Hybrid spectrum access model for cognitive femtocell networks

Zuo Xuzhou Liu Jishun Xia Weiwei Shen Lianfeng

(National Mobile Communications Research Laboratory, Southeast University, Nanjing 210096, China)

There has been increasing interest in deploying femtocells in a long term evolution (LTE) system. A femtocell (home base-station) has a coverage of 10 to 50 m and it is usually placed indoors. Due to a dense spatial reuse of the available radio spectrum, using femtocells can greatly improve the capacity and coverage of the networks[1]. However, as femtocells reuse the radio resources of macrocells, they may cause cochannel interference (CCI) to macrocell users. Interference decreases the network capacity and results in performance degradation. Handling the interference is one principal challenge for femtocell usage[2].

Cognitive radio (CR) techniques are proposed to alleviate the interference problem, which allow femtocells to sense the radio resource utilization of macrocells and autonomously adapt to the wireless environment and utilize the spectrum so as to avoid interference to macrocells[3].

After spectrum sensing, femtocells can employ different approaches to utilize the spectrum. One is for the femtocells to opportunistically access the spectrum only when it is not occupied by the macrocell. This approach is called spectrum overlay[4-7]. The other is that the femtocells can always access the spectrum if the interference caused to the macrocell users is under a pre-determined threshold. This approach is called spectrum underlay[8-11].

Both approaches have their merits and drawbacks. For spectrum overlay, the interference caused by femtocells to macrocells can be maintained at a low level if the sensing performance can be guaranteed. But the available amount of spectrum for femtocells depends on the spectrum usages of the macrocell. For spectrum underlay, a femtocell may have more available amount of spectrum, but it may cause higher interference to macrocells. Besides, it requires instantaneous knowledge of interference channels.

A novel dynamic access model is introduced to the femtocell in a two-tier environment taking advantage of the above two CR strategies. In this model, the femtocell flexibly operates in hybrid underlay/overlay spectrum access mode. Spectrum sensing is the first step to detect the interference conditions in a specified band. Then, femtocell users dynamically adjust their access mode based on the sensing results.

This model is motivated by the hybrid automatic repeat request (HARQ) transmission scheme of the LTE network discussed in Ref.[12]. Spectrum sensing is performed accordingly in two stages: the principal sensing to detect the macrocell base-station (evolved node-B (eNB)) downlink transmission, and the assisted sensing to detect the corresponding user equipment (UE) uplink acknowledgement.

The favorable usage mode is selected according to the sensing results in the two stages. The sensing results reflect the link quality with actual pathloss and relevant spatial conditions between the femtocell and macrocell users. By introducing the hybrid spectrum access model to handle the cross-tier interference, the interference recognition can be well managed and higher capacity can be provided via exploiting the spectrum opportunities efficiently. Simulations show that the throughput of the femtocell is greatly improved with tolerable interference to macrocell users.

1 Network Model

1.1 Systemconfiguration

Fig.1 shows a two-tier LTE network of time division duplex (TDD) type. In the circular macrocell, the UE is served by a high-power eNB, which is usually located at the center with an omnidirectional antenna. The macrocell is overlaid with random distributed femtocells, which has very small signal coverage. For each femtocell, there is a femtocell base station (home evolved node-B (HeNB)) located at the center and multiple home user equipment (HUE) which is connected to the HeNB. The HeNB is in charge of infrastructure functionalities such as service provisioning and profile management for the HUE.

In the macrocell, the transmission is frame-based. As shown in Fig.1, each frame is partitioned into 10 successive TDMA channels. Such a channel is the unit for radio resources scheduling[12]. The channel is used for either uplink (UL) transmission or downlink (DL) transmission.

Fig.1 Infrastructure-based macrocell and multiple overlaying femtocells in a two-tier TDD-LTE network

An important feature of a cellular wireless network is its asymmetry. The available bandwidth for DL transmission is much more than that for UL transmission. So, in this paper we consider the DL transmission of the macrocell.

In the macrocell, HARQ is utilized for transmission, that is, the transmission of a data packet is followed by an acknowledgement from the receiver. As shown in Fig.1, the DL data transmission of the macrocell is a two-step process between an eNB and its corresponding UE. First, the eNB transmits data packets to the UE on the DL channel. Then, the UE replies an acknowledgement on the UL channel.

