[ieee 2013 36th international conference on telecommunications and signal processing (tsp) - rome,...

5
Using Multiple Co-Channel Femtocells as Relays to Increase the Performance of the Outdoor User Samuel Baraldi Mafra, Hirley Alves, Richard Demo Souza, Evelio M. G. Fernandez, and Jo˜ ao Luiz Rebelatto Abstractβ€”The deployment of femtocells comes as a solution to increase indoor coverage although it may also increase the interference level in the network. In this paper we evaluate the use of multiple co-channel femtocells as relays to the outdoor user in a cognitive radio context, where the outdoor user can be seen as a primary user and the indoor users are seen as secondary users. We show that multiple co-channel femtocells, besides allowing several secondary users to access the spectrum, may considerably increase the performance of the primary user if the femtocells act as relays to the primary signal. Moreover, results show that with the increase of the number of cooperative co-channel femtocells the capacity constraints on the backhaul can be relaxed. Keywordsβ€”Cognitive Radio, Cooperative Communications, Femtocells, Heterogeneous Networks, Outage Analysis I. I NTRODUCTION F EMTOCELLS are covered by low cost and low power base stations (BS), providing coverage in indoor en- vironments where the macrocell signal is limited or not detectable. Femtocells often provide access to the cellular network through a digital subscriber line or optical fiber connection. With the inclusion of femto base stations, indoor users perceive an improved signal quality due to the proximity between transmitter and receiver, resulting in a highly efficient communication and large transmission rates [1]. Moreover, femtocells can operate either under the so-called open access mode, where all users (indoor and outdoor) may access the femtocell, or under the closed access mode, in which only previously allowed users can connect to the femtocell. The co-existence of macrocell and femtocells gives rise to a two-tier architecture. In this scenario we have [2]: i) cross-tier interference, in which the femtocell interferes on the macrocell and vice-versa; and ii) co-tier interference, which is the interference between neighbor femtocells. For instance, the interference can be addressed by means of power control techniques [3], [4], OFDMA [5], [6], cognitive radio solutions [7]–[12], etc. For instance, in [11], the authors propose a scheme of spectrum reuse based on cognitive radio, where the channel reuse is determined according to each femtocell’s channel environment, leading to a great SINR Manuscript received February 11, 2013. This work was supported by CAPES (Brazil), CNPq (Brazil), and Infotech Graduate School, University of Oulu (Finland). Samuel Baraldi Mafra, Richard Demo Souza, and Jo˜ ao Luiz Rebelatto are with Federal University of Technology-ParanΒ΄ a (UTFPR), Curitiba-PR, Brazil (e-mails: [email protected],{ richard, jlrebelatto}@utfpr.edu.br) Hirley Alves is with Centre for Wireless Communications (CWC), Oulu, Finland.(e-mail: [email protected].fi). Evelio M. G. Fernandez is with Federal University of ParanΒ΄ a (UFPR) Curitiba-PR, Brazil (e-mail: [email protected]). improvement. A cognitive WiMAX architecture is proposed in [12], where the base station and users intelligently adjust parameters such as power, channel, etc. The results show a substantial performance improvement in comparison to the protocol without the cognitive radio technique. Another al- ternative to deal with the interference is through the use of cooperative communications [13], where one or more nodes help the communication between source and destination by acting as relays. The way the relay behaves is determined by the cooperative protocols [13]. One of the most well known cooperative protocols is the decode-and-forward (DF), in which the relay tries to decode the source message and then re-encodes and forwards it to the destination. Two methods of cooperation between secondary users (fem- tocell) and primary users (macrocell) in a cognitive hetero- geneous network are investigated in [14]: i) the cooperative relay model, where the secondary user relays the message of the primary user, but is not able to transmit its own message concurrently with the primary user; and ii) the interference model, where the secondary user is allowed to transmit its own message concurrently with the primary user, as long as the interference imposed to the primary user remains below a required threshold. In [15], the use of femtocell as a relay for both indoor and outdoor users is proposed. The system performance is evaluated in terms of outage probability. The use of the femtocell home BS (or HBS) as a relay increases the network performance. In closed access it is possible to obtain the same outage probability than in a system without femtocells, thus completely eliminating the cross-tier interference. In open access, the authors demonstrate that performance is even better than in a system without femtocells, as both users benefit of the relaying provided by the HBS. In this paper we consider a similar scenario as in [15], but assuming the existence of multiple co-channel underlaid HBSs. We show that having multiple co-channel HBSs, be- sides allowing more secondary indoor users to access the spectrum, can increase the performance of the primary outdoor user when the multiple co-channel HBSs act as relays to the outdoor user signal. Moreover, we show that by having multiple underlaid HBSs, besides increasing the performance of the outdoor user, we can reduce the backhaul capacity requirement. The remainder of this paper is organized as follows. Sec- tion II introduces the system model, while Section III describes the outage probability analysis of the indoor and the outdoor users. Section IV presents some numerical results and, finally, Section V concludes the paper. 166 978-1-4799-0404-4/13/$31.00 Β©2013 IEEE TSP 2013

Upload: joao-luiz

Post on 16-Feb-2017

213 views

Category:

