Inter-Cell Interference Management in LTE-A Small-Cell Networks

Yong-Ping Zhang*, Shulan Feng*, Philipp Zhang, Liang Xia, Yu-Chun Wu*, and Xiaotao Ren *Research Department of Hisilicon, Huawei Technologies Co., Ltd, Beijing, P. R. China. Research Department of Hisilicon, Huawei Technologies Co., Ltd, Plano, Texas, USA.

Research Department of Wireless Network, Huawei Technologies Co., Ltd, Beijing, P. R. China. Emails: {yongping.zhang, shulan.feng, pzhang, xialiang, wuyuchun, renxiaotao}@huawei.com

AbstractThis paper is concerned with inter-cell interference

in co-channel Long-Term Evolution-Advanced (LTE-A) small-cell networks (SCNs). A practical SCNs architecture with small-cell group muting (SCGM) is proposed to mitigate the interference between the hosting macro-cell and small-cells as well as interference between adjacent small-cells. Lower power consumption can also be achieved in the proposed architecture. Our system level simulation results show that the proposed solution outperforms todays default operational methods in terms of both macro-cell average and cell-edge throughput.

KeywordsSmall-cell networks (SCNs), small-cell group muting (SCGM), interference management.

I. INTRODUCTION Owing to the high demand for broad services and new

applications, todays wireless networks are facing the challenge of supporting exponentially increasing data traffic [1]. Recent efforts geared to small-cell networks (SCNs) promise to have great practical value because they are capable of offering the jack of both trades. By a very dense deployment of low-cost, low-power base stations, both the spatial reuse of radio resource and transmit power efficiency can be potentially improved. It is envisioned that the next generation wireless networks will consist of macro-cells and a high density of small-cells with different capabilities including transmit power and coverage range [2].

With full frequency reuse and an increasing density of small-cell deployments, the interference between macro-cells and small-cells as well as interference between adjacent small-cells is always a serious concern in SCNs [2]. The widely-used inter-cell interference mitigation technique in homogeneous networks is soft frequency reuse (SFR) [3]. By dynamic spectrum access, SFR can effectively minimize interference. However, as applying SFR in SCNs directly, the bandwidth assignment is facing a tough challenge due to the irregular distribution of a huge number of small-cells in SCNs.

The network-centric enhanced inter-cell interference coordination (eICIC) technique with time domain muting has been intensively studied in the 3rd Generation Partnership Project (3GPP) community [4]. And it is now part of LTE-A. However based on the premise of low density of small-cells, the interference between adjacent small-cells is not considered and only the interference from macro-cells to small-cells is therefore partly mitigated in eICIC.

Other extensively-investigated inter-cell interference mitigation techniques are downlink coordinated multi-point

(CoMP) transmission [5]-[7] and interference alignment (IA) [8]. The former aims to mitigate inter-cell interference by coordinating the transmission of several geographically separated cells. With synchronized cells, inter-cell interference in SCNs can be mitigated and even exploited by applying multi-user based joint transmission techniques [6]. In IA, by restricting all inter-cell interference at each receiver into a small space, a higher degree-of-freedom (DoF) than conventional methods can be achieved. Since both CoMP and IA require reliable channel state information, significant message exchange between the cooperative cells, and high computational complexity, their benefits for practical SCNs are still unclear. Overall, how to efficiently and practically mitigate inter-cell interference in SCNs with a very dense deployment of small-cells is still an open issue.

In this paper, we propose a practical SCNs architecture with small-cell group muting (SCGM). Specifically, the proposed SCNs architecture can exclude the interference between macro-cells and small-cells and achieve lower power consumption. Furthermore, a SCGM scheme implemented in the above SCNs architecture can mitigate inter-cell interference between adjacent small-cells. Using system level simulation results, we demonstrate the validity of the proposed SCNs architecture and SCGM method.

The rest of this paper is organized as follows. System model and the interference problem in SCNs are described in Section II. The proposed architecture and SCGM are discussed in detail in Section III. Section IV presents numerical results. We finally give concluding remarks in Section V.

