Vous êtes sur la page 1sur 5

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts

for publication in the WCNC 2010 proceedings.

Interference Avoidance and Adaptive Fraction


Frequency Reuse in a Hierarchical Cell Structure
AN Ruihong, ZHANG Xin, CAO Gen, ZHENG Ruiming and SANG Lin
Wireless Theories and Technologies Lab (WT&T), Beijing University of Posts and Telecommunications
School of Information and Communication Engineering, Beijing University of Posts and Telecommunications
Beijing, P.R. China, 100876
anruihongbupt@gmail.com
AbstractDue to high data rate requirement for indoor radio
coverage, femtocells have been proposed as a potential good
solution in recent years. Under this configuration, a main design
objective is to coordinate the mutual interference between macro
and femto cells. In this paper, the flexible assignment of downlink
resource in a hierarchical cell structure is investigated. A
dynamic frequency assignment technique called adaptive fraction
frequency reuse (AFFR) is proposed, by avoidance of severe
interference between macro and femto cells. A network model
and system level simulations for LTE macrocell/femtocell
scenarios are presented. The results show that, AFFR has better
average and cell-edge performance compared with co-channel
frequency allocation (co-channel FA) and orthogonal frequency
allocation (OFA) for both macro and femto cells.
Keywords- femtocells; adaptive fraction frequency reuse; cochannel interference; frequency allocation

I.

INTRODUCTION

It is reported that the majority of data services and phone


calls will take place in indoor environment. Therefore, indoor
coverage will be necessary. One solution to improve indoor
coverage is the so-called home base stations [1]. Introduction
of femtocells can enable indoor connectivity through existing
broadband Internet connections. The deployment of femtocells
has a few benefits such as improved indoor coverage, reduced
bandwidth load in the macrocell networks, reduced indoor call
costs and savings of phone battery [2].
Special attention must be paid to interference mitigation
when femtocells are deployed [3], because the mutual
interference can not be handled simply by network planning.
The deployment ratio and positions of the femtocells are
unpredictable, so that it is difficult for operators to make
appropriate coordinations. The co-channel interference
between femtocells and macrocells has been widely studied
[3]-[7]. A conventional mechanism to operate femtocell
systems is that femtocell base stations use full frequency bands
of macro networks, which called co-channel frequency
allocation (co-channel FA) [3]. But this method can increase
cross-tier interference. In [4], another way called orthogonal
frequency allocation (OFA) is proposed to control co-channel
interference. However, it will drive the operators to a reduced
spectrum utility, which is extremely expensive and undesirable.

Therefore, co-channel deployment seems to be more


appropriate, but technically more challenging.
Using OFDMA as a multiple-access technique, a good
alternative to control co-channel interference between femto
and macro cells in LTE system is dynamic assignment of
physical resource blocks (PRBs). Femto users and macro users
who are interfering with each other will be allocated different
PRBs. In this paper, a new dynamic frequency assignment
technique is introduced, considering interference and coverage
issues, which called adaptive fraction frequency reuse (AFFR).
This AFFR scheme makes maximum usage of the resources,
and meanwhile reduces the interference brought by femtocells.
Ref. [5] and [7] only concern about the effects from femtocells
to macrocells, while in this paper, the performance of femtocell
users is also discussed. Simulation results show that AFFR has
better average and cell-edge performance for both macrocells
and femtocells compared with co-channel FA scheme and OFA
scheme.
The rest of the paper is organized as follows: In section II, a
system model of two layer deployment is introduced, and the
main design objective is expressed with formula. Section III
introduces the existing frequency allocation schemes, such as
co-channel FA and OFA, and then presents an AFFR algorithm
as dynamic interference avoidance technique for hybrid
network scenarios. In section IV, system level simulation
results show the increase in capacity and coverage when using
AFFR compared with co-channel FA and OFA in this
hierarchical cell structure. Finally, in the last section some
conclusions are drawn.
II.

SYSTEM MODEL AND PROBLEM FORMULATION

A. System Model and Assumptions


The setup consists of a network of seven cells with
wraparound, each of which has L femtocells. The performance
of the central base station (BS), called as BS0, is evaluated.
Only one femtocell UE (FUE) per femtocell BS is assumed for
simplicity. The number of users per macrocell BS is M, thus,
the central cell has M macrocell users (MUEs) and L FUEs.
Dense femto cell deployment modeling [8] is used in our
model. In each macrocell, a number of femtocell blocks are
dropped randomly.