Documents


1 download

TRANSCRIPT

Using Multiple Co-Channel Femtocells as Relays toIncrease the Performance of the Outdoor User

Samuel Baraldi Mafra, Hirley Alves, Richard Demo Souza, Evelio M. G. Fernandez, and Joao Luiz Rebelatto

Abstractβ€”The deployment of femtocells comes as a solutionto increase indoor coverage although it may also increase theinterference level in the network. In this paper we evaluate theuse of multiple co-channel femtocells as relays to the outdooruser in a cognitive radio context, where the outdoor user canbe seen as a primary user and the indoor users are seen assecondary users. We show that multiple co-channel femtocells,besides allowing several secondary users to access the spectrum,may considerably increase the performance of the primary userif the femtocells act as relays to the primary signal. Moreover,results show that with the increase of the number of cooperativeco-channel femtocells the capacity constraints on the backhaulcan be relaxed.

Keywordsβ€”Cognitive Radio, Cooperative Communications,Femtocells, Heterogeneous Networks, Outage Analysis

I. INTRODUCTION

FEMTOCELLS are covered by low cost and low powerbase stations (BS), providing coverage in indoor en-

vironments where the macrocell signal is limited or notdetectable. Femtocells often provide access to the cellularnetwork through a digital subscriber line or optical fiberconnection. With the inclusion of femto base stations, indoorusers perceive an improved signal quality due to the proximitybetween transmitter and receiver, resulting in a highly efficientcommunication and large transmission rates [1]. Moreover,femtocells can operate either under the so-called open accessmode, where all users (indoor and outdoor) may access thefemtocell, or under the closed access mode, in which onlypreviously allowed users can connect to the femtocell.

The co-existence of macrocell and femtocells gives riseto a two-tier architecture. In this scenario we have [2]: i)cross-tier interference, in which the femtocell interferes onthe macrocell and vice-versa; and ii) co-tier interference,which is the interference between neighbor femtocells. Forinstance, the interference can be addressed by means of powercontrol techniques [3], [4], OFDMA [5], [6], cognitive radiosolutions [7]–[12], etc. For instance, in [11], the authorspropose a scheme of spectrum reuse based on cognitive radio,where the channel reuse is determined according to eachfemtocell’s channel environment, leading to a great SINR

Manuscript received February 11, 2013. This work was supported byCAPES (Brazil), CNPq (Brazil), and Infotech Graduate School, Universityof Oulu (Finland).

Samuel Baraldi Mafra, Richard Demo Souza, and Joao Luiz Rebelatto arewith Federal University of Technology-Parana (UTFPR), Curitiba-PR, Brazil(e-mails: [email protected],{ richard, jlrebelatto}@utfpr.edu.br)

Hirley Alves is with Centre for Wireless Communications (CWC), Oulu,Finland.(e-mail: [email protected]).

Evelio M. G. Fernandez is with Federal University of Parana (UFPR)Curitiba-PR, Brazil (e-mail: [email protected]).

improvement. A cognitive WiMAX architecture is proposedin [12], where the base station and users intelligently adjustparameters such as power, channel, etc. The results show asubstantial performance improvement in comparison to theprotocol without the cognitive radio technique. Another al-ternative to deal with the interference is through the use ofcooperative communications [13], where one or more nodeshelp the communication between source and destination byacting as relays. The way the relay behaves is determinedby the cooperative protocols [13]. One of the most wellknown cooperative protocols is the decode-and-forward (DF),in which the relay tries to decode the source message and thenre-encodes and forwards it to the destination.

Two methods of cooperation between secondary users (fem-tocell) and primary users (macrocell) in a cognitive hetero-geneous network are investigated in [14]: i) the cooperativerelay model, where the secondary user relays the message ofthe primary user, but is not able to transmit its own messageconcurrently with the primary user; and ii) the interferencemodel, where the secondary user is allowed to transmit itsown message concurrently with the primary user, as long asthe interference imposed to the primary user remains below arequired threshold.

In [15], the use of femtocell as a relay for both indoorand outdoor users is proposed. The system performance isevaluated in terms of outage probability. The use of thefemtocell home BS (or HBS) as a relay increases the networkperformance. In closed access it is possible to obtain thesame outage probability than in a system without femtocells,thus completely eliminating the cross-tier interference. In openaccess, the authors demonstrate that performance is even betterthan in a system without femtocells, as both users benefit ofthe relaying provided by the HBS.

In this paper we consider a similar scenario as in [15],but assuming the existence of multiple co-channel underlaidHBSs. We show that having multiple co-channel HBSs, be-sides allowing more secondary indoor users to access thespectrum, can increase the performance of the primary outdooruser when the multiple co-channel HBSs act as relays tothe outdoor user signal. Moreover, we show that by havingmultiple underlaid HBSs, besides increasing the performanceof the outdoor user, we can reduce the backhaul capacityrequirement.

The remainder of this paper is organized as follows. Sec-tion II introduces the system model, while Section III describesthe outage probability analysis of the indoor and the outdoorusers. Section IV presents some numerical results and, finally,Section V concludes the paper.

