power control game performance in cognitive femtocell network · power control game performance in...
TRANSCRIPT
Power Control Game Performance in Cognitive Femtocell
Network
Anggun Fitrian Isnawati, Wahyu Pamungkas, and Jans Hendry Faculty of Telecommunication and Electrical Engineering, Institut Teknologi Telkom Purwokerto, 53147, Indonesia
Email: [email protected]; [email protected]; [email protected]
Abstract—The growing number of the implementation of a
cognitive femtocell communication system as a micro-scale
communication system is caused by the fact that it presents a
solution to the macro-scale communication system regarding
the deployment flexibility, low cost in network development,
and improving user quality at cell edge area. However, this
condition also brings disadvantages related to user interference.
One way to overcome the problem is by applying a power
control system. Due to the dynamic nature of the user in the
cognitive femtocell network, the user's own power control
method is appropriate to be applied. The self-power control
method in this study was based on the game theory approach or
commonly known as the power control game (PCG). The
number of users was firmly related to the performance of PCG.
The results show that as the number of users increased, the
power consumed and the number of iterations needed to achieve
the condition of Nash equilibrium also increased. But the
increase in the number of users reduced the achieved SINR. Index Terms—power control, cognitive femtocell, interference,
PCG, Nash equilibrium
I. INTRODUCTION
The main problem for mobile network micronization is
that the infrastructure network is still expensive to
develop. The latest development of network
micronization is the femtocell network, also called base-
stations, with short, and low-cost scope areas, and can be
installed by consumers for improved indoor voice and
data reception [1].
Utilization of cognitive radio technology on cellular
networks (cognitive radio network, CRN) strongly
supports the development of mobile communication,
especially femtocell network technology. Development of
cognitive femtocell network (often referred to as
cognitive femtocell network, CFN) can provide cost-
effective improvement solutions in several scenarios
related to spectrum scarcity [2].
The disproportionate use of transmitting power from
any user causes interference. Interference between users
happens to multiple users either in one femtocell or
different femtocell but using the same channel (resource
block, RB) at a time [3]. Therefore, it is necessary to have
an uplink power control system applied to the user side to
control the amount of generated interference amongst
cells so as to minimize the interference that occurs [4].
Manuscript received July 25, 2018; revised January 16, 2019. Corresponding author email: [email protected].
doi:10.12720/jcm.14.2.121-127
The purpose of power control is to ensure that the
transmitting power of each transmitter can attain a high
enough level to be detected by the receiver and low
enough to avoid interference to other users. However, the
purpose is inseparable from the trade-off on the mobile
system.
Higher user SINR yield better communication service,
but as consequence it requires higher power consumption,
thus, the battery lifetime will be shorter and interferes
other users's signals. Therefore, an effective power
control algorithm must be applied to control interference
to all users by reducing the transmitted power of each
user to ensure the optimal power distribution in the
system [5]. In this case, the user should lower the
transmitting power if user is close to the base station and
increases the transmitting power when far away from the
base station.
The use of game theory algorithm on power control is
originated from the existence of conflicts among self-
organized users in non-cooperative power control. Game
theory is in accordance with the character of the user who
is distributed, organized independently (self-organized),
and non-cooperative. Non-cooperative game theory
characteristics that use strategy methods to users without
having to obtain global information from the entire user
are suitable if applied to femtocell networks with user
characteristics as mentioned. A previous study reported a
game theory applied to many players with incomplete
information because the players are not in mutual
coordination and not cooperative in the strategy selection
[6]. In the wireless communication, each user has the
same desire that selfishly raises the maximum power to
meet the target SINR; thus, generating interference to
other users. Hence, the need for game theory method on
the power control system for the selection of user power
strategy independently is essential.