978-1-4244-6398-5/10/$26.00 2010 IEEE

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the WCNC 2010 proceedings.

where Ptx and Ptx denote the transmit power of macrocell BS


and femtocell BS on PRB n, respectively. PLm represents the

pathloss between MUE m and its serving BS, while PLml

denotes the pathloss between MUE m and femtocell BS l. N 0


is the additive white Gaussian noise plus inter-cell interference.
If PRB n is used by femtocell BS l, x nl = 1 , else x nl = 0 .
One FUE per femtocell is assumed, and the received SINR
of FUE in femtocell BS l on PRB n can be expressed as
Figure 1. A femtocell block

In a dense-urban femtocell modeling, each block represents


two stripes of apartments, each of which has 2 by N
apartments ( N is 10 in our model). Each apartment is of size
10m 10m . There is a street between the two stripes, with
width of 10m, see Fig. 1. Each femtocell block is of size
10( N + 2)m 70m . This is to make sure that the femtocell BSs
from different femtocell blocks are not too close to each other.
Each femtocell block has six floors [8].

The femtocell BS is also randomly placed in each


femtocell. All of the FUEs are located in the femtocell
apartment, which are dropped randomly in the active
femtocells. MUE are dropped randomly throughout the cell. It
is possible that some MUEs will be dropped into the femtocell
blocks.
A number of different deployment configurations have
been considered for femtocell BSs [9]. In this paper, closed
subscriber group (CSG) mode is investigated. If femtocell BSs
are configured with CSG mode, when a MUE approaches a
femtocell BS and the signal from the femtocell BS is larger
than that from macrocell BS, this MUE still can not hand over
to that femtocell BS. For this reason, the interference from
femtocell BS may cause dead zones in macrocell, as
handover can not take place.
The following assumptions are employed for ensuring the
analysis:
Perfect knowledge of users channels
Exact knowledge of interference source
Information exchange is supported between macrocell
BSs and femtocell BSs
B. Problem Formulation
As stated, the main objective is to coordinate the mutual
interference between macro and femto cells. To formulate the
frequency allocation problem, consider an OFDMA system
with N PRBs, and assume one femtocell block per each
macrocell for simplicity. Use the definitions in Section A, the
received signal to interference and noise ratio (SINR) of MUE
m on PRB n can be expressed as

SINRm (n) =

Ptx PLm
L

tx

l =1

x nl + N 0
PLml

(1)

SINRl (n) =

Ptx PLl
L

tx

PLlk x nk + Ptx PLl y n + 0

(2)

k =1,k l

where PLl represents the pathloss between FUE in femtocell l


and its serving femtocell BS, while PLlk denotes the pathloss
between FUE in femtocell l and femtocell BS k. PLl
represents the pathloss between FUE in femtocell l and
macrocell BS0. N 0 is the additive white Gaussian noise plus
inter-cell interference. If PRB n is used by BS0, y n = 1 , else
yn = 0 .
After SINR is calculated, user throughput can be obtained
by Shannon theorem and can be calculated in terms of PRBs.
Using (1) and (2), one user can get its received SINR on PRB n
if PRB n is allocated to this user. The throughput of this user
can be expressed as follows:
N

Cuser =

log 2 (1 + SINRn ) z n ,

(3)

n =1

where B0 is the bandwidth of one PRB. If PRB n is assigned


to this user, z n = 1 , else z n = 0 .
The throughput of femtocell/macrocell BS will be the sum
of its serving UEs.
The main idea of the proposed algorithm is to mitigate the
mutual interference between macro and femto cells, and then
improve the received SINR of both MUEs and FUEs, so that
the throughput of both systems can be promoted.
EXISTING AND PROPOSED FREQUENCY ALLOCATION
SCHEME
This section presents two existing downlink frequency
allocation schemes between macrocells and femtocells in the
hierarchical cell structure, and then introduces the proposed
adaptive fractional frequency reuse (AFFR) scheme. 10MHz
bandwidth (BW) for both macrocells and femtocells and
another 10MHz BW for macrocell only is assumed. Only the
performance of the first 10MHz BW is investigated, and
femtocells have priority over macrocells to occupy the PRBs
III.

