inter-cell interferences in lte radio networks · 0.3 mbit/s for a high loaded network, and even...

1

Inter-Cell Interferences in LTE Radio Networks

Joana Falcão, Luís M. Correia Instituto Superior Técnico / INOV

University of Lisbon Lisbon, Portugal

[email protected], [email protected]

João Almeida Ericsson

Lisbon, Portugal [email protected]

Abstract - The main scope of this work is to study the impact of frequency reutilization schemes in the performance of an LTE MIMO 2x2 network, in terms of satisfied users, throughput and SINR, quantifying its key performance indicators in urban scenarios regarding low and high load of the network, for 2600, 1800 and 800 MHz bands. This purpose was achieved through a development of a suitable simulator for LTE DL, which recreates the users’ allocation in a certain instant of time, implementing a real urban mobile network, a realistic users density for the city of Lisbon, using typical urban distributions of users along the pedestrian, vehicular and indoor scenarios, and six frequency reuse schemes, including Full Frequency Reuse. The low and high load results show that frequency reuse schemes improve the average user SINR between 5 to 10 dB, some even served a similar number of satisfied users as Full Frequency Reuse. Although, always with a penalty in the average user throughput, which is around 0.3 Mbit/s for a high loaded network, and even reaches 1 Mbit/s for a low loaded network. For now, in the operators’ point of view, frequency reutilization schemes are not a reliable interference mitigation technique. Keywords - LTE, interference, frequency reutilization schemes, urban scenario, Lisbon.

I. INTRODUCTION n recent years, not only the number of mobile subscribers has increased tremendously, as also mobile

telecommunications have know great technological developments that had a great impact at social and economical level. The great challenge for mobile communications has been to respond to the demand of data communications, not only the number of mobile users is constantly increasing, as with the introduction of smartphones, the data traffic have incredible grown in the later years. Data traffic became clearly dominant with the introduction of High Speed Downlink Packet Access (HSDPA), with the third generation systems (3G). However, with the demand for higher data rates and lower latencies, the Long Term Evolution (LTE) was adopted, marked as 4th generation system (4G) [1].

As one can see in Figure 1, the predicted traffic growth for the next years in the different regions is that 2013 doubles the traffic from 2012, and that the traffic growth keeps an

exponential growth over the next four years, 2014 to 2017. Although, as shown in Figure 2, revenues do not follow the huge data growth over time, which force the operators to be more innovative and search for more cost-efficient investments that provide higher system capacity aimed.

Figure 1. Global mobile data traffic forecast by region 2012-2017 [2].

Figure 2. Traffic & Revenue: Mobile Data Gap [3].

Due to high data rates, spectrum efficiency and system latency, LTE is a technology with a great potential. However, the great challenge for LTE is the urban areas. In urban scenarios, not only the probability of a User Equipment (UE) being in Line of Sight (LoS) with an evolved Node B (eNodeB) is much lower, as also there are much more indoor users, which add a even higher attenuation. In addiction, urban environments usually have capacity limited cells, which lead to cell areas that are much smaller that in rural environments and to an increase of the overlapping coverage area between

I

2

neighbouring cells, creating interference, which proved to be a key issue for LTE, being one of the main performance limiting factors.

The main scope of this work is to study the performance of an urban LTE network for different frequency reutilization mechanisms. Several frequency reutilization mechanisms are analysed and compared, for different bandwidths and frequency bands, in terms of satisfied users, throughput and SINR. In order to perform this study, a model was developed and implemented in a simulator. The main results show the performance of the several frequency reutilization mechanisms for 2600, 1800 and 800 MHz bands, considering high and low loaded LTE networks with MIMO 2x2.

II. THEORETICAL MODELS A model was developed in order to analyse and compare

Frequency Reuse Schemes (FRS) in terms of performance on a LTE MIMO 2x2 urban network.

A. Frequency Reuse Schemes (FRS) The following list outlines the six FRS considered for

analysis in this work. Note that the schemes 4, 5 and 6 are based on schemes presented in [4].

1) Full Frequency Reuse (Reuse-1): the sectors are not limited at spectrum level, meaning that all RBs are available in each sector.

2) Fractional Frequency Reuse (FFR): this schemes is illustrated in Figure 3, each sector is divided in two regions, inner and outer region. The outer region has 1/3 of all resources available and they are divided between the three sectors. These resources will be only allocated to cell-edge users, which are at least at a certain defined distance to the eNodeB, referred as inner region limit. The remaining resources are divided between the three sectors inner region and allocated to inner region users.