Power control Game method proposed by Koskie Gajic
or known as KG algorithm has not accommodated the
effect of channel addition to SINR and user transmitting
power. Even though the convergence speed is almost the
same, it cannot achieve the SINR target [7], [8]. Similar
to KG, the method presented by Al Gumaei [9] also could
not meet the SINR target. Meanwhile, Thalabani [10]
presented a technique where the user consumed a lesser
power than the KG algorithm when reaching convergence,
but the SINR target was not met. The obtained SINR was
even lower than the value in the KG algorithm. In
121©2019 Journal of Communications
Journal of Communications Vol. 14, No. 2, February 2019
methods proposed by Zhao Chenglin [11] and Luyong
Zhang [12] the obtained SINR was remarkably higher
than the SINR target. However, the provided Quality of
Service (QoS) was excessive in a communication system
that does not need high quality. It also consumed more
extensive power than necessary. Both methods improved
the KG algorithm in terms of reaching the SINR target,
but the obtained value was excessive.
The condition of users who dynamically changes will
affect system performance. In systems with self-
organizing users, changes in the number of users can
affect how quickly the system can achieve convergence
(stable) conditions while maintaining SINR according to
the target at optimal power. The effect of increasing the
number of users on femtocell communication networks
also affects system performance. Previous studies have
presented methods with low power and high convergence,
but they have low qualities in terms of achieving the
SINR target. The SINR obtained by previous methods are
lower than the target or remarkably higher than the target,
which consumes more extensive power than necessary.
This study analyzed the performance of the proposed
Power Control Game (called PCG Proposed) compared to
the previous methods under changing user conditions.
The performance was assessed in terms of user SINR
parameters, user power, and the time and the number of
iteration required to reach convergence (Nash
equilibrium).
The rest of this paper is shown as follows. The system
model is described in Section II, and Section III explains
the power control game in the cognitive femtocell
network. Section IV shows the simulation results of the
comparison among power control game methods and
distributed power control method, and conclusions are
given in section V.
II. SYSTEM MODEL
The system model used in this study consists of several
pairs of femtocell user equipment (FEU) as a transmitter
and a Femtocell Access Point (FAP) as a receiver. The
system model used three user schemes, which were three-
user, five-user, and ten-user schemes. The use of the three
schemes aimed to analyze the effect of user addition on
the PCG performance. The number of channels used in
each scheme was identical, which was four channels.
There are some pairs of cognitive user transmitter and
receiver in the system model. The solid lines represent
the communication link, while the dashed ones indicate
the interference links, as shown in Fig. 1.
Fig. 2 shows the flowchart of the Proposed PCG that
explains the process of changing the power strategy based
on the power update iteration method. The power update
formula was generated from the novel utility function of
the Proposed PCG method.
User 1
User 2
User 3Ch 2
Ch 1
Ch 3
User
Group
Channel
GroupCh 1
Ch 2
Ch 3
Ch 3
Ch 4
FUE 1
FUE 2FUE 3
FAP 1
FAP 2
FAP 3
Signal Interference
Fig. 1. System model for three-user scheme
Start
End
Initialization Pi
and Pmax
Pi < Pmax ?
Increasing Pi by Power Update
Proposed Power Update Formula
P(t+1) = P(t)?
No
Yes
|P(t+1)-P(t)|�Ɛ?
P(t+1) Pmax?
P(t+1) = Pmax = P*
Convergence
condition with
P(t+1)=P*
No
No
Yes
Yes
Yes
No
Proposed Utility
Function
Fig. 2. Flowchart of the proposed PCG
The utility function and power update formulas in this
study were [13]:
2
2 2 tar
i i i i k i i iU a p b c (1)
122©2019 Journal of Communications
Journal of Communications Vol. 14, No. 2, February 2019
( ) ( ) 3
( 1)
( ) ( ) 2
( )
( )
t tt tar i i i
i k t t
i i i
p a pp
c
(2)
with ai, bi and ci are the constants, pi is the i-th user power,
k is the channel sharing factor, γi is the user SINR, γtar
is
the SINR target, pi(t)
and pi(t+1)
are the i-th user power at
time (t) and (t+1).
Start
End
Initialization Pi
and Pmax
Pi < Pmax ?
Increasing Pi by Power Update
DPC Power Update Formula
P(t+1) = P(t)?