166978-1-4799-0404-4/13/$31.00 Β©2013 IEEE TSP 2013

II. SYSTEM MODEL

We consider the uplink of a two tier heterogeneous networkcomposed of a base station (BS), 𝑁 femtocells, and a macrooutdoor user, as depicted in Fig. 1. As we focus on a singlechannel, we assume one indoor user per femtocell. The macrooutdoor user can be seen as the primary user while the indooruser can be seen as the secondary user in a cognitive radiocontext. The 𝑁 indoor users and the outdoor user transmit atthe same time and frequency. It is also assumed that eachHBS has 𝐿 antennas, while the macro BS and all usersare considered to be single antenna devices1. Each HBStries to decode the messages of the outdoor user and of itscorresponding indoor user. Therefore, the femtocells operatein the open access mode with respect to the outdoor user,while the messages of other femtocell indoor users are treatedas noise. Each HBS is connected to the macro BS by meansof a wired, assumed error free, backhaul link.

If the HBS of the π‘˜-th femtocell, π‘˜ ∈ {1, 2, β‹… β‹… β‹… , 𝑁},decodes the messages of the outdoor and its indoor user, thenthat femtocell allocates π›Ώπ‘˜ β‹…πΆπ‘˜ bits per channel use (bpcu) of itsbackhaul capacity to the π‘˜-th indoor user and (1βˆ’π›Ώπ‘˜)β‹…πΆπ‘˜ bpcuto the outdoor user, where πΆπ‘˜ is the total backhaul capacity ofthe π‘˜-th femtocell and 0 ≀ π›Ώπ‘˜ ≀ 1. We impose an additionalconstraint in the choice of π›Ώπ‘˜ so that π›Ώπ‘˜ β‹… πΆπ‘˜ is always atleast as large as the indoor user required capacity. If the π‘˜-thHBS correctly decoded only one of the two messages, then itdedicates the total backhaul capacity πΆπ‘˜ for either the outdooror the corresponding indoor user. The macro BS only tries todecode the outdoor user, as decoding the indoor users at themacro BS does not bring considerable additional gains, unlessthe femtocell is very close to the macro BS. Moreover, if themacro BS tries to decode all indoor users then the complexitygrows very large with the increase of 𝑁 .

The channel fading between the indoor user of the π‘˜-thfemtocell and the macro BS is denoted as π‘”π‘˜, while π‘”π‘œ is thechannel fading between the outdoor user and the macro BS.Moreover, β„Žπ‘–

𝑗,π‘˜ corresponds to the channel between the indooruser of the 𝑗-th femtocell and the 𝑖-th antenna of the HBS inthe π‘˜-th femtocell, π‘˜ ∈ {1, 2, β‹… β‹… β‹… , 𝑁} and 𝑖 ∈ {1, 2, β‹… β‹… β‹… , 𝐿},while β„Žπ‘–

π‘œ,π‘˜ is the channel between the outdoor user and the𝑖-th antenna of the HBS in the π‘˜-th femtocell. All channelsare assumed to be independent of each other and follow aRayleigh quasi-static fading model. Indoor and outdoor userstransmit with rates π‘…π‘˜ and π‘…π‘œ bpcu, respectively.

The received signal at the macro BS is given by

𝑦𝑏 =

βˆšπ‘ƒπ‘œ

πœ…π‘œ,π‘π‘”π‘œπ‘₯π‘œ +

π‘βˆ‘π‘˜=1

βˆšπ‘ƒπ‘˜

πœ…π‘˜,π‘π‘”π‘˜π‘₯π‘˜ + 𝑀𝑏, (1)

where π‘ƒπ‘œ and π‘ƒπ‘˜ are the transmit powers of the outdoor userand the indoor user in the π‘˜-th femtocell, respectively, πœ…π‘œ,𝑏 β‰₯ 1is the path loss between the outdoor user and the macro BS,πœ…π‘˜,𝑏 is the path loss between the π‘˜-th indoor user and themacro BS, π‘₯π‘œ and π‘₯π‘˜ are the unity energy Gaussian complex

1Including additional antennas to the macro BS would not qualitativelychange our conclusions, and therefore we decided to consider a single antennamacro BS for the sake of simplicity. However, this extension can be consideredin a future work.

IndoorUser

HBS

Outdoor User

TransmissionLink

Interference Link

Base Station

IndoorUser

HBS

Fig. 1. System model with 𝑁 = 2 femtocells and one macrocell user

messages from the outdoor user and the π‘˜-th indoor user,and 𝑀𝑏 is additive white Gaussian noise with variance 𝑁0/2per dimension, while without loss of generality we assumehereinafter that 𝑁0 = 1. The messages transmitted by theindoor users are seen as noise at the macro BS, as it tries todecode only the outdoor user message π‘₯π‘œ.