Figure 3. FFR

3) Soft Frequency Reuse (SFR): this scheme is presented in Figure 4, and it is very similar to FFR, the difference is that the resources reserved to a sector outer region are transmitted at higher power than those transmitted in its inner region.

4) Fractional Sector Frequency Reuse (FSFR): this scheme is illustrated in Figure 5, and similarly to SFR the resources allocated to a sector’s outer region are transmitted at higher power than those transmitted in its inner region. The resources reserved to the inner region are divided in three groups and each sector reserves one of these groups to it, the same happening for the outer region resources. Users are preferably

allocated on the RBs intended for their serving sector, hence, in low-load conditions this FRS behaves exactly like SFR. However, when the reserved bandwidth is fully allocated, the sectors can start reusing inner RBs from their neighbouring sectors of its serving eNodeB.

Figure 4. SFR

Figure 5. FSFR [4]

5) FSFR with frontier users (FSFR-F): this scheme is presented in Figure 6, and unlike FSFR, this scheme takes into account the users near the frontier with the neighbouring sectors. A user is defined as frontier user if the difference between the powers received from the neighbouring sectors of the same eNodeB is smaller than a defined power difference limit. Each sector of a cell is divided in three regions, inner, outer and frontier region. However, in this FRS, the outer region will share its resources with the frontier region, and the resources reserved for these two regions are 1/3 of all available bandwidth.

Figure 6. FSFR-F [4]

3

6) FSFR with independent frontier region (FSFR-IF): this scheme is illustrated in Figure 7, and it is similar with FSFR-F scheme, but in this case, the frontier region has a part of the inner resources reserved only for it, not sharing resources with any region. Based on the assumptions taken on [GuHC12], it will be considered that the frontier region represents approximately 30% of the inner region and uses 1/3 of inner region resources.

Figure 7. FSFR-IF [4]

B. SNR, SINR and Throughput First, for each user is calculated the SNR for the central

frequency and its user type is defined, inner, outer or frontier region user. Then the users are ordered according to their service priority and their SNR, where the user with the higher priority service and higher SNR is the first to be served. To calculate SNR is necessary to now the average power at the receiver, which is calculated through equation (1), considering COST 231 Walfisch-Ikegami Model for pathloss:

𝑃![!"#] = 𝑃![!"#] + 𝐺![!"#] + 𝐺![!"#] − 𝐿! !" (1)

Where:

• 𝑃!: Transmitted power;

• 𝐺!: Gain from the transmitting antenna;

• 𝐺!: Gain from the receiving antenna;

• 𝐿!: Pathloss.

The pathloss, including additional can be calculated as showed in (2):

𝐿! dB = 𝐿!, !"#$ !"# !" [dB] + 𝐿!, !"#$%&"'!"( [dB] +𝑀!" [!"] (2)

Where: • 𝐿!,!"#$ !"# !" : Pathloss from the COST 231

Walfisch-Ikegami Model;

• 𝐿!,!"#$%&"'!"(: Pathloss from the user environment;

• 𝑀!": Slow fading margin.

SNR is calculated taking into account noise and the power at the receiver as shown in (3):

𝜌! [dB] = 𝑃! [dBm] − 𝑁 dBm (3)

Being:

• 𝑃!: Average signal power at the receiver;

• 𝑁: Average noise power at the receiver;

The noise power at the receiver can be obtained from (4):

𝑁[dBm] = −174 + 10 log!" ∆𝑓[Hz] + 𝐹! [dB] (4)

Where:

• ∆𝑓: Bandwidth of the radio channel being used;

• 𝐹!: Noise figure at the receiver.

Users are allocated according to their order, and they can only receive RBs reserved for their region (inner, outer or frontier region). If possible, the maximum throughput for the user’s service is guaranteed, if not, the user is only allocated if its service minimum throughput is guaranteed. After all users are allocated, the interference power felt by each user is calculated. As one can see in Figure 8, the interfering power felt by a user comes from the neighbouring sectors from its serving eNodeB and from the neighbouring sectors from other eNodeBs that the user is in range.

Figure 8. ICI and ISI.

The interference experienced by a user in a specific RB is calculated from (5), and it is sum of all powers received by the user from the neighbouring sector’s antennas, of the serving eNodeB or the neighbouring eNodeBs, that are transmitting that specific RB at the same time. For example, if the serving antenna is the only one allocating a specific RB, the user will not experience interference. But if two neighbouring sectors are allocating that specific RB too, the interference experienced by the UE will be the sum of the two interfering signal powers received at the UE.