No
Yes
|P(t+1)-P(t)|�Ɛ?
P(t+1) Pmax?
P(t+1) = Pmax = P*
Convergence
condition with
P(t+1)=P*
No
No
Yes
Yes
Yes
No
Fig. 3. Flowchart of the DPC
As can be seen in Fig. 2, when the difference between
the most recent power and the previous power value was
large (larger than ε), the process returned to the start,
which was recalculating the user SINR user and
conducting the power update process again. When the
difference between the most recent power and the
previous power was tiny (smaller than ε) then the
convergent condition or Nash equilibrium (NE) was
obtained, and the resulting power was determined as the
user transmitting power. The ε was set to a tiny value that
was predetermined as the reference of iteration
convergence of the system, which also shows the
computation accuracy and the ε value in this study was
2.10-13
.
In Fig. 3 on the flowchart of Distributed Power Control
(DPC), it is explained that the power control process
starts from the user's power initialization and the user's
maximum power determination. As the initial stage, the
power of the user is compared to the maximum power, if
greater than maximum power then user power is set to
maximum power.
The power update process will only be done if the user
power is less than the maximum power, using the DPC
power update equation. The DPC power update equation
or better known as Power Balancing Algorithm (PBA) is
only affected by the SINR target, SINR user and previous
power users. This process will stop if the difference
between current power and the previous power is tiny
(smaller than ε) which means that convergent conditions
have been reached and the value of power being used by
the user is denoted by P*.
III. POWER CONTROL GAME
A. The Concept of Power Control
Each telecommunication device requires power to
support telecommunication devices and for
communication processes, in this case emitting power to
send signals containing data/ information.
Telecommunication devices that require power control
systems in this study may be cell phones, laptops or other
mobile devices.
Based on the two types of power control namely
Centralized Power Control (CPC) and Distributed Power
Control (DPC), the network with centralized power
control can achieve better performance than with
distributed control. Centralized algorithms can efficiently
achieve optimal solutions with low communication costs
and computationally uncomplicated when on a small
scale system. Because of the increase in femtocell
network size and the uncertainty about the location of the
Femto User Equipment (FUE) and the number of
Femtocell Access Points (FAP), it becomes impractical to
use centralized algorithms. The development of a DPC
algorithm is a very challenging task because the
algorithm must work without global information and
coordinated by the central control station. But distributed
power control is able to avoid the bottleneck effect of
centralized power control and can improve reliability by
eliminating the failure effects at the central station, so this
is advantageous from the point of view of implementation.
Distributed power control algorithm works without
global information and is coordinated by central control.
The distributed power control avoids the bottleneck effect
of centralized power control and can improve reliability
123©2019 Journal of Communications
Journal of Communications Vol. 14, No. 2, February 2019
by eliminating the effects of failure in the central;
therefore, it is beneficial to be implemented.
In the DPC, the iteration process is perfomed because
each user functions as a controller either for himself or
for other users. Power updates are performed by each
user to achieve convergent conditions. Determining new
power value in the power update process related to the
previous power value used by the user. The equation is
expressed as follows [14]:
( 1) ( )
( )
tart ti
i it
i
p p
(3)
The equation is commonly known as Power Balancing
Algorithm (PBA). Power updates on the DPC are
obtained based on the user SINR and user power
conditions.
B. The Power Control Game (PCG) Method
On the power update equation of PCG, the power
update component is in the first part of Equation (3). The
component also appears on previous PCG methods, such
as:
1. Koskie Gajic (KG) [15]:
2( ) ( )
( 1)
( ) ( )
t tt tar i i i
i i t t
i i i
p a pp
c
(4)
2. Zhao Chenglin (ZC) [11]:
( ) ( )( 1)
( ) ( )
21t t
t tari i ii it t
i i i i
p b pp
c
(5)
3. Al Gumaei (AG) [9]:
( ) ( )( 1)
( ) ( )
t tt tari i i
i it t
i i i
p b pp
c
(6)
4. Luyong Zhang (LZ) [12]:
1 12 2( )
( 1)
( ) 2
tt tari i
i it
i i i
p bp
c
(7)
5. Thalabani (TB) [10]:
,
( ) ( ) ( ) 3( 1) ( )
( ) ( ) ( ) 2
( )
( )tar i
t t tt tar ti i i i
i it t t
i i i i
p p b pp
c
(8)
IV. SIMULATION RESULTS
The study used three schemes, which were three-user,
five-user, and ten-user schemes using four channels in
every scheme. Three parameters were analyzed in every
scheme, user power requirement, SINR value, and the
number of iteration that represented the convergence rate.