Both the macrocells and the femtocells are based on the same communication technology and share identical spectra. The macrocell may suffer from interference if the channel they occupied is being used by a nearby femtocell. As a result, the HeNB shall avoid scheduling the channels occupied by the nearby macrocell to prevent interference.

By incorporating the cognitive capacities, the HeNB can sense its operating environment. Thus, the operation of the femtocell can be adaptively modified by using the radio resource of the macrocell. According to the concept of CR, the macrocell is considered as the primary user, and the femtocell is considered as the secondary user.

It is assumed that neighboring cells follow a synchronized slot structure. In other words, the coupling of the UL/DL channels at the macrocell is known by the femtocells, which enables them to listen to the data sent by the macrocell users in their respective UL and DL channels. The sensing is performed during the guard interval at the beginning of each channel.

1.2 Energy detection

We focus on energy detection in this paper, since it can be simply implemented and used without prior knowledge of primary signal structures.

Before the input signal is applied to the energy detector, it is filtered by the bandpass filter with a bandwidth ofW. Lety(t) denote the resulting bandpass signal. The energy detector makes a decision between the following two hypotheses,

H0:y(t)=n(t) 0

H1:y(t)=hs(t)+n(t) 0

(1)

whereTdenotes the observation time;s(t) is the transmitted signal from the primary transmitter;n(t) is the zero-mean additive white Gaussian noise (AWGN) with varianceσ2; andhis the amplitude gain of the channel. The energy detector squaresy(t) and integrates it over the interval ofT. Then, the decision statisticYis compared with the thresholdλ. IfY>λ, the primary signal is determined to be present; otherwise, to be absent.

(2)

Following the work of Ref.[8],Yis shown to have the following distribution,

(3)

The probabilities of detection and false alarm can be represented as[13]

(4)

whereΓ(·) andΓ(·, ·) are the complete and incomplete gamma functions, respectively;Qm(·, ·) is the generalized MarcumQ-function defined as

(5)

whereIm-1(·) is the modified Bessel function of the (m-1)-th order.

For simplicity, we ignore channel fading and the path loss is calculated based on the basic path loss model. The propagation loss can be expressed asL(d)=d-α, wheredrepresents the distance andαdenotes the path loss factor which is a fixed known constant.

2 Hybrid Spectrum Access Model

2.1 Overview

In order to efficiently utilize the spectrum, we present a hybrid spectrum access model based on spectrum sensing. Using this model, a femtocell can choose its transmission mode when accessing the spectrum.

In this model, the HeNB monitors its operating radio environment and searches for the available channels by periodical sensing. Taking advantage of the HARQ transmission scheme of macrocell users, an HeNB performs principal and assisted sensing in the DL channel and the corresponding UL channel, respectively, and detects whether there are any macrocell users in the vicinity.

Normally, channels are available for the femtocell when no primary transmission is detected, which corresponds to the case of both principal and assisted sensing indicating the absence of primary transmission. On the other hand, channels are unavailable for the femtocell when primary transmission is detected, which corresponds to the case of both principal and assisted sensing indicating the presence of primary transmission. In both cases, the sensing results are the same.

However, when the sensing results of the two cases are different, the femtocell is switching to the underlay mode. The first case is that the principal sensing in the DL channel detects the signal of the eNB while the assisted detection in the UL channel does not detect any UE.

The probability of this case, under the condition that primary transmission is on, is

P1=p{Y1<λ1,Y2>λ2|H1}=

(6)

whereY1andY2are the decision statistics of the two phases; andλ1andλ2are the predefined sensing thresholds of the two phases, respectively;γ1andγ1are the SNR of the eNB signal in the DL channel and the SNR of the UE signal in the UL channel, respectively.

Then, the SNR of the eNB signal and the UE signal can be, respectively, represented as

(7)

(8)

whereQeNBandQUEare the transmitting power of teh eNB and the UE, respectively; andd1andd2are distances from the femtocell users to the eNB and the correspond UE, respectively. Because the coverage of the femtocell is small, we consider that the distances of femtocell users to the eNB and the UE are the same.