The signal received at the 𝑖-th antenna of the π‘˜-th HBS is

π‘¦π‘–π‘˜ =

βˆšπ‘ƒπ‘˜

πœ…π‘˜,π‘˜β„Žπ‘–π‘˜,π‘˜π‘₯π‘˜+

βˆšπ‘ƒπ‘œ

πœ…π‘œ,π‘˜β„Žπ‘–π‘œ,π‘˜π‘₯π‘œ+

π‘βˆ‘π‘—=1𝑗 βˆ•=π‘˜

βˆšπ‘ƒπ‘—

πœ…π‘—,π‘˜β„Žπ‘–π‘—,π‘˜π‘₯𝑗+𝑀𝑖

π‘˜, (2)

where 𝑃𝑗 is the transmit power of the indoor user in the 𝑗-thfemtocell, πœ…π‘—,π‘˜ is the path loss between the indoor user in the𝑗-th femtocell and the HBS in the π‘˜-th femtocell, the other πœ…β€™sfollow a similar definition, respecting the notation used so far,while 𝑀𝑖

π‘˜ is additive Gaussian noise just as 𝑀𝑏. The first termin (2) corresponds to the message transmitted by the indooruser of the π‘˜-th femtocell while the second term correspondsto the message of the outdoor user. The HBS tries to decodethese two messages, while the messages of the other 𝑁 βˆ’ 1indoor users (third term) are seen as noise.

In this work, as in [16], [17], we consider a path loss modelwhich describes a real femtocell scenario, so that the path lossbetween an indoor or outdoor user and a HBS and betweenan indoor user and the macro BS is defined as

πœ…(𝑑𝐡) = 39.676 + 20 log10 𝑑+ π‘žπ‘Š +π‘Š β€², (3)

where the first term corresponds to the free-space loss, 𝑑 isthe distance between transmitter and receiver in meters, π‘Š isthe wall partition loss (assumed to be 5dB), π‘ž is the number of

167

walls between transmitter and receiver, and π‘Š β€² is the outdoor-indoor penetration loss (assumed to be 10dB) and which isused only when applies. Moreover, the outdoor-to-macro BSpath loss model is defined as πœ…(𝑑𝐡) = 15.3 + 37.6 log10 𝑑.

The average signal-to-noise ratio (SNR) of the outdoor andindoor users to the macro BS are respectively given by π›Ύπ‘œ,𝑏 =π‘ƒπ‘œ

πœ…π‘œ,𝑏and π›Ύπ‘˜,𝑏 = π‘ƒπ‘˜

πœ…π‘˜,𝑏, while the SNR of the link between

the outdoor and indoor user of the 𝑗-th femtocell to the 𝑖-th antenna of the HBS in the π‘˜-th femtocell are respectivelygiven by 𝛾𝑖

𝑗,π‘˜ = 𝛾𝑗,π‘˜ = π‘ƒπ‘˜

πœ…π‘—,π‘˜and 𝛾𝑖

π‘œ,π‘˜ = π›Ύπ‘œ,π‘˜ = π‘ƒπ‘œ

πœ…π‘œ,π‘˜.

III. OUTAGE PROBABILITY ANALYSIS

First we consider the outage probability seen at the π‘˜-thfemtocell. Thus, let π’«π‘˜,π‘π‘œπ‘‘β„Ž be the probability that the HBS ofthe π‘˜-th femtocell correctly decoded its corresponding indooruser and the outdoor user. Then, supposing a multiple accesschannel (MAC) model and using some of the results in [18],we can write

π’«π‘˜,π‘π‘œπ‘‘β„Ž = 1βˆ’ π’«π‘˜,π‘œπ‘’π‘‘ βˆ’ π’«π‘˜,π‘œπ‘’π‘‘, (4)

where π’«π‘˜,π‘œπ‘’π‘‘ is the probability that the HBS correctly decodedonly the outdoor user, while π’«π‘˜,π‘œπ‘’π‘‘ is the probability that theHBS did not decode the outdoor user. Moreover, π’«π‘˜,π‘œπ‘’π‘‘ canbe defined as π’«π‘˜,π‘œπ‘’π‘‘ = Pr{π’œβˆ©β„¬}, where events π’œ and ℬ are

π’œ := log2

(1 +

βˆ‘πΏπ‘–=1 βˆ£β„Žπ‘–

π‘œ,π‘˜βˆ£2π›Ύπ‘œ,π‘˜1 +

βˆ‘πΏπ‘–=1 βˆ£β„Žπ‘–

π‘˜,π‘˜βˆ£2π›Ύπ‘˜,π‘˜ + πœƒ

)β‰₯ π‘…π‘œ, (5)

ℬ := log2

(1 +

βˆ‘πΏπ‘–=1 βˆ£β„Žπ‘–

π‘˜,π‘˜βˆ£2π›Ύπ‘˜,π‘˜1 + πœƒ

)< π‘…π‘˜, (6)

where we assume that the HBS applies maximum ratio com-bining (MRC) among the receive antennas, where πœƒ is givenby:

πœƒ =

π‘βˆ‘π‘—=1,𝑗 βˆ•=π‘˜

πΏβˆ‘π‘–=1

βˆ£β„Žπ‘–π‘—,π‘˜βˆ£2𝛾𝑗,π‘˜. (7)

Note that (5) and (6) are found considering the capacityregion of the two-user MAC where the outdoor user is decodedfirst (seeing the π‘˜-th indoor user as noise), while in theattempt to decode the π‘˜-th indoor user the interference fromthe outdoor user is removed as the latter could be decoded.Moreover, πœƒ represents the interference from the other 𝑁 βˆ’ 1indoor users (which are not decoded at the HBS).