The number of users was varied in each scheme to
analyze the effect of the addition of user on the system
performance based on the assessed parameters.
Fig. 4. Comparison of SINR generated by each method and SINR target
on the three-user scheme
Fig. 5. Comparison of SINR generated by each method and SINR target
on the five-user scheme
Fig. 6. Comparison of SINR generated by each method and SINR target
on the ten-user scheme
The determination of the performance parameters
value on DPC and PCG methods was performed when the
user reached convergence. The results of SINR
comparison reached by the users were based on DPC
method and some PCG methods for the three schemes are
displayed in Figs 4, 5 and 6. The figures show that the
SINR value achieved in the DPC method had the
maximum limit equal to the SINR target of 5 dB.
Meanwhile, the PCG methods had different results. The
KG, TB and AG methods could not meet the SINR target,
124©2019 Journal of Communications
Journal of Communications Vol. 14, No. 2, February 2019
while other PCG methods such as LZ, ZC, and Proposed
PCG could meet the SINR target, even surpassing it.
As can be seen on Figs. 4, 5, and 6, the LZ and ZC
methods always had the remarkably higher SINR value
compared to the other methods. This is good for the
system because it means the achieved quality of service
(QoS) was outstanding, but the power-used by both
methods were extremely large, which was wasteful and it
generated larger interference to other.
Changes in the number of users affected the limits of
SINR values achieved by the user. Figures 4, 5, and 6
showed that as the number of users increased, the
achieved SINR also decreased. It is caused by the
decreasing quality of the signal obtained by the user due
to the increasing number of users who communicate
simultaneously in the system. The comparison of the
achieved SINR and the number of users can be seen in
Table I.
Table I displays the SINR achieved by three methods
of KG, TB, and AG in all three schemes had similar
characteristics that were lower than the SINR target
during convergent conditions. Meanwhile, the three
methods of Proposed, ZC, and LC on all three schemes
had the SINR values greater than the SINR target in
convergent conditions.
TABLE I: COMPARISON OF USER SINR
Method User SINR (dB)
Three-user Five-user Ten-user
AG 4.83826 4.87746 4.81895
KG 4.99990 4.99839 4.99504 TB 4.87500 4.87479 4.87308
DPC 5 5 5
Proposed 5.49999 5.49977 5.49783 ZC 29.93543 11.20542 8.542164
LZ 78.60130 19.44684 11.16670
Other than the SINR value, the required power is also
an essential parameter in analyzing the system
performance. The user power requirements on the three
schemes are shown in Figs. 7, 8, and 9. The figures show
that as the number of users increased, the required power
also rose. This is caused by the power required to provide
the adequate signal quality increases as the number of
users increases because the number of competition
between users in the system increases. Figure 7, 8, and 9
also show that the LZ and ZC method required the largest
of power among other methods, which means that these
methods wasted power. However, this generated a high
SINR value in convergence. Other methods of KG, TB,
Proposed, AG, and DPC require nearly the same power
on average for every scheme. The required transmitting
power is related to the SINR achieved by each method.
Changes in the number of users affected the power
required by the users. Fig. 7, Fig. 8, and Fig. 9 displayed
that as the number of users increased, the power required
by the user to reach convergence increased. This is
caused by the power required to provide the adequate
signal quality increases as the number of user increases.
The comparison of power required by the user on every
scheme can be seen in Table II, whereas the comparison
of the number of iterations is shown in Table III.