Substituting Eqs.(7) and (8) into Eq.(6), we can obtain a relationship betweenP1andd1,d2for the givenQeNBandQMUEas

P1=f(d1,d2)

(9)

As shown in Fig.2, this is mainly due to the fact that a femtocell is located close to the eNB and far from the operating UE. In this case, the femtocell is unable to detect the UE signal. Meanwhile, the further the femtocell is located from an operating UE, the weaker the DL interference the UE suffers. In this situation, the femtocell can access the DL channel in an underlay way; that is, the femtocell can utilize the DL channel as long as the interference to the UE is under its tolerable interference threshold.

Fig.2 Spectrum sensing in DL and UL channels

The second case is that the principal sensing in the DL channel does not detect the signal of the eNB while the assisted detection in the UL channel detects a nearby UE. As shown in Fig.2, this is mainly due to the fact that the femtocell is located far from the eNB and located close to the operating UE. In this case, the femtocell can access the UL channel in an underlay way, which means that the interference to the eNB is under its tolerable interference threshold.

Based on the information obtained from spectrum sensing, the femtocell can flexibly operate in an underlay or overlay mode. It is normally working in an overlay mode if the sensing results in both stages are the same, and it switches to the underlay mode when the sensing results in both steps are varied. Both cases will be discussed in the following.

2.2 Overlay mode

When primary users are transmitting, the probability that the sensing results in both phases indicates that the presence of primary transmission can be expressed as

Pde=p{Y1>λ1,Y2>λ2|H1}=

(10)

When primary users are not transmitting, the probability that the sensing results in both phases indicate the presence of primary transmission can be expressed as

Pfa=p{Y1>λ1,Y2>λ2|H0}=

(11)

When a femtocell fails to detect the primary signal, it will transmit in the channel simultaneously. In this case, the average throughput of the femtocell can be expressed as

(12)

When primary transmission is off and there is no false alarm in detection, the average throughput of the femtocell can be expressed as

(13)

whereWis the spectrum bandwidth;QSUis the transmitting power of the femtocell user; andris the distance from the femtocell users to the eNB and the corresponding HUE, which is betweenrmaxandrmin. Because the distribution of HUE within the femtocell is random uniform, being subject to the maximum and minimum separation constraints.

(14)

The average throughput of the femtocell in the overlay mode is

Toverlay=Pp(1-Pde)Toverlay-1+(1-Pp)(1-Pfa)Toverlay-2

(15)

wherePpis the probability that primary transmission is present.

The coupling between UL and DL channels is of great importance for the implementation of two phase sensing. Compared with traditional single phase sensing, the proposed model improves the reliability of sensing due to the introduction of the combination of two phases. Meanwhile, compared with collaborative sensing among multiple users, it does not need signaling overhead.

2.3 Underlay mode

To maintain enough available spectra for usage of a femtocell, it is necessary for the femtocell to access a spectrum at times despite the presence of primary transmission. The selection of the underlay strategy and interference management must be carefully investigated.

Femtocells switch to the underlay mode when the sensing results in both phases are varied.

2.3.1 The first case

The first case is that the principal detection in the DL channel indicates absence of the primary signal while the assisted detection in the UL channel indicates presence of the primary signal.

As shown in Fig.3, in this situation, the femtocell can access the UL channel in an underlay way. But note that the femtocell must carefully evaluate the potential interference in order not to cause unacceptable interference to the eNB. That is, the interferenceRsfrom the femtocell to the eNB must be beyond a thresholdrb, andrbis determined by the eNB’s reception sensitivity and QoS.

Fig.3 Channel usage and interference scenario for the two-tier network

The sensing thresholdλbis used for the HeNB to measure the link between itself and the eNB and make a decision for access or not. If decision statisticY1<λb, the femtocell can access the UL channel. On the contrary, if the decision statisticY1>λb, the femtocell cannot access the UL channel.