The probability π’«π‘˜,π‘œπ‘’π‘‘ can be defined as π’«π‘˜,π‘œπ‘’π‘‘ = π’«π‘˜,π‘›π‘œπ‘›π‘’+π‘ƒπ‘˜,𝑖𝑛, where π’«π‘˜,π‘›π‘œπ‘›π‘’ and π’«π‘˜,𝑖𝑛 are the probabilities that theHBS could not decode both users and the probability thatthe HBS could decode only the corresponding indoor user,respectively. Moreover the probability π’«π‘˜,𝑖𝑛 is given by

π’«π‘˜,𝑖𝑛 = Pr{π’ž ∩ π’Ÿ}, (8)

where, still considering a two user MAC, events C and D are

π’ž := log2

(1 +

βˆ‘πΏπ‘–=1 βˆ£β„Žπ‘–

π‘˜,π‘˜βˆ£2π›Ύπ‘˜,π‘˜1 +

βˆ‘πΏπ‘–=1 βˆ£β„Žπ‘–

π‘œ,π‘˜βˆ£2π›Ύπ‘œ,π‘˜ + πœƒ

)β‰₯ π‘…π‘˜, (9)

π’Ÿ := log2

(1 +

βˆ‘πΏπ‘–=1 βˆ£β„Žπ‘–

π‘œ,π‘˜βˆ£2π›Ύπ‘œ,π‘˜1 + πœƒ

)< π‘…π‘œ. (10)

The probability π’«π‘˜,π‘›π‘œπ‘›π‘’ is given by

π’«π‘˜,π‘›π‘œπ‘›π‘’ = Pr{β„° ∩ β„± ∩ 𝒒}, (11)

where events β„° , β„± and 𝒒 are

β„° := log2

(1 +

βˆ‘πΏπ‘–=1 βˆ£β„Žπ‘–

π‘˜,π‘˜βˆ£2π›Ύπ‘˜,π‘˜1 +

βˆ‘πΏπ‘–=1 βˆ£β„Žπ‘–

π‘œ,π‘˜βˆ£2π›Ύπ‘œ,π‘˜ + πœƒ

)< π‘…π‘˜, (12)

β„± := log2

(1 +

βˆ‘πΏπ‘–=1 βˆ£β„Žπ‘–

π‘œ,π‘˜βˆ£2π›Ύπ‘œ,π‘˜1 +

βˆ‘πΏπ‘–=1 βˆ£β„Žπ‘–

π‘˜,π‘˜βˆ£2π›Ύπ‘˜,π‘˜ + πœƒ

)< π‘…π‘œ, (13)

𝒒 := log2

(1 +

βˆ‘πΏπ‘–=1 βˆ£β„Žπ‘–

π‘œ,π‘˜βˆ£2π›Ύπ‘œ,π‘˜ + βˆ£β„Žπ‘–π‘˜,π‘˜βˆ£2π›Ύπ‘˜,π‘˜

1 + πœƒ

)< π‘…π‘œ+π‘…π‘˜.

(14)Next, we write the outage probability of the outdoor user

at the macro BS, π’ͺπ‘π‘ βˆ£e, conditioned on the correspondingvector of decoding events e = [𝑒1, ..., 𝑒𝑁 ] at the 𝑁 femtocells.A given decoding event at the π‘˜-th femtocell is defined asπ‘’π‘˜ ∈ Ξ¦ = {π‘œπ‘’π‘‘, 𝑖𝑛, π‘π‘œπ‘‘β„Ž, π‘›π‘œπ‘›π‘’}, π‘˜ ∈ {1, 2, ..., 𝑁}, whereπ‘œπ‘’π‘‘ means that only the outdoor user could be decoded atthe π‘˜-th femtocell, 𝑖𝑛 represents the situation in which onlythe corresponding indoor user is decoded, π‘π‘œπ‘‘β„Ž indicates thatboth users could be decoded, and π‘›π‘œπ‘›π‘’ means that neither theoutdoor nor the indoor user could be decoded. For instance,suppose that 𝑁 = 2 and that the vector of decoding events ise = [π‘œπ‘’π‘‘, π‘π‘œπ‘‘β„Ž]. Then, e indicates that in the first femtocellonly the outdoor user could be decoded while in the secondfemtocell both the corresponding indoor user and the outdooruser could be decoded.