Fig. 7. Comparison of power requirement between DPC and PCG on the
three-user scheme
Fig. 8. Comparison of power requirement between DPC and PCG on the
five-user scheme
Fig. 9. Comparison of power requirement between DPC and PCG on the
ten-user scheme
Tables I and II showed that the addition of the number
of users affected the user SINR and the power required
by the user to reach convergence. Table I shows that the
Proposed PCG results still reached the SINR target (5 dB).
Even though the values decreased as the number of users
125©2019 Journal of Communications
Journal of Communications Vol. 14, No. 2, February 2019
increased, the SINR values achieved still larger than the
SINR target, with the number in the three-, five-, and ten-
user schemes were 5.4999 dB, 5.4977 dB, and 5.49783
dB, respectively. Meanwhile, Table II displays that in all
methods, the user power increased as the number of users
increased. It is caused by the increased number of users
created more competition when sharing the same channel.
As the number of user increases, the SINR decreases and
the power needed to maintain the quality increases.
TABLE II: COMPARISON OF USER POWER
Comparison of simulation time for the three user
schemes is presented in Table III. The simulation time
was influenced by the addition of the number of users. As
the number of users increased, the time required to reach
convergence also increased. The rate of convergence is
represented by the number of iterations required by the
user to convergent.
TABLE III: COMPARISON OF THE AVERAGE OF ITERATION TIME
Method Average of Iteration Time (second)
Three-user Five-user Ten-user
AG 0.0409 0.0721 0.1799
KG 0.0355 0.0780 0.1873 TB 0.0384 0.0884 0.2326
DPC 0.0311 0.0725 0.1758
Proposed 0.0388 0.0810 0.2123 ZC 0.0520 0.1113 0.2448
LZ 0.0536 0.1169 0.2017
V. CONCLUSIONS
The results show that the addition of the number of
users affected the consumed power. As the number of
users increased, the power consumption also increased.
Similarly, the number of iterations required to achieve
convergence also increased as the number of users
increased. In contrast, the addition of the number of users
decreased the achieved SINR even though the decrease
was insignificant because the addition of the user could
still be accommodated by the system. In this case, the
system was still feasible.
ACKNOWLEDGMENT
The authors wish to thank to Institut Teknologi
Telkom Purwokerto.
REFERENCES
[1] V. Chandrasekhar, J. G. Andrews, and A. Gatherer,
“Femtocell networks: A survey,” IEEE Commun. Mag.,
vol. 46, no. 9, pp. 59–67, 2008.
[2] X. Li, L. Qian, and D. Kataria, “Downlink power control
in co-channel macrocell femtocell overlay,” in Proc. 43rd
Annual Conference on Information Sciences and Systems,
2009, pp. 383–388.
[3] I. W. Mustika, K. Yamamoto, H. Murata, and S. Yoshida,
“Potential game approach for self-organized interference
management in closed access femtocell networks,” in
Proc. IEEE 73rd Vehicular Technology Conference (VTC
Spring) , 2011, pp. 1–5.
[4] Z. Li, Z. Jiang, Y. Wang, and D. Yang, “A Modified
Power Control Scheme in OFDMA Uplink,” in Prco. 4th
International Conference on Wireless Communications,
Networking and Mobile Computing, 2008, pp. 1–5.
[5] X. Qin, B. Guo, Z. Wang, and X. Yan, “Power control
based on game theory in cognitive radio,” in Proc. 6th
International Congress on Image and Signal Processing,
2013, vol. 2, pp. 1169–1173.
[6] B. Wang, Y. Wu, and K. J. R. Liu, “Game theory for
cognitive radio networks: An overview,” Comput. Netw.,
vol. 54, no. 14, pp. 2537–2561, Oct. 2010.
[7] F. Li, X. Tan, and L. Wang, “A new game algorithm for
power control in cognitive radio networks,” IEEE Trans.
Veh. Technol., vol. 60, no. 9, pp. 4384–4391, Nov. 2011.