II. SYSTEM MODEL AND PROBLEM FORMULATION Consider a cellular system with N=19 sites where each site

has three macro-cells, M small-cells randomly distributed within macro-cell area. The maximum transmit power of macro-cell (denoted by P1) and small-cell (denoted by P2) are 46dBm (i.e., 40w) and 30dBm (i.e., 1w) respectively.

A fundamental issue in designing SCNs is the following: what is the appropriate density to offer a good tradeoff between the ratio of UEs offloaded to small-cells and total power consumption. In this paper, the density of small-cells M is given by

1

2

PMP

=

(1)

978-1-4673-6337-2/13/$31.00 2013 IEEE

where denotes the floor function and is pre-defined and used to adjust the tradeoff between the ratio of offloaded UEs and total power consumption. Clearly, the latter is decided by P1, P2, N and M. Actually, P1, P2 and N are explicit here. We can choose the proper value of M to restrict the total power consumption within a certain range by adjusting . In what follows, we show how to estimate the ratio of offloaded UEs when is given. Take = 2 as an example. Recall that P1 = 46dBm and P2 = 30dBm. From (1), we obtain M = 20, i.e., 20 small-cells randomly distributed in macro-cell area. Then the ratio of offloaded UEs can be obtained by the Monte Carlo method. In particular, we first randomly place plenty of UEs, e.g., two hundred thousand UEs, in the networks. Then serving cell selection based on measurements of reference signal received power (RSRP) is performed. Note that an offset of 10log (i.e. 3dB) is applied to the RSRP measured from small-cells. After gathering the number of UEs served by small-cells, the ratio of offloaded UEs we can obtain is as high as 76.96%. This offloading ratio implies that the encouraging performance can be achieved in SCNs even without eICIC [9]. Note that the assumption of M = 20 will be adopted in the rest of this paper unless otherwise stated.

Next, we examine the effect of inter-cell interference on the performance of SCNs. We assume that only one UE can be scheduled by each cell simultaneously. The sum spectrum efficiency in one macro-cell (denoted by r) is then given by

1

,1log(1 )

M

m km

r q+

=

= + (2) where

3 ( 1)

, , , 01,

N M

m k m k i ki i m

q p p N+

=

= + denotes the received

signal to interference plus noise ratio (SINR) at UE k, pi,k the received power of UE k from cell i and N0 the additive noise. Note that in (2), we assume UE k is served by cell m. From (2), we can see that, when M increases, both the time-frequency resource reuse factor and inter-cell interference3 ( 1)

,1,

N M

i ki i m

p+

= increase accordingly. And the increasing of the

latter results in the logarithmical reduction of r. Due to the random distribution of small-cells, to theoretically trace the impact of M on r is very difficult. Hence, using system level simulation, we investigate the relationship between M and r in Fig. 1. Note that r in Fig. 1 is normalized by r with M = 0.

From the results, the approximately linear relationship between the normalized r and M can be observed when M is less than 10. It implies that at that time, inter-cell interference is not serious. However when M is larger than 10, inter-cell interference, which is critical to the system performance, prevents the system to achieve the reasonable gain with the increasing density of small-cells. Thus, the ultimate goal of this paper is to investigate how to effectively and robustly mitigate inter-cell interference in SCNs with the high density of small-cells.

Fig. 1. Comparison of normalized spectrum efficiency per macro-cell of SCNs with various numbers of small-cells

III. A PRACTICAL SCNS ARCHITECTURE WITH SCGM In conventional LTE-A SCNs architecture, any UE select

one macro-cell or small-cell as its serving cell according to RSRP. Then both control signaling and data are received by UE from its corresponding serving cell. In this section, we develop a practical SCNs architecture assuming a very dense deployment of small-cells. In this architecture, the interference between macro-cells and small-cells is excluded completely. SCGM is further proposed to mitigate inter-cell interference between adjacent small-cells.

A. A Practical SCNs Architecture The key idea of our proposed architecture is to configure

the transmitting of control channel at macro-cells and data channel at small-cells (Fig. 2 shows an example).