Thus, the outage probability of the outdoor user at themacro BS given a specific vector of decoding events at the𝑁 femtocells is given by

π’ͺπ‘π‘ βˆ£e = log2

(1 +

βˆ£β„Žπ‘œ,π‘βˆ£2π›Ύπ‘œ,𝑏1 +

βˆ‘π‘π‘˜=1 π›½π‘˜ β‹… βˆ£β„Žπ‘˜,π‘βˆ£2π›Ύπ‘˜,𝑏

)

<

(π‘…π‘œ βˆ’

π‘βˆ‘π‘˜=1

(π‘Žπ‘˜ β‹… (1βˆ’ π›Ώπ‘˜)πΆπ‘˜)βˆ’π‘βˆ‘

π‘˜=1

π‘π‘˜ β‹… πΆπ‘˜

), (15)

where the binary variable π‘Žπ‘˜ is equal to one if the π‘˜-thfemtocell could decode both users, so that a capacity of(1 βˆ’ π›Ώπ‘˜)πΆπ‘˜ over the backhaul is used to convey informationon the outdoor user message to the macro BS. The binaryvariable π‘π‘˜ is equal to one if the π‘˜-th femtocell could decodeonly the outdoor user, while the binary variable π›½π‘˜ is equal toone if the π‘˜-th femtocell could not decode the indoor user2.The possible values of π‘Žπ‘˜, π‘π‘˜ and π›½π‘˜ are given in Table I.

In order to write the overall outdoor user outage probabilityat the macro BS for a given number 𝑁 of co-channel fem-tocells, we need to consider all possibilities to the vector ofdecoding events e weighted by their corresponding probabili-ties. Thus, for the general case with 𝑁 femtocells, the overalloutdoor user outage probability π’ͺ𝑏𝑠 can be written as:

π’ͺ𝑏𝑠 =βˆ‘βˆ€π‘–βˆˆΞ¦

...βˆ‘βˆ€π‘—βˆˆΞ¦

𝒫1,𝑖...𝒫𝑁,𝑗 β‹… π’ͺπ‘π‘ βˆ£e=[𝑖,...,𝑗]. (16)

2We assume that the backhaul capacity is always at least equal to therequired capacity of the indoor user. Therefore, if the indoor user is decodedat the HBS, then its interference can be removed at the macro BS.

168

TABLE IVALUES OF π‘Žπ‘˜ , π‘π‘˜ AND π›½π‘˜ FOR THE POSSIBLE DECODING EVENTS AT THE

FEMTOCELLS.

π‘’π‘˜ π‘Žπ‘˜ π‘π‘˜ π›½π‘˜

π‘œπ‘’π‘‘ 0 1 1π‘π‘œπ‘‘β„Ž 1 0 0π‘›π‘œπ‘›π‘’ 0 0 1𝑖𝑛 0 0 0

Since the cardinality (number of elements) of the set Ξ¦ is∣Φ∣ = 4, it turns out that the number of elements in (16) isgiven by βˆ£Ξ¦βˆ£π‘ . For instance, in the case of a single co-channelfemtocell (𝑁=1), (16) reduces to:

π’ͺ𝑏𝑠 =βˆ‘

βˆ€π‘–βˆˆ{π‘œπ‘’π‘‘,𝑖𝑛,π‘π‘œπ‘‘β„Ž,π‘›π‘œπ‘›π‘’}𝒫1,𝑖 β‹… π’ͺπ‘π‘ βˆ£e=[𝑖]

=𝒫1,π‘π‘œπ‘‘β„Ž β‹… π’ͺπ‘π‘ βˆ£π‘π‘œπ‘‘β„Ž + 𝒫1,π‘œπ‘’π‘‘ β‹… π’ͺπ‘π‘ βˆ£π‘œπ‘’π‘‘+ (17)

𝒫1,π‘›π‘œπ‘›π‘’ β‹… π’ͺπ‘π‘ βˆ£π‘›π‘œπ‘›π‘’ + 𝒫1,𝑖𝑛 β‹… π’ͺπ‘π‘ βˆ£π‘–π‘›.

Additionaly, the indoor user outage probability π’ͺπ‘˜,𝑖𝑛 at theπ‘˜-th femtocell is given by

π’ͺπ‘˜,𝑖𝑛 = π’«π‘˜,π‘œπ‘’π‘‘ + π’«π‘˜,π‘›π‘œπ‘›π‘’. (18)

As a reference, we also consider the case where thefemtocells are absent. In this case the outdoor user outageprobability at the macro BS is given by

π’ͺ𝑁𝐹 = Pr(log2(1 + βˆ£β„Žπ‘œπ‘βˆ£2π›Ύπ‘œ,𝑏) < π‘…π‘œ)

= 1βˆ’ exp

(βˆ’2π‘…π‘œ βˆ’ 1

π›Ύπ‘œ,𝑏

). (19)

IV. NUMERICAL RESULTS

In this section we discuss some numerical results regardingthe performance of the outdoor user at the macro BS. Weassume 𝑁 ∈ {1, 2, 3}, 𝐿 ∈ {1, 2}, π‘ž = 4 (number of wallsbetween two femtocells), π‘ž = 2 for (in)outdoor users to HBSand indoor user to BS links, and π‘…π‘œ = 𝑅𝑖 = 1 bpcu.

