can put the total number of available resources block according to the number of clusters C
group first, then assigned the different groups of blocks resources into different FBS.
Assume that each cluster of the available bandwidth is equal, for each FBS usable
spectrum bandwidth:
B W/ C (4)
0 0
, V(G) [ ( , )] f ff f v va v v A

 ISSN: 2502-4752
142
Where, W with the total bandwidth available for femtocell network.
According to the literature [8], within each femtocell, can allocate sub-channel by using
Equation (5).
,
,
P
( ) argmax
G
f k f
f
k f
k n

 (5)
Where, ( )fk n refers to n sub-channels assign to service users k of femtocell f.
(2) Power allocation
After frequency reuse between the FBS, due to large disturbance of femtocell using
orthogonal spectrum, increase the FBS downlink transmition power does not necessarily cause
disturbance to the adjacent femtocell, but under the condition of satisfying user’s QoS, larger
femtocell transmission power will cause energy waste. However, reduce the transmission power
can reduce energy consumption, but it is likely to reduce the user’s QoS, increase outage
probability of network. Obviously, the appropriate FBS downlink transmition power configuration
is on the premise of guarantee the quality of customer service is the key to reduce the network
energy consumption, therefore, effective power control must be conducted on the basis of
spectrum allocation.
The existing power control scheme based on genetic algorithm use the energy
efficiency as fitness function:
K N
,
1 1
_
C
EE
P
f
n
k f
k n
f
f in
f
 
 


(6)
Where,
EEf
is energy efficiency of femtocell f, defined as the ratio between femtocell
throughput and FBS actual consumption power;
f is the bandwidth of the sub-channel, which
unit is Hz; ,Cn
k f is transmission capacity on the unit frequency band, can be obtained by
truncation Shannon formula:
 
,min min
min max
m
, 2 ,
ax , ax
,
m
C
C log 1
C
n
k f
n n n
k f k f k f
n
k f

   

     
   



(7)
Where,  is attenuation factor, means the signal attenuation in the process of actual
transmission.
The fitness function of Equation (6) only considers the energy efficiency of the femtocell
and in the process of power adjustment can reduce the user’s QoS. In order to ensure that the
user service is not interrupted, introduce the concept of weighted energy efficiency:
min
_
_
,0 P( ) 0
EE
C EE
n
k
f w
f n f
k f
else
    
 
 (8)
Where, , minP( ) 0n
kk f   
is the user’s outage probability, suggests that when user’s
SINR is less than the threshold values mink , communication will interrupt. Because of the
communication between the user’s business is often not identical, according to the required

143
transfer rate of users calculate user outage probability is more tally with the actual; _Cf n
is the
normalized unit frequency band transmission capacity:
_
max
C
C
C
f
f n 
(9)
Where,
Cf
is femtocell f the average transmission capacity on the unit band:
K
,1
C
C
K
f
k fk
f
f



(10)
The advantage of introducing _Cf n
as energy efficiency has weight avoiding the user in
the process of power control by the user’s transmission rate and the femtocell energy efficiency
limit caused by communication interrupt.
Using weighted energy efficiency as an individual fitness function is to provide the
safeguard for user communication interrupt. This is because when a user receives the signal
SINR is too low and can’t communicate, the attribution of FBS fitness value is 0, so that they
can avoid some FAP because of the transmitted power is too small and not services. Therefore,
employing the weighted energy efficiency in Equation (8) as an individual fitness function in
power control can ensure users QoS in the premises of improving the efficiency of the network.
3. The femtocell in the network based on clustering of resource allocation mechanism
First of all, clustering the femtocelling; then, sub-channel and power allocation [9] within
each cluster. Simplified flow chart of resource allocation scheme is shown in Figure 3.
Start
Constructing Conflict
graph、adjacency matrix
Clustering
Subchannel allocation
Power allocation
End
Figure 3. The resource allocation scheme based on clustering in femtocell networks

 ISSN: 2502-4752
144
Based on the conflict graph and adjacency matrix clustering detailed steps are as
follows:
Initialization initialize the cluster number ( Z )l l 
 to 1; Y is isolated point set;
Step 1 Add all vertex with degree 0 to Y, and remove the vertices in V;
Step 2 If V  , add all the femtocell to clusters Cl ,and go to step 8, otherwise go to
step 3;
Step 3 If V  , blocking the adjacency matrix, then each partitioned matrix
corresponding to a FBS group, denoted as G ( 1,2, , )i i w , which is the number of FBS
group ,set the number of vertices in a FBS group ( 1,2, , )i i w for Ni ;
Step 4 ACHA algorithm is used to clustering each FBS group, each of the FBS group
clustering results are obtained, denote the required number of clusters for the th a FBS group
as Li ， maxL =i
i
C ；
Step 5 Find any FBS group G j with the number of clusters C , according to the
clustering results will step 4 vertices joined in this FBS group C in a cluster, and remove the
vertices from the V, If V  , then go to step 7, otherwise go to step 6;
Step 6 If V  , there are
L
A i
C kinds of scheme of adding the vertex in the ( )i i j
FBS into C cluster. Based on the principle of largest FBS distance in cluster , choose the
solution meet 0
0 0
,
1 , C ,
max
l
f f
l f f f f
d
  