If the femtocell utilizes the UL channel, the interference of the secondary signal to the eNB can be expressed as

(16)

When a primary transmission is present, the probability that any femtocell user transmits in the UL channel can be expressed as

Punderlay-11=P{Y1<λb,Y1<λ1,Y2>λ2|H1}

(17)

In this case, the average throughput of the femtocell can be expressed as

(18)

Then the probability that the interference to the eNB is beyond the threshold when the femtocell uses the UL channel can be expressed as

Pmb=P{R5>Rb|Y1<λb,Y1<λ1,Y2>λ2|H1}=

(19)

When primary transmission is absent, the probability that any femtocell user transmits in the UL channel can be expressed as

Punderlay-12=P{Y1<λb,Y1<λ1,Y2>λ2|H0}

(20)

In this case, the average throughput of the femtocell can be expressed as

(21)

The average throughput of the first case is

Tunderlay-1=PpPunderlay-11Tunderlay-11+(1-Pp)Punderlay-12Tunderlay-12

(22)

2.3.2 The second case

The second case is that the principal detection in the DL channel indicates the presence of the primary signal while the assisted detection in the UL channel indicates the absence of the primary signal.

As shown in Fig.3, in this situation, femtocells may occupy the DL channel in an underlay way. But note that the femtocell must carefully evaluate the potential interference in order not to cause unacceptable interference to the UE. That is, the interferenceRufrom the femtocell to the UE must be beyond a thresholdru, which is determined by the UE’s reception sensitivity and QoS.

The sensing thresholdλuis used for the HeNB to measure the link between itself and the eNB and make a decision for use the channel or not. If decision statisticY2<λu, the femtocell can utilize the DL channel. On the contrary, if the decision statisticY2<λu, the femtocell cannot utilize the DL channel.

If the femtocell utilizes the DL channel, the interference of the secondary signal to the UE can be calculated as

(23)

When a primary transmission is present, the probability that any femtocell user transmits in the DL channel can be expressed as

Punderlay-21=P{Y2<λu,Y1<λ1,Y2<λ2|H1}

(24)

In this case, the average throughput of the femtocell can be expressed as

(25)

Then the probability that the interference to the UE is beyond the threshold when the femtocell utilizes the DL channel is

Pine=P{Ru>ru|Y2<λu,Y1<λ1,Y2>λ2|H1}=

(26)

When a primary transmission is absent, the probability that any femtocell user transmits in the DL channel can be expressed as

Punderlay-22=P{Y2<λu,Y1<λ1,Y2<λ2|H0}

(27)

In this case, the average throughput of the femtocell can be expressed as

(28)

The average throughput of the femtocell of the second case is given by

Tunderlay-2=PpPunderlay-21Tunderlay-21+(1-Pp)Punderlay-22Tunderlay-22

(29)

Finally, the average throughput of the femtocell is given by

T=Toverlay+Tunderlay=Toverlay+Tunderlay-1+Tunderlay-2

(30)

In this model, the key for success is to keep interference caused by femtocells low enough to ensure a low impact on the performance of an existing macrocell, but there are enough channels for femtocells to achieve high data rates. In the next section, the performance of this model is investigated by system-level simulations.

3 Simulation Results

In this section, the throughput performance of our proposed hybrid access model is evaluated. We also compare the proposed scheme with the conventional cognitive-based overlay scheme.

The simulation parameters are summarized in Tab.1, which are in line with 3GPP LTE specifications. In our simulation, the overlaying LTE network consists of macrocells and femtocells. We consider one macrocell with a 500 m cell radius, where 10 femtocells and 50 UEs are uniformly distributed in the center macrocell. The network operates in the 2 GHz band with a system bandwidth of 5 MHz. As shown in Tab.1, macrocell users transmit at an equal fixed power level and have an omni-directional antenna pattern with antenna gain. Each macrocell sector has 10 UEs associated with it on average. UEs are considered to be uniformly distributed in the system. 10 femtocells are uniformly distributed in each macrocell. HUEs are uniformly distributed in a finite circular area around each femtocell (The cell radius is 20 m). Each of the femtocells operates at an equal fixed power level and an omni-directional antenna pattern with antenna gain.

In the simulation, we adopt the Okumura-Hata model to calculate outdoor path loss and use the COST-231 multi-wall model for indoor path loss. When considering outdoor UEs with respect to a femtocell, a penetration loss of 20 dB is assumed. A closed-access policy is assumed, where only an authorized set of UEs can be associated with a femtocell. In particular, the full load is assumed for each femtocell; thus, each HeNB will select as many channels as possible.

We compare the performance of three schemes. The first scheme is to combine the sensing results in the two phases according to the OR rule. The second scheme is to combine the sensing results in the two phases according to the AND rule. The third scheme is to implement our hybrid model.