In particular, macro-cells are muted in data channel and their functions are restricted to guaranteeing the coverage, assisting local area radio access, mobility management and so on. All UEs in SCNs first get access into macro-cells and then receive data service from small-cells by the assistance of macro-cells. Note that there is no control or common signaling transmitted by small-cells. The above restrictions are dictated by the following reasons.

1) Inter-cell interference between macro-cells and small-cells is excluded completely in this architecture. In LTE-A, different OFDM symbols are allocated to transmit control channel and data channel. Therefore transmission resource at macro-cells and small-cells is orthogonalized in the time division multiplexing (TDM) manner. Note that this architecture can be extended easily from co-channel scenario to multi-carrier scenario by configuring one control-dedicated component carrier (CC) at macro-cells and multiple data-dedicated CCs at small-cells.

2) Lower power consumption can be achieved compared with the conventional SCNs. The total consumed power in SCNs is calculated as

P = (1-1)P13N + (1-2)P23NM (3) where 1 and 2 denote the muting ratios of macro-cell and small-cell respectively. In LTE-A, at most 4 OFDM symbols out of one subframe, which consists of 14 OFDM symbols, are allocated to transmit control signaling. And the other symbols are allocated to transmit data. Thus, we have 1 = 2 = 0 in the conventional SCNs architecture and 1 = 10/14, 2 =

0 4 5 6 8 10 15 20012

4

6

8

10

12

M

Nor

mal

ized

spec

trum

eff

icie

ncy

Multiplexing gain:11.38 x

4.53 x 5.25 x

5.97 x

7.25 x

8.40 x

Fig. 2. Illustration of the proposed SCNs architecture.

1 - 1 = 4/14 in our proposed architecture. Recall that P1 = 46dBm, P2 = 30dBm and M = 20. The reduction of power consumption is at least1 57.14%. Note that full buffer traffic is assumed in the above estimation. The impact of cells which are powered off due to no active UEs is not taken into account.

3) The high density of small-cells and exclusion of the interference between macro-cells and small-cells guarantee the seamless coverage (i.e., the coverage of control signaling and data). In particular, UEs receive control signaling from macro-cell in our proposed architecture. Thus, the coverage of control signaling can be ensured by the original design of LTE-A homogenous networks. The only issue is whether the coverage of data channel is inconsistent. This issue is analyzed below with the help of the geometry (which is also referred to as wideband SINR). In Fig. 3, we plot the geometry of all UEs when the data channel of macro-cells is on or muted. In the former case, any UE connects to macro-cell or small-cell according to its RSRP. In the latter case, all UEs receive data from small-cells. No interference is received from the muted macro-cells. Without loss of any generality, in our simulation example, UEs are distributed randomly in the entire networks. After macro-cell muting, the 60% worst UEs geometry is improved. It implies that the coverage of data channel in our proposed architecture is enhanced. The loss only happens at the peak geometry. It is reasonable and expectable due to the large transmit power gap between macro-cell and small-cell. It is worthy to note that, the loss of the peak performance does not result in any degradation of the data channel coverage.

B. Small-Cell Group Muting The high density of small-cells guarantee the coverage

continuity, but result in stronger inter-cell interference between adjacent small-cells. To overcome this problem, SCGM is proposed below.

In SCGM, small-cells distributed in the same hosting macro-cell area2 are separated into two groups, i.e., G1 and G2. 1 When 1~3 OFDM symbols are allocated to transmit control channel, the more power consumption reduction can be obtained. 2 Actually, small-cells deployed at the edge of adjacent hosting macro-cells may have the overlapping coverage area and cause interference between each other. However due to the lower transmit power of small-cells, the interference level is therefore low. Moreover in this paper, to relax the requirement on backhaul, SCGM is restricted within macro-cell, i.e., intra-macro-cell SCGM. We leave the design of inter-macro-cell SCGM as the future work.

Fig. 3. The geometry of UEs with different distribution when macro-cells in SCNs are on (the red line) or muted (the blue line)

Fig. 4. Illustration of the muting patterns with various muting ratios. The periodicity of the pattern is 8 subframes (8ms in LTE-A).

Different grouping criterions can be adopted in SCGM, e.g., the number of serving UEs (i.e., the load) and the statistic link quality (i.e., reference signal received quality (RSRQ) in LTE-A).