[8] J. Jiao, L. Jiang, and C. He, “A novel game theoretic
utility function for power control in cognitive radio
networks,” presented at the Fifth International Conference
on Computational and Information Sciences (ICCIS),
2013, 2013, pp. 1553–1557.
[9] Y. A. Al-Gumaei, K. A. Noordin, A. W. Reza, and K.
Dimyati, “A new power control game in two-tier
femtocell networks,” in Proc. 1st International
Conference on Telematics and Future Generation
Networks (TAFGEN), 2015, pp. 131–135.
[10] A. Al Talabani, A. Nallanathan, and H. X. Nguyen, “A
novel chaos based cost function for power control of
cognitive radio networks,” IEEE Commun. Lett., vol. 19,
no. 4, pp. 657–660, Apr. 2015.
[11] Z. Chenglin and G. Yan, “A novel distributed power
control algorithm based on game theory,” presented at the
5th International Conference on Wireless
Communications, Networking and Mobile Computing,
Beijing, China, 2009.
[12] L. Zhang, S. Zhang, L. Wu, Y. Liu, and C. Zhao, “A Nash
Game Algorithm for Distributive Power Control with
Faster Convergence in Cognitive Radio,” presented at the
Communications and Information Technology, 2009.
ISCIT 2009. 9th International Symposium on, 2009.
[13] A. F. Isnawati, R. Hidayat, S. Sulistyo, and I. W. Mustika,
“A novel utility function of power control game in multi-
channel cognitive femtocell network,” Int. J. Commun.
Netw. Inf. Secur. IJCNIS, vol. 9, no. 3, 2017.
[14] S. A. Grandhi and J. Zander, “Constrained power control
in cellular radio systems,” in Proc. 44th Vehicular
Technology Conference, 1994, pp. 824–828.
[15] S. Koskie and Z. Gajic, “A nash game algorithm for SIR-
based power control in 3G wireless CDMA networks,”
IEEEACM Trans. Netw., vol. 13, no. 5, pp. 1017–1026,
Oct. 2005.
Method User Power (mW)
Three-user Five-user Ten-user
AG 3.9030 62.1425 191.199
KG 4.0015 64.2088 198.204 TB 3.9012 62.6131 193.350
DPC 4.0016 64.2295 198.401 Proposed 4.4035 70.6859 218.221
ZC 24.0725 145.4974 340.408
LZ 63.6121 257.1110 447.761
126©2019 Journal of Communications
Journal of Communications Vol. 14, No. 2, February 2019
Anggun Fitrian Isnawati was born in
Cilacap, Indonesia, in 1978. She
received Doctoral Degree in
Telecommunication Engineering from
Universitas Gadjah Mada, Yogyakarta,
Indonesia in 2018. She joined as a
Lecturer in the Institut Teknologi
Telkom Purwokerto since 2002. Her
research interests include wireless communication, cellular
network, optical communication, MIMO, game theory and
cognitive radio.
Wahyu Pamungkas was born in
Purwokerto, Indonesia, in 1978. He
obtained a Bachelor Degree in Electrical
Engineering, Universitas Gadjah Mada,
Indonesia in 2002. He received his
Master Degree from Telkom University,
Bandung, Indonesia in 2010. By now, he
pursue Doctoral Program at Electrical
Engineering, Institut Teknologi Sepuluh Nopember Surabaya.
He joined Institut Teknologi Telkom Purwokerto since 2002,
and become senior lecturer. His research interest include
Wireless Communications, Cellular Network, Vehicle
Communications and Satellite Communications.
Jans Hendry received the Bachelor and
Master Degree in Electronics and Signal
System, Universitas Gadjah Mada,
Indonesia. In 2007, he started working as
A/V system researcher in Samsung
Electronics Indonesia to develop various
DVD-Player Models. He joined Institut
Teknologi Telkom Purwokerto since
2017. His interests are Electronics, Control, Instrumentation,
Optimization, Machine Learning, and Signal Processing.
127©2019 Journal of Communications
Journal of Communications Vol. 14, No. 2, February 2019