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the WCNC 2010 proceedings.

on the first 10MHz BW. 10MHz BW means 50PRBs is


available.
A. Existing Frequency Allocation Scheme
Fig. 2 shows a co-channel FA scheme. Full frequency
bandwidth is allocated to both macrocells and femtocells.
Since the co-channel FA scheme shares all frequency band
between macro and femto cells, the mutual interferences is
very critical.
In order to avoid the serious co-channel interference, OFA
scheme is proposed [4]. is defined as the frequency
allocation fraction for macrocell BS. That is to say,

Figure 2. Co-channel frequency allocation

Macro cells bandwidth


.
Total system bandwidth

Macrocell BSs use a part of frequency band of the total


bandwidth, while femtocell BSs can only use remaining
frequency band. Fig. 3 shows an OFA scheme with = 0.8 .
Although orthogonal FA scheme can avoid the co-channel
interferences between macro and femto cells, both macrocell
BSs and femtocell BSs can only use part of the full frequency
band, which lead to waste of frequency resources.
B. Proposed Frequency Allocation Scheme
Co-channel FA scheme will create serious mutual
interference while OFA scheme will reduce the proportion of
resources used by both macrocells and femtocells. For these
reasons, AFFR is proposed. The main idea of the proposed
scheme is to suppress the downlink interference from femtocell
BSs to both MUEs and FUEs. In order to maintain the
performance of MUEs, all 50 PRBs can be used at macro
Layer. AFFR runs at each time slot and the algorithm is
divided into three steps.
1) Share the available spectrum on a non-overlapping
basis between neighboring femtocells
Using the conception of interference graph [11], when a
femtocell BS is powered on, its FUE measures the pathloss
from its serving femtocell BS and the surrounding femtocell
BSs [12]. Based on these measurements, this femtocell BS
determines the set of femtocell BSs that are likely to have an
interfering relationship with it by checking whether the
pathloss difference from this femtocell BS and its neighboring
femtocell BSs to its FUE is above a certain threshold PLTH. The
interfering set of femtocell BS l can be expressed as:

I l = { femtocell BS k | PLlk PLl > PLTH }, k = 1, 2, ... , L .(4)


Then, using a cyclic iterative algorithm, the femtocell BS
will be assigned those PRBs (or subbands) that are nonoverlapping with femtocell BSs in its interfering set [10]. That
means, on each loop, every femtocell BS chooses one PRB
from 50 PRBs randomly except the PRBs this femtocell BS
and femtocell BSs in its interfering set have already used. At
the end of each loop, the scheduler makes a judgment to decide

Figure 3. Orthogonal frequency allocation ( = 0.8).

whether new allocation is done during this round. If there is no


PRB that has been allocated during this loop, the iterative
algorithm is over.
After this step, the resources assigned to femtocell BSs that
are interfered with each other will be orthogonal.
2) Resource reuse between femtocell BSs
In this step, one femtocell BS seeks all the remaining PRBs
to find whether there are any other PRBs that can be used
based on SINR requirement. Assume that there is only one
FUE associated with a femtocell BS, where the FUE and
femtocell BS share the same index for notation simplicity.
In order to decide whether PRB n can be used by femtocell
BS l, firstly, the SINR of FUE l on PRB n( SINRl (n) ) is
calculated supposing PRB n is used by femtocell BS l. If
SINRl (n) is above a certain threshold SINR1TH :
SINRl (n) SINR1TH ,

(5)

then femtocell BS l decides that it can use PRB n, and sends


reuse request to femtocell BSs in I l . Then the FUE served by
femtocell BS k in I l calculates its received SINR on PRB n
assuming PRB n is allocated to femtocell BS l. If SINRk (n) is
above a certain threshold SINR2TH :
SINRk (n) SINR2TH , HeNBk I l ,

(6)

the femtocell BS k sends reuse permission back to femtocell


BS l, otherwise, femtocell BS k sends reuse rejection. If all
the received signals are reuse permission, femtocell BS l
decides that PRB n is assigned to femtocell BS l.

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the WCNC 2010 proceedings.