After grouping, through the backhaul links, two different periodic muting patterns with different muting ratios are sent from the hosting macro-cells to small-cells belonging to G1 and G2 respectively. Fig. 4 shows the muting patterns used in our implementation. Finally, small-cells transmit or mute at the corresponding subframes according to their muting patterns.

Note that the UEs served by small-cells belonging to G1 may experience significantly different interference levels when small-cells belonging to G2 are during normal or muting subframes, and vice versa. In particular, we assume that UE k is served by small-cell m belonging to G1. Let

(1),m kq and

(2),m kq

be the corresponding SINR of UE k while small-cells belonging to G2 are during normal and muting subframes respectively. Then (1),m kq and

(2),m kq can be written as

1 2

,(1),

, , 0,

m km k

i k i ki G i m i G

pq

p p N

=

+ + , (4)

1

,(2),

, 0,

m km k

i ki G i m

pq

p N

=

+ . (5) It is therefore necessary to configure the UEs to report (1),m kq

and (2),m kq separately. The corresponding small-cells should use the appropriate channel state information feedback during link adaptation and scheduling decision.

-5 0 5 10 15 200

0.2

0.4

0.6

0.8

1

Geometry (dB)

CD

F

Macro-cell onMacro-cell muting

IV. SIMULATION ASSUMPTIONS AND RESULTS In this section, we show the throughput performance of the

proposed SCNs architecture with SCGM. The baseline performance is obtained under default assumptions of closed-loop SU-MIMO without range extension (RE) or eICIC. We implement the proposed SCNs architecture with SCGM in a quasi-dynamic system level simulator. This simulator includes explicit modeling of major radio resource management (RRM) algorithms such as packet scheduling, closed-loop MIMO with precoding and rank adaptation, link adaption, hybrid automatic repeat request (HARQ). In this paper, we assume a perfectly synchronized backhaul connection between small-cells and their hosting macro-cells. We leave the effects of synchronization delay as the future work. Table I summarizes the default simulation parameters.

Note that the macro-cell average and 5% quantiles (denoted by 5%-ile below) UE throughput are adopted as performance metrics. In particular, the macro-cell average throughput, which is obtained by averaging the sum throughput of UEs in each macro-cell, indicates the overall performance. And the 5%-ile UE throughput, which is obtained at the 5% point of the cumulative distribution function (CDF) curve, indicates the cell-edge performance.

A. Benefits of the Proposed SCNs Architecture At first, we present the throughput performance for the

proposed SCNs architecture. Fig 5 shows the throughput (marked by the light grey bars) after adopting the proposed SCNs architecture without SCGM. A significant gain of 26.64% is observed at the 5%-ile UE throughput. However without SCGM, the proposed SCNs architecture leads to marginal macro-cell average throughput improvements.

The benefits of the proposed SCNs architecture on inter-cell interference cancellation as well as power efficiency improvements have been analyzed in Section III and therefore we omit the details here.

B. Effects of Various Grouping Criterions and Muting Patterns

Secondly, we show that when SCGM with proper grouping criterion and muting pattern is involved in the proposed SCNs architecture, both the macro-cell average throughput and 5%-ile UE throughput can be further improved.

In our implementation, half small-cells with highest load (Criterion 1, marked by the deep grey bars in Fig. 5) or best link quality (Criterion 2, marked by the black bars in Fig. 5) are grouped as G1 and the others are grouped as G2. For the sake of simplicity, small-cells belonging to G1 are configured by pattern 0 (i.e., without muting) while small-cells belonging to G2 are configured by one of the other patterns.