𝐼[!"] = 𝐼! [!"]!!!!! (5)

Where:

• 𝐼!: Interference power coming from transmitter i;

• 𝑁! : Number of interfering signals reaching the

receiver.

When the total interfering power received by a RB is known, its SINR can be calculated by (6):

4

𝜌!" [dB] = 10 log!"!! [mW]

! [mW]!! [mW] (6)

When the SINR for a RB is known it is possible to calculate the real throughput given by that RB to the user. The equations used to calculate throughput were extrapolated by [5] from the latest 3GPP measurements. Equation (7) gives the throughput for QPSK modulation, equation (8) for 16-QAM and equation (9) for 64-QAM. The modulation that guarantees the best throughput is the one chosen.

𝑅! [!"#] =!.!"#∗!!"#$$%.!"!"!#"$"%∗!∗!.!""

!".!!"#$!!%&"#!"&#!!!!.!""#$%$&&%&'()*'∗!! !"

(7)

𝑅! [!"#] =!"#$%.!"!#$%&%'&#∗!∗!.!"#

!.!"#$#%&"!&'#((((!!!!.!"#$%$&'!("$"$#∗!! !"

(8)

𝑅! [!"#] =!"#$%.!"#"#$$!%&#%∗!∗!.!"#

!.!""!#$%$""!&'!"%"%!!!!.!""#$%!""$&%''"#∗!! !"

(9)

The user real throughput will be the sum of throughput for all RBs allocated to it.

III. RESULTS ANALYSIS

A. Scenarios Description The reference scenario in this work is a mobile network in

the city of Lisbon. For a more realistic approach, Lisbon is divided in two zones, centre and off-centre zone, and each zone has different density and distribution of users and eNodeBs.

The values considered in simulation for COST 231 Walfisch-Ikegami propagation model parameters are shown in TABLE I. Dense Urban is considered for the centre zone, and Urban for the off-centre zone, as the density of users and eNodeBs is higher in the centre zone and lower in the off-centre zone, in order to recreate a more realistically mobile network load for the city of Lisbon.

TABLE I. COST 231 WALFISCH-IKEGAMI PARAMETERS

Propagation model parameter Urban Dense Urban

eNodeB height (ℎ!) [m] 26 26

Buildings height (𝐻!) [m] 24 24

MT height (ℎ!) [m] 1.70 1.70

Streets widths (𝑤!) [m] 35 30

Inter buildings distance (𝑤!) [m] 75 50

Departing angle from the closest building (Φ) [m] 90 90

Four different environments were considered by assigning additional pathloss to the user in order to characterize outdoor, vehicular and indoor users. The pedestrian environment consists of a user at the street level with low attenuation margins; the vehicular one considers users performing services moving at high speed; the indoor environment characterises

users performing services inside buildings and has two variants, low and high loss, where the latter is used to consider users in deep indoor locations with higher penetration attenuation. The values for percentage of users, slow fading and indoor margin defined for the environments are shown in TABLE II. The type of MT considered for the reference scenario is a smartphone.

TABLE II. SLOW FADING, PENETRATION MARGIN AND PERCENTAGE

VALUES

Environment

Pedestrian Vehicular Indoor

Low-Loss

Indoor

High-Loss

Percentage

[%] 30 15 20 35

𝑀!" [dB] 8.8 8.8 8.8 8.8

𝐿!"# [dB] 0 11 11 21

In this thesis, six scenarios are considered, each one corresponding to one of the six FRS presented below. For each FRS, the number of RBs and transmission power per RB for each region is shown in TABLE III, for 20 MHz of bandwidth, and in TABLE IV, for 10 MHz of bandwidth.

TABLE III. NUMBER OF RBS RESERVED AND TRANSMISSION POWER

PER RB, CONSIDERING 20 MHZ BANDWIDTH

FRS

Inner Region Outer Region Frontier

Region Remaining RBs

RBs 𝑻𝑿𝑹𝑩

[dBm] RBs

𝑻𝑿𝑹𝑩

[dBm] RBs

𝑻𝑿𝑹𝑩

[dBm] RBs

𝑻𝑿𝑹𝑩,𝑴𝑨𝑿

[dBm]