3) Maximum SINR scheduler at macrocell BSs


In this step, one MUE measures its received SINR on all
PRBs and reports this information to its main serving BS. First,
a macrocell BS selects one MUE randomly, and it means that
all MUEs get resources with equal probability. Then once
MUE m is granted, its main serving BS selects the PRB on
which its SINR is max.
n * = arg max ( SINRm (n))

TABLE II.
Frequency
Allocation Scheme

User Throughput(Mbps)
Average user throughput

10% user throughput

Co-channel FA

1.31

0.076

OFA

1.15

0.063

AFFR

1.42

0.101

(7)

n=1,... N

MACRO USER THROUGHPUT

1
0.9

If the received SINR on PRB n for MUE m is above


SINR3TH , that is

0.8

Co-channel FA
OFA
AFFR

0.7

SINRm (n ) >

SINR3TH

(8)

CDF

0.6
*

0.5
0.4

macrocell BS decides that PRB n can be used by MUE m, and


can not be allocated to any other MUEs in its serving set.
Macrocell BSs continue doing this algorithm until there is no
PRB available.
This step can avoid MUEs be assigned to the PRBs where
the interference brought by femtocell BSs is severe. The
received SINR of MUEs can be improved and DL throughput
of macro layer can be promoted.
IV.

SIMULATION RESULTS

The considered OFDMA system consists of seven base


stations with N = 50 PRBs and M = 10 users. The number of
femtocells in each cell is assumed to be 50,100,150 and 200.
The detailed parameters are presented in Table I.
In the urban deployment, channel model in Table 7 in [10]
is used, with minimum distance between UE and femto/macro
BS set to be 1m.
A. Throughput and SINR of Different Algorithms
In this section, throughput and SINR of three different
algorithms described in section III are compared. Fig. 4 shows
TABLE I.

SIMULATION PARAMETERS

Parameter
Carrier frequency

Value
2000 MHz

Macrocell BS antenna gain

14 dBi

Femtocell BS antenna gain

0 dBi

UE noise figure

9 dB

Maximum femtocell BS TX power

20 dBm

Maximum macrocell BS TX power

46 dBm

0.8
TH

PL

SINR1TH / SINR2TH / SINR3TH

30dB
2dB/2dB/0dB

0.3
0.2
0.1
0
-40

-30

-20

-10
0
10
20
Femtocell User SINR(dB)

30

40

Figure 4. Femtocell user SINR CDF of three algorithms

the SINR CDF for the downlink of three schemes with 200
femtocells per macrocell. It is shown that the SINR
improvement of FUEs as a result of OFA scheme is only
marginal, that is because OFA can not handle the interference
caused by other femtocell BSs. However, when AFFR scheme
is used, interference from neighboring femtocell BSs can be
limited, so AFFR scheme has the best FUE SINR, especially
for the cell-edge users.
Table II shows the average macro user throughput of three
schemes when there are 200 femtocells per macrocell. The
average throughput of the MUEs using OFA scheme is the
smallest, and the proposed AFFR scheme shows the best
performance. The average MUE throughput of AFFR scheme
is 1.42Mbps, and shows improved performance by 8.4% than
the 1.31Mbps of co-channel FA scheme. The average
throughput performance of OFA is worst, which is 1.15Mbps.
That is due to the reduced spectrum utility.
Fig. 5 shows the simulation results of the throughput for
FUEs with 200 femtocells per cell. From Fig. 5, we can see
that the throughput of FUEs using OFA scheme is significantly
lower than that using co-channel FA and AFFR. That is
because when OFA is used, the available bandwidth of each
femtocell is only 20% of the total bandwidth. It is also shown
that data rate coverage of AFFR is better than that of OFA and
co-channel FA.
B. Throughput of Different Femtocell Deployment Density
Fig. 6 and Fig. 7 show the average cell throughput of
femtocells and macrocells vs. different femtocell deployment
density. The result of Fig. 6 shows that when femtocell density
is very low, such as 50 or 100 femtocells per one macrocell, the

Average Throughput of Macrocell(Mbps)

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the WCNC 2010 proceedings.