In Fig.4, the throughput of the femtocell of the three schemes with various primary transmission probabilitiesPpis shown. We can see a clear throughput improvement in our scheme using the hybrid model compared with the scheme using the OR rule, and only a slight throughput decrease in our scheme compared with the scheme using the AND rule. This is because the AND rule scheme utilizes the spectrum in an aggressive way while the OR rule scheme utilizes the spectrum in a conservative way, and the hybrid model utilizes the spectrum according to the spectrum usage and related spatial conditions of macrocell users. Furthermore, we can see that the interference probability of femtocell to macrocell users in Fig.5 when one time sample is 100 frames andPpis 0.5. The AND rule scheme causes much more interference than the other two schemes.

From Fig.4 and Fig.5, it is shown that the hybrid model can achieve a tradeoff between the conservative way and the aggressive way. By introducing the hybrid spectrum access model to handle the cross-tier interference, the interference recognition can be well managed and higher capacity can be provided via exploiting the spectrum opportunities efficiently. The throughput of the femtocell is greatly improved with tolerable interference to macrocell users.

Fig.4 Average femtocell throughput according to Pp of the three schemes

Fig.5 Interference probability of femtocell to macrocell users

4 Conclusion

In this paper, we present a hybrid underlay/overlay access model for femtocells in the two-tier LTE network. In this model, spectrum sensing is performed in DL channels and the corresponding UL channels, respectively. The favorable usage mode is selected according to the sensing results. Simulation results show that the throughput of the femtocell is greatly improved with tolerable interference to the macrocell.

[1]Chandrashekhar V, Andrews J. Femtocell networks: a survey [J].IEEECommunicationsMagazine, 2008,46(9): 59-67.

[2]Yavuz M. Interference management and performance analysis of UMTS/HSPA+femtocells [J].IEEECommunicationsMagazine, 2009,47(9): 103-108.

[3]Lien S Y, Tseng C C, Chen K C, et al. Cognitive radio resource management for QoS guarantees in autonomous femtocell networks [C]//IEEEInternationalConferenceonCommunications. Cape Town, South Africa, 2010:5502784-1-5502784-6.

[4]Gur G, Bayhan S, Alagoz F. Cognitive femtocell networks: an overlay architecture for localized dynamic spectrum access [J].IEEEWirelessCommunications, 2010,17(8): 62-70.

[5]Meerja K A, Ho P H, Wu B. A novel approach for co-channel interference mitigation in femtocell network [C]//GlobalTelecommunicationsConference. Houston, USA, 2011:6133956-1-6133956-6.

[6]Zhao G, Ma J, Li Y, et al. Spatial spectrum holes for cognitive radio with directional transmission [J].IEEETransactionsonWirelessCommunications, 2009,8(10): 5270-5279.

[7]Chen Y, Zhao Q, Swami A. Joint design and separation principle for opportunistic spectrum access in the presence of sensing errors [J].IEEETransactionsonInformationTheory, 2008,54(5): 2053-2071.

[8]Oh D C, Lee H C, Lee Y H. Cognitive radio based femtocell resource allocation [C]//InternationalConferenceonInformationandCommunicationTechnologyConvergence. Hangzhou, China, 2010: 274-279.

[9]Huang J W, Krishnamurthy V. Cognitive base stations in LTE/3GPP femtocells: a correlated equilibrium game-theoretic approach [J].IEEETransactionsonCommunications, 2011,59(12): 3485-3493.

[10]Lien S Y, Tseng C C, Chen K C, et al. Cognitive radio resource management for QoS guarantees in autonomous femtocell networks [C]//IEEEInternationalConferenceonCommunications. Cape Town, South Africa, 2010:5502784-1-5502784-6.

[11]Kaimaletu S, Krishnan R, Kalyani S, et al. Cognitive interference management in heterogeneous femto-macro cell networks [C]//IEEEInternationalConferenceonCommunications. Kyoto, Japan, 2011:5962617-1-5962617-6.

[12]3GPP. TS 36.213 Evolved universal terrestrial radio access (E-UTRA): physical layer procedures (Release 10) [S]. 3rd Generation Partnership Project,2010.

[13]Digham F F, Alouini M S, Simon M K. On the energy detection of unknown signals over fading channels [C]//IEEEInternationalConferenceonCommunications.Anchorage, USA, 2003: 3575-3579.