This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2011 proceedings.

A Novel Spectrum Handoff Scheme With Spectrum Admission Control in Cognitive Radio Networks Chengyu Wu, Chen He, Lingge Jiang

Yunfei Chen

Department of Electronic Engineering Shanghai Jiao Tong University, China Email: {jerry916, chenhe, lgjiang}@sjtu.edu.cn

School of Engineering University of Warwick Coventry, UK, CV4 7AL Email: [email protected]

Abstract—In this paper, we propose a new spectrum handoff scheme with spectrum admission control (SAC). In the proposed scheme, the secondary users (SUs) make up secondary user groups (SUGs) to achieve appointed detection probability of primary user signals, and perform spectrum handoff to an available spectrum when the primary users (PUs) reuse the spectrum in cognitive radio networks (CRNs). A simple Markov model is adopted to analyze the performance of spectrum handoff in terms of blocking probability, forced termination probability and throughput of the cognitive radio system. Numerical results show that the new handoff strategy is suitable for multi-user cognitive radio systems, and that spectrum admission control and cooperative sensing can effectively increase the efficiency of spectrum handoff.

I. I NTRODUCTION Cognitive radio (CR) improves the usage efficiency of the under-utilized radio spectrum, and this has motivated researchers to study CR. One of the most important functionalities in CR is spectrum mobility [1]. To achieve spectrum mobility, spectrum handoff is required. In a cognitive radio network with both primary users (PUs) and secondary users (SUs), the PUs do not always occupy their frequency bands. Consequently, the SUs may temporarily occupy the frequency bands which are not occupied by the PUs. However, the SUs must vacate these frequency bands when the PUs come back and reuse them. In this case, the communication links of the SUs established over the PUs’ frequency bands will be lost and therefore, the SUs must find another idle frequency bands to resume their communication links. This is called spectrum handoff [2]. Spectrum handoff is less investigated in the literature compared with spectrum sensing. Most of the literature research the spectrum handoff is applying a Markov chain model to analyze the performance. In [2], [3], a Markov chain model is proposed to analyze the performance of spectrum access, and then a channel reservation scheme for spectrum handoff is designed to make tradeoff between the forced termination probability and the blocking probability. A three-dimensional Markov chain model is developed in [4] to analyze and study the cognitive radio performance given certain primary traffic, and also a channel reservation scheme for the primary system is considered to reduce the forced termination probability. And some models for spectrum handoff is based on queuing theory. A queuing system is designed to maximize the system

throughput under maximum-delay quality of service (QoS) constraints in the primary activity considered in [5]. In [6], the authors propose a novel priority virtual queue interface to analyze the multimedia users’ interactions and determine the required information exchange according to the priority queuing analysis. A preemptive resume priority (PRP) M/G/1 queuing network is developed in [7] to analyze the reactiveor proactive-sensing spectrum handoff considering the sensing time. In order to maintain a given QoS requirement for SUs, a proper spectrum admission control (SAC) scheme, which is known as an effective functionality of the media access control (MAC) layer in ensuring the QoS of wireless networks, should be integrated into the MAC layer. Since the number of available frequency bands always changes dynamically due to the random variations of the PU traffic, SAC is an indispensable part of a CR network to improve the flexibility and to satisfy the QoS of the SUs’. In [8], a soft-QoSbased spectrum handoff mechanism is integrated into SAC by using the Markov chain model to describe the procedures of SAC and spectrum handoff. A nonlinear optimization is formulated to minimize the dropping probability with the constraint of blocking probability. In [9], a fractional guard channel reservation scheme is applied to tradeoff between the blocking probability of SUs and the forced termination probability. In this paper, a new spectrum handoff scheme with SAC is proposed for cognitive radio networks (CRNs) by introducing cooperation among SUs into sensing to reduce the probability of missed detection and the probability of false alarm. Also, we develop secondary user group (SUG) as a unit to access the available spectrum, the members of which can cooperate with each other to prevent hidden-terminal problem and to increase the detection probability. A suitable approach for SUGs’ spectrum access is also developed. Based on this, a Markov model is proposed to analyze the performance of the spectrum handoff in terms of blocking probability, forced termination probability and system throughput. Numerical results show that the new spectrum handoff scheme is ideal for multi-user cognitive radio systems, the approach of SAC for SUGs can effectively improve the system performance. The remainder of the paper is organized as follows. In Section II, the system model is described along with the

978-1-4244-9268-8/11/$26.00 ©2011 IEEE

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2011 proceedings.