Reuse-1 100 26 - - - - - -

FFR 22 30.68 11 30.68 - - 1 26

SFR 22 30 11 31.76 - - 1 31.76

FSFR 66 26.9 11 28 - - 1 28

FSFR-F 66 26.9 11 28 - - 1 28

FSFR-IF 42 26.9 11 27.9 7 28.8 4 28.8

TABLE IV. NUMBER OF RBS RESERVED AND TRANSMISSION POWER

PER RB, CONSIDERING 10 MHZ BANDWIDTH

FRS

Inner Region Outer Region Frontier

Region Remaining RBs

RBs 𝑻𝑿𝑹𝑩

[dBm] RBs

𝑻𝑿𝑹𝑩

[dBm] RBs

𝑻𝑿𝑹𝑩

[dBm] RBs

𝑻𝑿𝑹𝑩,𝑴𝑨𝑿

[dBm]

Reuse-1 50 27.78 - - - - - -

FFR 10 31.76 5 31.76 - - 5 31.76

SFR 10 31.46 5 32.04 - - 5 32.04

FSFR 30 28.45 5 29.54 - - 5 29.54

FSFR-F 30 28.45 5 29.54 - - 5 29.54

FSFR-IF 21 29.03 5 30 3 28.8 5 30

In TABLE V are presented the default parameters for the two bandwidths considered, 10MHz and 20 MHz. Note that Reuse-1 is the reference scheme. Two main analyses were performed, trying to evaluate performance of the schemes by the users’ and operators’ perspectives. One is done with few users in the network, which better represents the actual

5

situation of an LTE mobile network in Lisbon, where the cells load is still very low. The other one is done with the network highly loaded, with much more users than the ones that can be served at an instant, where schemes take place to share the bandwidth along the users and according to Quality of Service (QoS) criteria. For low load analysis, it was considered approximately 1000 covered users in the whole network, and for high load analysis, approximately 8000 users covered.

TABLE V. DEFAULT PARAMETERS FOR REFERENCE SCENARIO

Parameters DL Values for

a 20 MHz bandwidth

DL Values for a 10 MHz

bandwidth eNodeB DL Transmission

Power [dBm] 46 44.77

Frequency [MHz] 2600 Bandwidth [MHz] 20 10

MIMO Configuration 2x2 Modulation QPSK, 16QAM, 64QAM

Maximum eNodeB Antennas Gain [dBi] 17.6

UE Antenna Gain [dBi] 0 User Losses [dB] 1 Cable Losses [dB] 3

Users Noise Figure [dB] 7 Slow Fading Margin [dB] 8.8

FRS Reuse-1 Inner Region Limit [m] 300

Max. Power Difference for Frontier Region [dB] 3

Scheduling Algorithm Proportional Fair

The users are served according to their service priority and SNR, hence, the first users to be served are the ones with the higher service priority and the higher SNR. Services minimum and maximum throughputs are presented in TABLE VI, with the respective QoS priority and penetration. The users were specified as covered or served, and the served users were classified as satisfied or unsatisfied after the interference calculation.

TABLE VI. SMARTPHONE SERVICES CHARACTERIZATION

Service QoS

Priority

Penetration [%]

Minimum Throughput [Mbit/s]

Maximum Throughput [Mbit/s]

Streaming 1 36 1.024 6

Chat 2 5.5 0.064 0.384 Web

Browsing 3 25 1.024 20

FTP 4 9.5 1.024 21.5 Email 5 6 1.024 8 P2P 6 18 1.024 5

B. High Load Scenarios Results Analysis

This section presents the influence of the six FRS under study for a mobile network in the city of Lisbon, in high load conditions. The six schemes are analysed and compared in terms of average user SINR and average user throughput. The 2600 MHz and 1800 MHz bands are analysed for a 20 MHz bandwidth and the 2600 MHz and 800 MHz bands, for a 10

MHz bandwidth. Note that the results are always shown for the Lisbon centre and off-centre zone.

As one can see in Figure 9 and Figure 10, when considering a 20 MHz bandwidth, Reuse-1 scheme has the worst average SINR for satisfied users for 2600 MHz and 1800 MHz. FFR and SFR present the best average user SINR, around 10 dB better than Reuse-1. FSFR, FSFR-F and FSFR-IF are approximately 5 dB better than Reuse-1. Note that the Lisbon off-centre has always a better average user SINR for the six FRS, which is expected due to the smaller density of users and eNodeBs in this zone.