By applying Criterion 2, only with pattern 1 and 2, the additional gains on both the macro-cell average throughput and 5%-ile UE throughput can be achieved. As the muting ratio increases, the macro-cell average throughput gain increases while 5%-ile UE throughput gain decreases. This phenomenon can be explained by the Matthew effect. In particular, after applying Criterion 2, the UEs in G1 are with

TABLE 1: SIMULATION ASSUMPTIONS Network

Carrier frequency 2 GHz

Cell layout 19 sites, 57 macro-cells, 500m macro-cells inter-site distance, wrap-around

Test Scenarios Macro-cell: ITU UMa Pico-cell: ITU UMi UE speed: 3 km/hr

Duplex mode FDD System bandwidth 10 MHz

Transmit power Macro-cell P1: 46dBm (i.e., 40w) Pico-cell P2: 30dBm (i.e., 1w)

Antenna configuration

Macro-cell: 2 antennas, 3D [10] Small-cell: 2 antennas, Omni UE: 2 antennas, Omni

Traffic model Traffic model Full buffer UE number 40 UEs/macro-cell

UE placement

30 UEs inside the hotspots (the area within 40m radius of each small-cell); 10 UEs are uniformly and randomly distributed within the macro-cell area. Feedback

CSI feedback delay 4ms CSI feedback period 5ms

Scheduler Scheduler Proportional fair (PF) UE receiver Interference unaware MMSE HARQ CC, Maximum 3 transmission

(a)

(b) Fig. 5. The macro-cell average (Fig. (a)) and 5%-ile UE (Fig. (b)) throughput comparison of baseline and our proposed method.

Pattern 1 Pattern 2 Pattern 3 Pattern 40

50

100

150174.35

250

300

Muting pattern

Mac

ro-c

ell a

vera

ge th

roug

hput

(Mbi

t/s)

BaselineMacro-cell MutingMacro-cell Muting + SCGM, Criterion 1Macro-cell Muting + SCGM, Criterion 2 15.27%

15.78%14.46%

0.37%7.59%7.54%

17.44%

12.17% 12.75%

Pattern 1 Pattern 2 Pattern 3 Pattern 40

0.25

0.5

0.75

0.94

1.25

1.5

1.75

2

Muting pattern

5%-il

e U

E th

roug

hput

(Mbi

t/s)

BaselineMacro-cell MutingMacro-cell Muting + SCGM, Criterion 1Macro-cell Muting + SCGM, Criterion 2

0.53%

38.50%

13.48%

63.96%57.01%

37.22% 29.09%26.64%

55.94%

relatively better channel condition. As the muting ratio of G2 increases, the throughput performance of the UEs in G1 is further enhanced at the cost of that of the UEs in G2.

From Fig. 5, we find that by applying Criterion 1, SCGM can offer the additional macro-cell average (7%~15%) and 5%-ile UE (12%~37%) throughput gains to the proposed SCNs architecture. Muting pattern 2, which maximizes the 5%-ile UE throughput, is further selected to be the optimal muting pattern. Note that the presented simulation results are obtained by adopting Criterion 1 and muting pattern 2 from now on unless otherwise stated.

C. Comparison with eICIC in LTE-A So far, we have determined the optimal grouping criterion

and muting pattern for SCGM. By using these optimal configurations, we make a performance comparison between our proposed method and eICIC with RE in LTE-A.

As pointed out by Strzyz et. al [11], the higher RE offset will damage the control channel reliability in LTE-A heterogeneous networks. Hence in simulation, the RE offset is set to 3 dB, which is recommended in [11] when there is no eICIC in control channel. Note that in eICIC, we assume that UEs can perfectly remove the interference from the cell-specific reference signal in almost blank subframe (ABS) transmitted by macro-cells.

From Fig. 6, as expected, we find that eICIC in LTE-A can hardly cancel inter-cell interference in SCNs with a very dense deployment of small-cells efficiently. When the muting ratio is set to 50%, eICIC has similar throughput performance as the baseline. And the macro-cell average throughput gain is only 1.95%, which is much lower than that of our proposed method.

When the muting ratio in eICIC is reduced to 20%, eICIC yields higher peak throughput for non-edge UEs while causes an obvious loss at the 5%-ile UE throughput. In particular, the macro-cell average and 5%-ile UE throughput gains are 8.92% and -52.17% respectively. The counterparts of our proposed method are 12.17% and 63.96%. Nevertheless, in Fig. 6, one can observe that eICIC can obtain better peak performance compared with our proposed method. This phenomenon can be explained by the large transmit power difference between small-cell and macro-cell. And the latter is muted at data channel in our proposed SCNs architecture.