1 SU1 SU7 SUG1

Others

2

RCTS Handoff Sense Handoff

SU 1 RRTS

3

SU 4

SU4

SUG2

SU6

ĂĂ

SU2

λp

CCC

MN-3

SU 1 DATA

SU 4

MN-1 MN

ACK

RCTS Handoff Sense Handoff

OR Handoff RRTS

DATAPeriodic

Handoff

Sensing

Handoff

Handoff Sense Handoff OR CRTS Handoff RTS

RCTS Handoff Sense Handoff CCC

Channel l

Handoff CCC

Fig. 2: SUG’s Shaping and Spectrum Handoff

•

•

•

II. S YSTEM M ODEL

As shown in Fig. 1, we assume that all members of each SUG are single radio devices, and they have the same outband common control channel (CCC) to communicate with each other. An SUG may consist of two or more SUs, which may join or leave the group at any time. In general, an SUG is shown as Fig. 2(1) and performs as follows:

DATA Channel j

CCC

(2) SUG's Spectrum Handoff for PU's activity with Periodic Sensing

Fig. 1: System Model

A. SUGs’ Shaping

Periodic Sensing

Handoff

Channel j

SU9

In the considered CR system, the PUs do not need to know anything about the SUs, so that the primary system should not be modified due to the activity of secondary users. We assume that a channel is a basic spectrum unit, and that the spectrum consists of M parallel primary bands with each band divided into N channels. Each band is allocated for a PU, each channel is allocated for an SU, and PUs have priority to use the channels over SUs. We assume the arrival of PUs is Poisson process with arrival rate λp , and the corresponding service time is exponentially distributed with rate μp . And we define the packets arrival of SUs as a distributed Poisson process with rate λs , the duration time as exponentially distribution with parameter μs , and the traffic load perform as η = λs /μs . The observation range of one single SU is small and typically less than its transmission range. So the SUs may cause interferences to the PUs inside their transmission ranges even though the SUs consider the channels available. In this case, we denote a group of SUs as an SUG [10] which could be cooperative with each other to prevent hidden-terminal problem and to increase the sensing accuracy.

CTS

CACK

(1) SUG's Shaping

MN-2

considered procedures and implementation schemes. Subsequently, in Section III, the cooperation spectrum sensing scheme is briefly introduced along with the main hypothesis and considerations, and the Markov model is illustrated and analyzed with a number of relevant performance metrics. The numerical results and discussions are given in Section IV. Finally, Section VI concludes the paper with some final remarks and future considerations.

DATA

PU Active Periodic Sensing

Others

SUG3 SU8

Handoff

Channel j

PU 1

SU5

SU3

Handoff Sense Handoff OR CRTS Handoff RTS

RCTS Handoff Sense Handoff

4 SUG Base

Handoff

•

•

•

•

Step 1: Initialization, the SUs sense channels and report the sensing results to the SUG base station, and SUG base station initializes the channels usage statistics. Step 2: Polling, SU 1 that want to communicate with SU 4 sends broadcast RRTS (Ready Request to Send, including the channel number j to be sensed) messages in the CCC. Step 3: Reporting, other SUs (including the specified SU 4), which receive the RRTS messages, sense the appointed channel and send RCTS (Ready Clear to Send, including the sensing result) messages to SU 1, SU 1 would not send ‘Reject’ messages to the certain SUs until the number of received RCTS at SU 1 up to δ, which denotes the optimal number of members in an SUG. Step 4: Decision, SU 1 merges the results with ‘OR’ rule to decide to access the channel or not, and report the results to the SUG base station. If not, go to Step 2. Step 5: Handoff , SU 1 sends CRTS (Control Request to Send) messages in CCC, and all the members of the SUG will receive CRTS and perform spectrum handoff to channel j, and also SU 1 switch to channel j. Step 6: Transmitting, SU 1 sends RTS (Request to Send) messages on channel j, SU 4 will receive RTS and send CTS(Clear to Send) back to establish communication with SU 1. The other SUs of the SUG continue to sense inband on channel j periodically. At the end, SU 4 sends ACK (ACKnowledge Character) if it completely receives the data. Step 7: Idle, SU 1 receives ACK , broadcasts CACK (Complete ACKnowledge Character) on channel j, and all the members of the SUG perform spectrum handoff to CCC after receiving CACK. Finally, go to Step 2.