Fig. 2 shows the outdoor user outage probability as a func-tion of π›Ύπ‘œπ‘. We assume π‘ƒπ‘˜(𝑑𝐡) = π‘ƒπ‘œ(𝑑𝐡)βˆ’20, π‘‘π‘œπ‘ = 200 m,π‘‘π‘œπ‘˜ = 25 m, π‘‘π‘˜π‘˜ = 5 m, π‘‘π‘˜π‘ = 200 m, π‘‘π‘—π‘˜ = 50 m, backhaulcapacity 𝐢 = 2 bpcu, π›Ώπ‘˜ = 1/2 and 𝐿 = 1 (single antenna).As expected, we conclude that the outdoor user experienceslarger diversity gains as 𝑁 grows, once the HBSs act as relaysto the outdoor user. For instance, for π’ͺ𝑏𝑠 = 10βˆ’3 and 𝑁 = 1 again of around 22dB can be observed with respect to the casewithout femtocells, and more 8dB can be added if 𝑁 = 2. If𝑁 = 3 we have only an additional 3dB SNR gain, showingthat we have diminishing returns with the increase of 𝑁 .

Fig. 3 depicts the outdoor user outage probability asa function of π›Ύπ‘œπ‘ and different backhaul capacities 𝐢 ∈{1.33, 1.4, 1.5, 2} bpcu. Notice that with two co-channel fem-tocells (𝑁 = 2) with reduced backhaul capacity (𝐢 =1.5 bpcu) the outage probability is very close to that with asingle femtocell with larger backhaul capacity (𝐢 = 2 bpcu).In the case of three femtocells (𝑁 = 3) the same performancecan be obtained with an even more reduced backhaul capacity(𝐢 = 1.33 bpcu). By increasing the number of co-channelfemtocells we can decrease the backhaul capacity requirement,

βˆ’20 βˆ’10 0 10 20 30 4010

βˆ’3

10βˆ’2

10βˆ’1

100

Ξ³ob

(dB)

Obs

No femtocellN=1 C=2 bpcuN=2 C=2 bpcuN=3 C=2 bpcu

Fig. 2. Outdoor user outage probability as a function of π›Ύπ‘œπ‘ for differentnumber of single antenna co-channel femtocells.

what may also decrease the overall implementation and oper-ational costs of the femtocells. Moreover, from the figure wecan also conclude that exists a strict trade-off between 𝑁 and𝐢 once we observe a larger diversity gain when 𝐢 growsfrom 1.4 to 1.5 bpcu for 𝑁 = 2. Such gains are in the orderof several dBs (> 10dB) and are even larger if the indoorusers can afford a higher backhaul capacity. Therefore, wecan see that the achievable diversity gain is directly related tothe backhaul capacity.

The deployment of multiple co-channel femtocells canincrease the performance of the outdoor user and reduce thebackhaul capacity. The increase in performance comes fromthe increased spatial diversity that can be exploited by theoutdoor user, once the HBSs act as relays, while the reducedbackhaul capacity comes from the fact that this additionalspatial diversity is distributed. It is illustrative to compare thecase of multiple single antenna co-channel femtocells withthe case of a single femtocell with multiple antennas. Thisis done in Fig. 4, considering 𝑁 ∈ {1, 2}, 𝐿 ∈ {1, 2} and𝐢 ∈ {1.8, 1.9, 2} bpcu. As expected, a single femtocell with

βˆ’20 βˆ’10 0 10 20 30 4010

βˆ’3

10βˆ’2

10βˆ’1

100

Ξ³ob

(dB)

Obs

N=2 C=1.4 bpcuN=1 C=2 bpcuN=2 C=1.5 bpcuN=3 C=1.333 bpcuN=2 C=2 bpcu

Fig. 3. Outdoor user outage probability as a function of π›Ύπ‘œπ‘, 𝑁 , and 𝐢.

169

βˆ’20 βˆ’10 0 10 20 30 4010

βˆ’3

10βˆ’2

10βˆ’1

100

Ξ³ob

(dB)

Obs

No femtocellN=1, L=2, C=1.8 bpcuN=1, L=2, C=1.9 bpcuN=1, L=1, C=2 bpcuN=2, L=1, C=2 bpcuN=1, L=2, C=2 bpcu

Fig. 4. Outdoor user outage probability as a function of π›Ύπ‘œπ‘, 𝐿, and 𝐢.

𝐿 antennas can offer the same diversity order to the outdooruser as 𝑁 = 𝐿 HBSs with a single antenna each. Moreover,the outage probability is even smaller with a single femtocellwith 𝐿 antennas due to the array gain (similar to what happenswhen comparing MRC and selection combining). However,note that by having multiple antennas in a single femtocell weare not able to reduce the backhaul capacity requirement with-out compromising performance. This result shows the benefitof having multiple co-channel femtocells, in which case weare able to reduce the backhaul capacity requirement withoutconsiderably compromising the performance of the outdooruser. Moreover, by having multiple co-channel femtocells weincrease the area spectral efficiency.

Finally, Fig. 5 shows the indoor user outage probability asa function of π›Ύπ‘˜,π‘˜, 𝑁 , and 𝐿. The performance of the indooruser improves considerably with additional receive antennasat the HBS. However, note that the channel conditions at thefemtocell are in general much better than in the macrocell, sothat the indoor user often operates at a high SNR π›Ύπ‘˜,π‘˜. Finally,note that both this improved indoor user performance seen inFig. 5 and the savings in backhaul capacity seen in Fig. 3, can

0 10 20 30 40 50 6010

βˆ’3

10βˆ’2

10βˆ’1

100

Ξ³k,k

(dB)

Ok,

in

N=2 L=1N=1 L=2

Fig. 5. Indoor user outage probability as a function of π›Ύπ‘˜,π‘˜ , 𝑁 and 𝐿.

be concurrently achieved if we consider the use of multipleco-channel HBSs with multiple antennas.