Figure 9. Average SINR for satisfied users with SINR higher than 5

dB, considering a 20 MHz bandwidth, in the 2600 MHz band

Figure 10. Average SINR for satisfied users with SINR higher than 5 dB, considering a 20 MHz bandwidth, in the 1800 MHz band

In Figure 11 and Figure 12 is presented the average SINR for satisfied users for the 2600 MHz and 800 MHz band, considering a 10 MHz bandwidth. The results are very similar to 2600 MHz and 1800 MHz bandwidth when considering 20 MHz bandwidth, being again Reuse-1 the scheme with the worst average user SINR. FFR and SFR are again the best ones, showing that the limitation of RBs really improves the average user SINR. Note that the average user SINR for the 800 MHz band is a little smaller for the six FRS than for the 2600 MHz band.

6

Figure 11. Average SINR for satisfied users, considering high load, a

10 MHz bandwidth, in the 2600 MHz band

Figure 12. Average SINR for satisfied users with SINR high than 5


The average user throughput for the 2600 MHz and 1800 MHz bands, considering a 20 MHz bandwidth, is shown in Figure 13 and Figure 14. It is clear that Reuse-1 has the best average user throughput for the two bands considered. On the other hand, FFR and SFR present the worst average user throughput for the two zones of Lisbon and the two bands, which means that the limitation of RBs really improve the average user SINR but with a penalty on average user throughput, being more severe for FFR and SFR schemes. Note that the average user throughput is lower for the 1800 MHz, but it was verified that this band serves more users than 2600 MHz band.

Figure 13. Average user throughput, taking interference into account,

a considering a 20 MHz bandwidth, in the 2600 MHz band

Figure 14.Average user throughput, taking interference into account,

considering a 20 MHz bandwidth, in the 1800 MHz band

The average user throughput for the 2600 MHz and 800 MHz band when considering a 10 MHz bandwidth is presented in Figure 15 and Figure 16. As happened for the 2600 MHz and 1800 MHz when considering a 20 MHz bandwidth, Reuse-1 scheme presents the best average user throughput despite of having the worst average user SINR. Note that the average user throughput in the 800 MHz band is lower than in the 2600 MHz, however it was verified that the 800 MHz band serves more satisfied users.





C. Low Load Scenarios Results Analysis

This section presents the influence of the six FRS under study for a mobile network in the city of Lisbon, in low load

7

conditions. The six schemes are again analysed and compared in terms of average user SINR and average user throughput. As was done for a high loaded network, the 2600 MHz and 1800 MHz bands are analysed for a 20 MHz bandwidth and the 2600 MHz and 800 MHz bands, for a 10 MHz bandwidth.

The average SINR of satisfied users with a SINR superior to 5 dB is shown in Figure 17 and Figure 18. As expected, for a low loaded network, the average user SINR for the six FRS, in the three frequency bands, is higher than for a high loaded network. However, Reuse-1 presents the worst average user SINR for the two zones of Lisbon and for the two frequency bands considered, as happened for a high loaded network. One can see that unlike what happened for high load of the network, FSFR, FSFR-F and FSFR-IF schemes have a similar average user SINR to FFR and SFR schemes, which is around 5 dB for the centre zone and 7 dB for the off-centre zone, in the 2600 MHz and 1800 MHz bands.



Figure 18. Average SINR for satisfied users with SINR superior to 5


Considering a 10 MHz bandwidth, the average user SINR for the 2600 MHz and 800 MHz band is presented in Figure 20 and Figure 19. Again, Reuse-1 has the worst average user SINR of all schemes. As for a 10 MHz bandwidth the there are less RBs available than for a 20 MHz bandwidth, FFR and SFR guarantee an average user SINR significantly superior to FSFR, FSFR-F and FSFR-IF, in the 2600 MHz and 800 MHz band.





In Figure 21 and Figure 22 is shown the average user throughput for the 2600 MHz and 1800 MHz band. For a low loaded network, Reuse-1 continues to have the better average user throughput, but in this case the difference to the other five FRS is even bigger. As expected, for a low loaded network the average user throughput is higher than for a high loaded network. Note that for Reuse-1 scheme, in low load of the network, 1800 MHz band guarantees an average user throughput higher than the 2600 MHz band for the two zones of Lisbon.