 
C
, add the FBS group vertices corresponding clusters, and
remove the vertices from the V, if V   ,then go to step 7, otherwise repeat step 6;
Step 7 According to the type (11), add the first vertex of Y into the cluster
* *
( {1,2, , })l l  C :
'
'
*
,Y
C
argmax
l
f ff
f
l d


  (11)
Remove the vertices from the Y, if ,then go to step 8, otherwise repeat step 7;
Step 8 End of the clustering.
Clustering process as shown in Figure 4.
If the total number of FBS is F, for the clustering algorithms, which don’t grouping FBS
in front of the clustering , clustering result can be obtained by a second Stirling number:
1 0
1
( 1) ( )
!
i jF i
F F
i j
i
j O F
ji

 
 
  
 
 
(12)
In this paper, the number of the clustering algorithm for clustering scheme may be:
L L
1
C Ai i
i
w
N
i
 C
(13)
We can see, in this paper, the algorithm of the optimal clustering scheme greatly
narrowed the search scope, thus reducing the complexity of clustering, and FBS in the network,
the more the degree of complexity, the more obvious. According to the steps 4 and 7, the
algorithm computational complexity is
( )O F .
i

145
Start
According to the position of FBS construct
conflict graph and adjacency matrix
Calculate the degree of vertices in the graph
Remove the vertices with degrees 0 and add
them to the independent vertex set
Update the Conflict graph and adjacency matrix
The conflict graph is empty
The adjacency matrix is chunked to group the
FBS
The FBS groups were clustered by ACHA
algorithm to obtain the clustering results of each
FBS group and the number of clusters
required，denoted as the maximum value of
The vertices in the FBS group corresponding to
the maximum value are added to the clusters
according to the clustering result of the previous
step，and remove these vertices
According to the principle of maximum FBS
distance in the cluster, the vertices in any FBS
group are added to the corresponding clusters,
and remove these vertices
No
No
The independent vertex set is empty
The vertices of the independent vertex are
added in turn to the cluster that makes the
maximum distance between the vertices in
the cluster
Yes
No
End
Yes
No
Yes
Yes
C
C
Figure 4. The femtocell base station clustering algorithm flow chart based on grouping

 ISSN: 2502-4752
146
3. The Results of Simulation and Analysis
The MATLAB simulation parameters and parameter values as shown in Table 1.
Table 1. The System Simulation Configuration Parameters
The parameter name Parameter values
MBS coverage radius(RM)/m 500m
FBS transmission power/ dB 46dB
FBS coverage radius(RF)/ m 50m
FBS maximum transmitted
power(Pmax)/ dBm
20dBm
FBSminimum transmitted
power(Pmin) /dBm
0dBm
Cmax/( bits/Hz), Cmin/ (bits/Hz) 4.4,0 bits/Hz
max dBγ min dBγ 19.5,-10 dB
The standard deviation of shadow
fading / dB
8dB
The minimum coupling loss (MCL)/
dB
70dB
Attenuation factor ( ) 0.6
The system total bandwidth (W)/
MHz
10MHz
AGWN unilateral power spectral density
(N0)/ (dBm/Hz)
-174 dBm/Hz
Carrier frequency( )/GHz 2 GHz
User noise factor/ dB 9 dB
In this section, compared with the high energy-efficient resource allocation scheme
based on user QoS guarantees and the resource allocation scheme based on the UFR, the
clustering algorithm that did not group the FBS before clustering, respectively.
(1) The the user average data rate of proposed resource allocation scheme based on
clustering algorithm is compared with the user average data rate of the whole frequency
multiplexing and the resource allocation scheme using the existing clustering algorithm.
As can be seen from Figure 5, with the increase in the number of FBS, the network
average user data rate showed a downward trend. This is because the more FBSs, FUEs sufer
from greater interference, SINR lower, the user data rate reduction. Compared to UFR, the
average user data rate based on clustering [10] resource allocation is higher because clustering
reduces the interference among with femtocells and increases the SINR of FUEs. In the case of
low complexity of the algorithm, the clustering algorithm of this paper can obtain the average
user data rate which is equivalent to the clustering algorithm which without grouping FBS before
clustering.
(2) Compared with high energy-efficient power control based on user QoS guarantees
and network energy efficiency and user average data rate based on energy efficiency power
control.


cf

147
Figure 5. The average user data rate under different clustering algorithms
Figure 6. The network energy efficiency comparison under different power control scheme
As can be seen from Figure 6, with the increase in the number of FBS, the network
energy efficiency is declining, because FBS is increasing, the network load is increasing, the
base station needs to increase the power to expand coverage. Compared with the power
control only based on energy efficiency, the network energy efficiency under the high energy
efficiency power control based on user’s QoS guarantee is a bit low, because the system as
much as possible to ensure the user’s QoS when it is making power adjustments to improve
energy efficiency. Therefore, it can provide a higher average user data rate, as shown in
Figure 7.