Co-channel FA
OFA
AFFR

0.9
0.8
0.7
CDF

0.6
0.5
0.4
0.3
0.2
0.1
0 2
10

15
14
13

Co-channel FA
OFA
AFFR

12
11

10
50
3

10
10
Femtocell User Throughput(Kbps)

10

75
100
125
150
175
Number of Femtocells per Macrocell

200

Figure 7. Average cell throughput of macrocell BSs

Figure 5. Femto user throughput CDF of three algorithms

Femtocell BS Throughput(Mbps)

30
Co-channel FA
OFA
AFFR

25

average throughput of macro/femto cells decreases as the


increasing of femtocell number. However, for all deployment
density, AFFR shows the best performance for macro layers.
To sum up, AFFR scheme offers data rate enhancement as well
as high spectrum utility in a hierarchical cell structure.

20

REFERENCES
15

[1]
[2]

10
5
0
50

75
100
125
150
175
Number of Femtocells per Macrocell

200

Figure 6. Average cell throughput of femtocell BSs

average throughput of femtocells using AFFR is higher than


using co-channel FA. OFA scheme has the worst average cell
throughput.
Fig. 7 shows the average throughput of macrocells vs.
different femtocell density. The average throughput of
macrocells decreases as the increasing of femtocells number.
Similar to Fig. 6, the results of Fig. 7 shows that AFFR has the
best macrocell performance, while OFA is the worst.
V.

CONCLUSION

In this paper, the flexibility of the co-channel deployment


of femtocells in an existing macrocell network is investigated.
Two existing methods are introduced: orthogonal and cochannel FA schemes. In addition, an adaptive fractional
frequency reuse in a hierarchical cell structure is proposed and
system level simulation results are analyzed. The simulation
results show that AFFR has better coverage and average cell
throughput compared with the co-channel FA scheme as well
as OFA scheme due to appropriate control of downlink
interference from femtocell BSs. Furthermore, the average cell
throughput for both macrocells and femtocells in different
femtocell deployment density is presented. It is shown that the

Femto forum. http://www.femtoforum.org.


V. Chandrasekhar and J. G. Andrews, Femtocell networks a survey,
IEEE Communications Magazine, vol. 46, pp. 5967, September 2008.
[3] H. Claussen, Performance of macro- and co-channel femtocells in a
hierarchical cell structure, IEEE 18th International Symposium on
Personal, Indoor and Mobile Radio Communications (PIMRC 2007), pp.
15, Athens, Greece, September 2007.
[4] K. Cho, W. Lee, D. Yoon, K. Hyun, and Yun-Sung Choi, Resource
alloation for orthogonal and co-channel femtocells in a hierarchical cell
structure, 13th IEEE International Symposium on Consumer
Electronics (ISCE2009), pp. 655-656, 25-28 May 2009.
[5] L. T. W. Ho and H. Claussen, Effects of user-deployed, cochannel
femtocells on the call drop probability in a residential scenario, IEEE
18th International Symposium on Personal, Indoor and Mobile Radio
Communications (PIMRC 2007), pp. 15, Athens, Greece, September
2007.
[6] I. Guvenc, M. R. Jeong, F. Watanabe, and H. Inamura, A hybrid
frequency assignment for femtocells and coverage area analysis for cochannel operation, IEEE Communications Letters, pp. 880882,
December 2008.
[7] D. Lpez-Prez, G. De La Roche, A. Valcarce, A. Juttner and J. Zhang,
Interference avoidance and dynamic frequency planning for WiMAX
femtocells networks, Communication Systems, 2008( ICCS 2008), 11th
IEEE Singapore International Conference, pp. 15791584, 19-21
November 2008.
[8] Femtoforum WG2 OFDMA Interference Study OFDMA Interference
Study: Evaluation Methodology Document, March 2009.
[9] 3GPP TR 25.967, v9.0.0, Home Node B Radio Frequency (RF)
Requirements (FDD), June 2009.
[10] R4-092872,Downlink interference coordination between HeNBs,
CMCC, August 2009.
[11] M. C. Necker, A Graph-Based Scheme for Distributed Interference
Coordination in Cellular OFDMA Networks, VTC Spring 2008. IEEE,
pp. 713 718, 11-14 May 2008.
[12] D. Lpez-Prez, A. Valcarce, G. De La Roche and J. Zhang, OFDMA
femtocells: A roadmap on interference avoidance, IEEE
Communications Magazine, vol. 47, no 9, pp. 41 - 48, September 2009.

Vous aimerez peut-être aussi