D. Effects of UE Placement and Transmit Antennas Number Finally, we examine the impact of UEs distribution and

transmit antennas number on system performance in Fig. 7. In general, the macro-cell average throughput gain is less

sensitive to the UE distribution. On the other hand, when UEs are distributed in the uniform manner, a more significant gain on the 5%-ile UE throughput can be obtained by our proposed method. It can be explained as follows. In uniform distribution, more UEs are located in the overlapping area between adjacent small-cells, and therefore inter-cell interference is relatively more severe. After applying our proposed SCNs architecture and SCGM, inter-cell interference between macro-cells and small-cells as well as interference between adjacent small-cells is mitigated. As a

Fig. 6. Performance comparison of eICIC with RE in LTE-A and our proposed method

(a)

(b) Fig. 7. Comparison of the macro-cell average (Fig. (a)) and 5%-ile UE (Fig. (b)) throughput of baseline and the proposed method with different UEs distributions and transmit antenna numbers.

consequence, it is reasonable that the gain on cell-edge throughput is relatively more significant in uniform distribution. In particular, comparing to the counterpart in hotspot distribution, the 5%-ile UE throughput gain is increased from 60%-plus to 80%-plus in uniform distribution.

Changing the number of transmit antennas affects the performance of both the baseline and our proposed method equivalently. Therefore, the relative gains over the baseline

0 5 10 15 20 25 300

0.2

0.4

0.6

0.8

1

UE Throughput(Mbit/s)

CD

F

BaselineeICIC: RE(3dB) + ABS(20%)eICIC: RE(3dB) + ABS(50%)Macro-cell Muting + SCGM

Uniform,2Tx Uniform,4Tx Hotspot,2Tx Hotspot,4Tx0

50

100

150162.89174.35

207.40227.70

250

300

Mac

ro-c

ell a

vera

ge th

roug

hput

(Mbi

t/s)

BaselineMacro-cell Muting + SCGM 9.13%

8.90%

12.17%10.82%

Uniform,2Tx Uniform,4Tx Hotspot,2Tx Hotspot,4Tx0

0.25

0.5

0.66

0.830.94

1.17

1.5

1.75

2

5-ile

UE

thro

ughp

ut (M

bit/s

)

BaselineMacro-cell Muting + SCGM

69.01%

81.00%

63.96%82.43%

are hardly affected by the number of transmit antennas and are stable at 9% (uniform distribution) and 11% (hotspot distribution) on the macro-cell average throughput, and 82% (uniform distribution) and 66% (hotspot distribution) on the 5%-ile UE throughput approximately.

V. CONCLUSIONS In this paper, from the aspects of architecture and

deployment, we have presented a practical architecture with SCGM for LTE-A SCNs with a very dense deployment of small-cells. By separating the transmission of control channel at macro-cells and data channel at small-cells, the proposed architecture can exclude the interference between macro-cells and small-cells completely and achieve lower power consumption. With SCGM, the interference between adjacent small-cells is mitigated. The significant potential gains of our proposed method can be observed in terms of both the macro-cell average throughput and 5%-ile UE throughput.

There are many interesting research directions that can be extended from this work including: (1) to design an efficient grouping criterion for SCGM, since the criterions used in this paper are not ingenious enough and the peak performance is not high relatively; (2) to design inter-macro-cell SCGM; (3) to evaluate the effects of imperfect backhaul on our proposed method.

REFERENCES

[1] C. Comaniciu, N. B. Mandayam, and H. V. Poor, Radio resource management for green wireless networks, in Proc. IEEE 70th Veh. Technol. Conf., Sept. 2009.

[2] J. Hoydis, M. Kobayashi, and M. Debbah, Green small-cell networks, IEEE Vehicular Technology Mag., vol. 6, no.1, pp. 37~43, 2011.

[3] M. L. Qian, W. Hardjawana, Y. H. Li, B. Vucetic, J. L. Shi, and X. Z. Yang, Inter-cell interference coordination through adaptive soft frequency reuse in LTE networks, in Proc. IEEE Wireless Commun. and Networking Conf., Apr. 2012.