B. Spectrum Handoff with Spectrum Admission Control The SUG should perform spectrum handoff to guarantee the QoS of PU in case a PU starts to transmit, so periodic sensing is necessary. As shown in Fig. 2(2), all the members of the SUG do spectrum handoff to CCC and sense channels to achieve an available channel when the SUG detects PU is active with periodic sensing. The sensing period determines

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2011 proceedings.

the maximum time during which the SUG will be unaware of a reappearing PU and hence may harmfully interfere with it. Designing an admission control scheme for SUGs is able to limit both the forced termination probability of SUGs and the interference probability upon PUs from SUGs. We define α ∈ [0, M N ] as the admission control threshold, considering the SUGs which perform spectrum handoff have priority over the new arrival SUGs to access the reservation channels. An admission decision is taken as follows: >α accept (1) M N − b(x + e2 ) ≤α reject

′+ (x 1

x2 μ s

μ 1)

p

ρ Qd p (x, κ )λ p

x1′, x2 , x3

ϕ (1-Q f )λs

) λp

Q

x1′ + 1, x2 − κ , x3

ϕ(

fo

rm

ρQ

d

p

=1

,κ (x

f

)

ϕ (1-Q f ) λ s

Qf 1-

)λ s

x1′, x2 − m, x3 − m

p(

x,

)

λ

And the detection probability of the individual SUs in the SUG is given as √ 1 · Q−1 (Pf ) − N 1 γ) Pd = Q( √ 2γ + 1 (5) √ 1 δ ˜ f ) − N 1 γ) = Q( √ · Q−1 (1 − 1 − Q 2γ + 1 where γ is the PU’s signal to noise ratio (SNR) received at the cooperative SU, Q(·) is the complementary cumulative distribution function of the standard Gaussian, and N1 = τs B is the time-bandwidth, where τs is the sensing duration and B is the bandwidth of channel j. The total Qd of the SUG is derived by substituting (5) into (2). The optimum number δ (less than the number of all SUs which denotes as n0 ) of SUs in an SUG to obtain the accredited Qd is computed from (4), (5) and (2).

ρ(

d

κ

Ideally, we would like to have Qd = 1.0 and Qf = 0.0, but these will be difficult to achieve in the practical situation. It is more reasonable to allow Qd = 0.9 ∼ 0.95 and Qf = 0.01 ∼ 0.1 [12]. In the ‘OR’ fusion rule, to achieve a targeted ˜ f , the false alarm probability of the false alarm probability Q individual SUs in the SUG is given as δ ˜f Pf = 1 − 1 − Q (4)

( x2 + 1) μ s

Q 1−

ϕ

(3)

x1′, x2 + 1, x3 + 1

x1′ + 1, x2 , x3 + m

( x2 + 1) μ s

Qf = 1 − (1 − Pf )

ϕ (1-Q f ) ⋅ λ s

(2)

δ

p x′1μ

Qd = 1 − (1 − Pd )δ

λp

A. SUG’s Cooperation Assumed that the detection probability for a channel of an SU is Pd , and the optimal number of cooperative SUs in an SUG is δ. The target detection probability and false alarm probability of the SUG for cooperative sensing using the OR fusion rule with energy detector is given as [12], [13]:

d

III. A NALYSIS

x1′ − 1, x2 , x3 − m ρQ

We define x = (x1 , x2 ) as the system vector, where x1 denotes the number of PUs that SUGs observed and x2 stands for the number of in-service SUGs. Considering the presence of sensing error(miss detection or false alarm), the actual number of PUs may be x1 = x1 + 1 or x1 = x1 − 1 [11]. So we adopt x = (x1 , x2 , x3 ) as the actual system vector, x3 denotes the number of interfered channels [9]. Let b(x) = N x1 +x2 +x3 denote the number of occupied channels at state x, and ei be a triple dimensional vector with position i set to 1 and other positions set to 0.

x1′, x2 − 1, x3

p

x1′, x2 − κ , x3

x1′, x2 + 1, x3 Fig. 3: Markov Model

B. Analysis with Markov Model To analyze the performance of spectrum handoff with SAC for SUGs, we develop a simple Markov model shown as Fig. 3. P r(H0 ) and P r(H1 ) denote the probability of the hypotheses H0 (noise only, denoting PU’s signal absence) and H1 (where the PU’s signal is present) respectively, and we denote ρ = P r(H1 ), ϕ = P r(H0 ), and ρ+ϕ = 1. Considering the detection probability and the false alarm probability of an SUG, we classify the event that trigger the transmission of the system as follows: 1) PU arrival, there are three situations for a PU arrival: • The SUG senses the arrival of PU, and the PU is actually there. The PU arrival will terminate κ SUGs, and trigger state(x1 , x2 , x3 ) ⇒ (x1 + 1, x2 − κ, x3 ) shown as Fig. 3, κ ∈ [0, min(x2 , N )] and the transmission rate of a new PU arrival with termination κ SUGs is ρQd · p(x, κ)λp , where p(x, κ) is given by N (M −x −1)N p(x, κ) =

κ

1

x −κ

(M −x2 )N x2

1

(6)