 ISSN: 2502-4752
148
Figure 7. The average user data rate comparison under different power control scheme
4. Conclusion
Aiming at the problem that the complexity of the existing femtocell clustering algorithm
is large, it is proposed to construct the conflict graph and adjacency matrix before clustering,
and to block the adjacency matrix. The number of clusters required for each FBS group is
calculated by using the adaptive clustering heuristic algorithm. The sum of the FBS distances in
the cluster is maximized according to the principle of maximizing the sum of FBS distances in
the cluster. In this paper, the clustering algorithm is used to reduce the search range of the
optimal clustering scheme, so as to reduce the computational complexity of the clustering
algorithm. Aiming at the problem that the improvement of the energy efficiency of the network
under the power control scheme which at the expense of increasing the outage probability and
the problem of the single service a power control scheme which based on the weighted energy
efficiency as the fitness function to ensure that the user’s communication is not interrupted,
meet the different needs of business and improving network spectrum efficiency during the
power adjustment process is proposed in this paper. According to the above-mentioned
clustering algorithms and power control schemes, an energy-efficient resource allocation
scheme based on user QoS guarantee is designed. The scheme first grouping for all femtocells
to eliminate the same level of interference among the base stations and to improve the
spectrum utilization of the network. Then dividing spectrum and advocating sub-channel
according to the clustering results for the femtocell. Finally, power control for femtocell to
improve the energy efficiency of the network. The simulation result shows that the average user
data rate based on the clustering resource allocation scheme is higher than that of UFR.
Compared with the clustering algorithm which does not grouping the femtocell before clustering,
the proposed clustering algorithm can obtain the same average user data rate in the case of low
computational complexity. Compared with the energy efficiency-based resource allocation
scheme, the network energy efficiency of the two schemes is comparable, but the average user
data rate of the energy-efficient resource allocation scheme based on user QoS guarantees is
higher.
Reference
[1] Y. L. Lee, J. Loo, T. C. Chuah and A. A. El-Saleh, Fair resource allocation with interference mitigation
and resource reuse for LTE/LTE-A femtocell networks. IEEE Transactions on Vehicular Technology.
2016; 65(10):8203-8217.
[2] M. Lin, S. Silvestri, N. Bartolini, T. F. La Porta. On selective activation in dense femtocell
networks. IEEE Transactions on Wireless Communications. 2016; 15(10):7018-7029.

149
[3] Y. J. Liang, Y. L. Tsai and K. Chiang. A dynamic resource allocation scheme in femtocell networks
considering user mobility, 2016 Asia-Pacific Conference on Communications (APCC), Yogyakarta,
2016: 191-196.
[4] Gibbons A. Algorithmic graph theory. Sydney: Cambridge University Press, 1994.
[5] Estrada R, Jarray A, Otrok H, et al. Base station selection and resource allocation in macro-femtocell
networks under noisy scenario. Wireless networks. 2014, 20(1): 115-131.
[6] Li Wei, Zheng Wei, Wei Xiangming, et al. Dynamic clustering based sub-band allocation in dense
femtocell environments. 2012 IEEE 75th Vehicular Technology Conference (VTC Spring). Yokohama,
2012: 1-5.
[7] Ji Lei, Chen Qianbin, Tang Lun. Load forecasting based femtocell clustering spectrum allocation
algorithm research in heterogeneous network. Science Technology and Engineering. 2015; 15(13).
[8] Abdelnasser A, Hossain E, Kim I. Clustering and resource allocation for dense femtocells in a two-tier
cellular OFDMA network. IEEE Transactions on Wireless Communications. 2014; 13(3): 1628-1641.
[9] Kim G, Kwon A, JW L. Utility-based subchannel allocation for OFDMA femtocell networks. 2011
Proceedings of 20th International Conference on Computer Communications and Networks (ICCCN).
Maui: IEEE Press, 2011: 1-6.
[10] Chatterjee M, Das K, Turgut D. WCA. a weighted clustering algorithm for mobile Ad Hoc networks.
IEEE Journal of Clustering Computing, 2002; 5(2): 193-204.

17 15969 31364-1-sm(edit)n

More Related Content

What's hot (19)

Similar to 17 15969 31364-1-sm(edit)n (20)

More from IAESIJEECS (20)

Recently uploaded (20)

17 15969 31364-1-sm(edit)n