[4] R. Razavi, S. Kucera, C. Androne, and H. Claussen, Characterisation of other-cell interference in co-channel WCDMA small cell networks, in Proc. IEEE 75th Veh. Technol. Conf., May. 2012.

[5] A. Tolli, M. Codreanu, and M. Juntti, On the Value of Coherent and Coordinated Multi-cell Transmission, IEEE ICC Workshop 2009, Jun. 2009.

[6] Y. P. Zhang, L. Xia, P. Zhang, S. L. Feng, J. Y. Sun, and X. T. Ren, Joint transmission for LTE-Advanced Systems with non-full buffer traffic, in Proc. IEEE 75th Veh. Technol. Conf., May. 2012.

[7] L. C. Wang, and C. J. Yeh, 3-cell network MIMO architectures with sectorization and fractional frequency reuse, IEEE J. Sel. Areas Commun. vol. 29, no. 6, Jun. 2011.

[8] X. S. Wang, Y. P. Zhang, P. Zhang, and X. T. Ren, Relay-aided interference alignment for MIMO cellular networks, in Proc. IEEE Int. Symp. Inf. Theory (ISIT), Jul. 2012.

[9] Y. Y. Wang, and K. I. Pedersen, Performance analysis of enhanced inter-cell interference coordination in LTE-Advanced heterogeneous networks, in Proc. IEEE 75th Veh. Technol. Conf., May. 2012.

[10] 3GPP, "Evolved Universal Terrestrial Radio Access (E-UTRA); Further advancements for E-UTRA physical layer aspects (Release 9)", TS 36.814, v9.0.0, Mar. 2010.

[11] S. Strzyz, K. Pedersen, J. Lachowski, and F. Frederiksen, Performance optimization of pico node deployment in LTE macro cells in Proc. Future Network Mobile Summit, Jun. 2011.

/ColorImageDict > /JPEG2000ColorACSImageDict > /JPEG2000ColorImageDict > /AntiAliasGrayImages false /CropGrayImages true /GrayImageMinResolution 200 /GrayImageMinResolutionPolicy /OK /DownsampleGrayImages true /GrayImageDownsampleType /Bicubic /GrayImageResolution 300 /GrayImageDepth -1 /GrayImageMinDownsampleDepth 2 /GrayImageDownsampleThreshold 2.00333 /EncodeGrayImages true /GrayImageFilter /DCTEncode /AutoFilterGrayImages true /GrayImageAutoFilterStrategy /JPEG /GrayACSImageDict > /GrayImageDict > /JPEG2000GrayACSImageDict > /JPEG2000GrayImageDict > /AntiAliasMonoImages false /CropMonoImages true /MonoImageMinResolution 400 /MonoImageMinResolutionPolicy /OK /DownsampleMonoImages true /MonoImageDownsampleType /Bicubic /MonoImageResolution 600 /MonoImageDepth -1 /MonoImageDownsampleThreshold 1.00167 /EncodeMonoImages true /MonoImageFilter /CCITTFaxEncode /MonoImageDict > /AllowPSXObjects false /CheckCompliance [ /None ] /PDFX1aCheck false /PDFX3Check false /PDFXCompliantPDFOnly false /PDFXNoTrimBoxError true /PDFXTrimBoxToMediaBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXSetBleedBoxToMediaBox true /PDFXBleedBoxToTrimBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXOutputIntentProfile (None) /PDFXOutputConditionIdentifier () /PDFXOutputCondition () /PDFXRegistryName () /PDFXTrapped /False

/CreateJDFFile false /Description > /Namespace [ (Adobe) (Common) (1.0) ] /OtherNamespaces [ > /FormElements false /GenerateStructure false /IncludeBookmarks false /IncludeHyperlinks false /IncludeInteractive false /IncludeLayers false /IncludeProfiles true /MultimediaHandling /UseObjectSettings /Namespace [ (Adobe) (CreativeSuite) (2.0) ] /PDFXOutputIntentProfileSelector /NA /PreserveEditing false /UntaggedCMYKHandling /UseDocumentProfile /UntaggedRGBHandling /UseDocumentProfile /UseDocumentBleed false >> ]>> setdistillerparams> setpagedevice