V. CONCLUSION

We investigated the use of multiple co-channel HBSs asrelays to the outdoor user. The performance of the outdooruser considerably increases with the increase in the numberof cooperative HBSs. Additionally, capacity constraints on thebackhaul can be relaxed with the increase of the number ofcooperative HBSs given a target outage probability. As a futurework we intend to consider multiple antennas at the macro BS,as well as to investigate the case of multiple co-channel HBSswith the inclusion of dedicated relays at the femtocells, in away to improve the performance of both outdoor and indoorusers.

REFERENCES

[1] V. Chandrasekhar, J. G. Andrews, and A. Gatherer, β€œFemtocell networks:a survey,” IEEE Comm. Mag., vol.46, no.9, pp.59-67, 2008.

[2] G. de la Roche, A. Valcarce, D. Lpez-Prez, and J. Zhang, β€œAccess controlmechanisms for femtocells,” IEEE Comm. Mag., vol.48, no.1, pp.33-39,2010.

[3] V. Chandrasekhar, J. G. Andrews, T. Muharemovic, Z. Shen, A.Gatherer, β€œPower control in two-tier femtocell networks,” IEEE Trans.Wireless Commun., vol.8, no.8, pp.4316-4328, 2009.

[4] V. Chandrasekhar, M. Kountouris, and J. G. Andrews, β€œCoverage inmulti-antenna two-tier networks,” IEEE Trans. Wireless Commun.,vol.8,no.10, pp.5314-5327, 2009.

[5] D. Lopez-Perez, A. Valcarce, G. de la Roche, and J. Zhang, β€œOFDMAfemtocells: a roadmap on interference avoidance,” IEEE Comm. Mag.,vol.47, no.9, pp.41-48, 2009.

[6] V. Chandrasekhar and J. G. Andrews. β€œSpectrum allocation in two-tiernetworks”. IEEE Trans. Commun., vol.57, no.10, pp.3059-3068, 2009.

[7] G. Gur, S. Bayhan, and F. Alagoz,β€œCognitive femtocell networks: anoverlay architecture for localized dynamic spectrum access”. IEEEWireless Commun., vol.17, no.4, pp.62-70, 2010.

[8] S.-M. Cheng, S.-Y. Lien, F.-S. Chu, and K.-C. Chen, β€œOn exploitingcognitive radio to mitigate interference in Macro/femto heterogeneousnetworks,” IEEE Wireless Comm. Mag., vol.18, no.3, pp.40-47, 2011.

[9] Y.-Y. Li, E.S. Sousa, β€œCognitive uplink interference management in 4Gcellular femtocells,” IEEE 21st International Symposium on PersonalIndoor and Mobile Radio Communications, pp.1567-1571, 26-30 Sept.2010

[10] A. Hayar, B. Zayen, S. Bouferda, β€œInterference management strategyfor deploying energy efficient femtocells network,” 7th InternationalICST Conference on Cognitive Radio Oriented Wireless Networks andCommunications, pp.208-211, 18-20 June 2012

[11] Y.-Y. Li, M. Macuha, E.S. Sousa, T. Sato, M. Nanri, β€œCognitiveinterference management in 3G femtocells,” IEEE 20th InternationalSymposium on Personal, Indoor and Mobile Radio Communications,pp.1118-1122, 13-16 Sept. 2009

[12] J. Jin, B. Li, β€œCooperative Resource Management in Cognitive WiMAXwith Femto Cells,”2010 Proceedings IEEE INFOCOM, pp.1-9, 14-19March 2010

[13] J. N. Laneman, D. N. C. Tse and G. W. Wornell, β€œCooperative diversityin wireless networks: efficient protocols and outage behavior,”IEEETrans. Inform. Theory, vol.50, no.12, pp.3062-3080, 2004.

[14] R. Urgaonkar, M.J. Neely, β€œOpportunistic cooperation in cognitivefemtocell networks,” IEEE J. Sel. Areas Commun., vol.30, no.3, pp.607-616, 2012.

[15] T. Elkourdi and O. Simeone, β€œFemtocell as a relay: an outage analysis,”IEEE Trans. Wireless Commun., vol.10, no.12, pp.4204-4213, 2011.

[16] C. H. M. de Lima, M. Bennis, K. Ghaboosi and M. Latva-aho,β€œInterference management for self-organized femtocells towards greennetworks,” IEEE 21st PIMRC, pp.352-356, 26-30 Sept. 2010.

[17] H. Alves, M. Bennis, R. D. Souza, M. Latva-Aho, β€œEnhanced per-formance of heterogeneous networks through full-duplex relaying”.EURASIP Journal on Wireless Communications and Networking, 2012.

[18] R. Narasimhan, β€œIndividual outage rate regions for fading multipleaccess channels,” IEEE ISIT, pp.1571-1575, 24-29 Jun. 2007.

170