8



The average user throughput for the 2600 MHz and 800 MHz band is presented in Figure 23 and Figure 24. As in all other analysis Reuse-1 has the best average user throughput, in the 800 MHz band the difference to the other FRS is 1 Mbit/s or more.




considering low load, a 10 MHz bandwidth, in the 800 MHz band

D. Global Results Analysis From the high and low load analysis one can see that FRS really improve the average user SINR, normally increasing it from 5 dB to 10 dB for the three frequency bands considered. Regarding only high load analysis, FFR and SFR are the schemes that most improve average user SINR, although the average number of satisfied users and the average user throughput in these schemes are always significantly lower

than in Reuse-1. It was verified that the schemes that consider inner region RBs reuse from the other neighbouring sectors of his serving cell, FSFR, FSFR-F and FSFR-IF, serve a similar number of users to Reuse-1. Although, for the lower frequency bands, especially for the 800 MHz band, Reuse-1 serves more satisfied users than the other FRS. Nevertheless, the most important parameter of comparison between FRS is the average user throughput, and in this parameter Reuse-1 always obtains the better results. This is even more noticeable for a low loaded network, where Reuse-1 even reach an average user throughput approximately 1 Mbit/s better than the second best FRS. So, Reuse-1 scheme offers the best average user throughput in all analysis, in spite of average user SINR being much worse than on the other FRS. This means that in the overall, frequency reutilization mechanisms improve the average user SINR but always with a penalty in average user throughput.

IV. CONCLUSIONS The obtained results show that for an LTE MIMO 2x2 network in a urban environment the FRSs implemented effectively decrease the interference felt by users and even sometimes serve more satisfied users than Reuse-1, but always with a penalty in the users’ throughput, being this penalty much more severe when the network have a low load. The penalty in the average user throughput comes with the limitation of RBs for each sector imposed by the frequency reuse mechanisms, which indeed improves the users’ channel conditions but in the overall give less RBs to each user, leading to an average user throughput below the average for Reuse-1. Due to the higher priority services that have a higher concentration of users and a higher maximum throughput, as is the case of Web, FFR and SFR schemes have the worst number of satisfied users and average user throughput. Note that when possible, the maximum throughput for all users’ service is guaranteed, not limiting the maximum number of RBs per user. As for these FFR and SFR schemes there are less RBs reserved each region, the resources on each sector obviously starve faster than for the other FRS, which leads to worst results. For a high loaded network FSFR serve about the same satisfied users and have an average user throughput not much below Reuse-1. For a low loaded network, Reuse-1 is clearly the best scheme, guaranteeing a significantly higher average user throughput than the other FRSs. The results for high and low load of the network lead to the conclusion that for today’s LTE MIMO 2x2 networks the frequency reutilization mechanisms do not bring great improvements, therefore supporting that the limitation of the spectrum does not bring the advantages expected and that for now, in the point of view of operators, they are not a reliable alternative to mitigate the interference in LTE. Regarding future work, despite of the conclusions obtained in this thesis, that shows that frequency reuse mechanisms can really improve SINR but always with a penalty in the users throughput, there are many improvements that can be made that are not addressed in this thesis and can lead to better results and different conclusions. MIMO 4x4 and 8x8 could be implemented, for these antenna configurations the throughput per RB is much higher than for MIMO 2x2, and the limitation of RBs imposed by the frequency reuse mechanisms may not

9

starve so much the resources for each sector leading to better results. It would also be interesting to study the impact of frequency reuse mechanisms for higher bandwidths, like 50 MHz or 100 MHz, as the number of RBs is much higher for this bandwidths, different distribution of the RBs for the sectors and regions could be tested. Moreover, if an allocation of resources with temporal depth was implemented, it would be possible to test the dynamic frequency reuse mechanisms, which are the most promising until now and have led to better results. In these dynamic FRSs, the RBs are distributed through the regions depending on the number of users and their distribution along the eNodeBs, better adapting the number of resources available in each region to users’ needs, in theory leading to better results.

REFERENCES

[1] Holma, H. and Toskala, A., LTE for UMTS: Evolution to LTE-Advanced, John Wiley & Sons, West Sussex, UK, 2011. [2] Cisco VNI Forecast, Cisco Visual Networking Index: Global Mobile Data Traffic Forecast Update, 2012-2017, Feb. 2013, (http://www.cisco.com/en/US/solutions/collateral/ns341/ns525/ns537/ns705/ns827/white_paper_c11-520862.pdf). [3] Openet, How Smart Operators are Closing the Mobile Data Revenue Gap, 2013, (http://www.openet.com/resources/whitepapers). [4] Guío, I., Hernández, Á., Chóliz, J., and Valdovinos, A., "Resource allocation strategies for full frequency reuse in tri-sectorized multi-cell orthogonal frequency division multiple access systems", 2012. [5] Almeida, D. X., Inter-Cell Interference Impact on LTE Performance in Urban Scenarios, M.Sc. Thesis, Instituto Superior Técnico, Lisbon, Portugal, 2013.

inter-cell interferences in lte radio networks · 0.3 mbit/s for a high loaded network, and even...

Documents