The SUG detects the arrival of PU, but the PU is not there. This is a false alarm, the results also would terminate κ SUGs, state (x1 , x2 , x3 ) ⇒ (x1 , x2 − κ, x3 ) with transmission rate ϕQf · p(x, κ)λp . • The SUG does not detect the arrival of PU, but the PU is there. This is a missed detection, state (x1 , x2 , x3 ) ⇒ (x1 +1, x2 , x3 +m) with transmission rate ρ(1−Qd )·λp , where m ∈ [0, min(x3 , N )]. 2) PU departure, there are two situations for a PU departure: • State (x1 , x2 , x3 ) ⇒ (x1 − 1, x2 , x3 ) with transmission rate x1 μp , denotes that the channels before the PU departs are only occupied by the PU. • State (x1 , x2 , x3 ) ⇒ (x1 − 1, x2 , x3 − m) with trans mission rate x1 μp , denotes that m channels occupied •

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2011 proceedings.

M

M N −N x1

x2

x1 =0

x2 =0

x3 =0

P (x1 , x2 , x3 )φ(x1 , x2 , x3 ) = 1

(8)

An SUG that is terminated by the new arrival of PUs, would be forced termination if there is no channel for spectrum handoff, and the forced termination probability of SUG PF can be derived as PF = 1 − Pr{l > 0} min(x2 ,N )

≈1−

κ=1

λp (ρQd + ϕQf )p(x, κ) Pr{α − κ < 0} μp (9)

where α is the remaining number of reservation channels, and α ≤ α because of partial reservation channels are occupied by PUs. Assuming that all the SUGs have the same data transmission rate (normalized 1 bit per unit time), the throughput of SUG can be expressed as the product of connection completion rate and the average duration of the successfully completed connections. The connection completion rate is (1 − PB )(1 − PF )λs ,

1 0.9 0.8

d

0.7

Q

by the PU and SUG before the PU departs, and m ∈ [0, min(x3 , N )]. 3) SUG arrival, there are three situations for an SUG arrival: • State (x1 , x2 , x3 ) ⇒ (x1 , x2 + 1, x3 ) with transmission rate ϕ(1 − Qf )λs , denotes that the new SUG accesses the available channel that not occupied by a PU. • State (x1 , x2 , x3 ) ⇒ (x1 , x2 , x3 ), denotes that the new SUG is blocked, and the transmission rate is 1. The new arrival of SUG would be blocked if the number of available channels below the threshold of SAC, so the blocking probability of SUGs with SAC scheme can be obtained as (7), where l denotes the number of channels which are available for SUGs, P (x1 , x2 , x3 ) is the state probability. Let Φ as the set of feasible state in the Markov model, and φ(x1 , x2 , x3 ) denote an indicator of Φ: φ(x1 , x2 , x3 ) = 1 if x = (x1 , x2 , x3 ) ∈ Φ, and 0 otherwise. • State (x1 , x2 , x3 ) ⇒ (x1 , x2 +1, x3 +1) with transmission rate ϕ(1−Qf )λs , denotes that the new SUG accesses the channel that also is occupied by a PU because of the false alarm. 4) SUG departure, there are two situations for an SUG departure: • State (x1 , x2 , x3 ) ⇒ (x1 , x2 − 1, x3 ) with transmission rate x2 μs , denotes that the SUG departs the channel that not occupied by a PU. • State (x1 , x2 , x3 ) ⇒ (x1 , x2 − m, x3 − m) with transmission rate ρQd p(x, κ), denotes that the SUG departs the interfered channel that also is occupied by a PU, or performs spectrum handoff for detecting the presence of a PU, and m ∈ [0, min(x3 , N )]. As shown in figure 3, we could get the balance equations, and

0.6 0.5 δ=1 δ=3 δ=50 δ=200

0.4 0.3 0.2 −20

−19

−18

−17

−16 γ (dB)

−15

−14

−13

−12

Fig. 4: SUG’s detection probability versus PU’s SNR γ

the average duration of the successful completed connection μs )−1 , and the throughput of SUG is given by [3] is ( 1−P F T hs = (1 − PB )(1 − PF )λs (

μs )−1 1 − PF

(10)

IV. N UMERICAL R ESULTS In this section, we present the numerical results of blocking probability, forced termination probability, and the throughput to explore the optimal and suboptimal performance of the system. The parameters setting for the system is shown in Table I. To maximize the throughput of SUs, we develop this optimization considering the effect of SUG and SAC respectively with parameters α and δ. TABLE I: Simulation Parameters Parameter

Value

Description

M N λp μp λs μs γ N1 n0 ˜f Q

3 6 0.2 0.3 0.1 ∼ 0.9 0.6 -20 ∼ -12 dB 5 200 0.1

Number of spectrum bands Number of channels in each band Arrival rate of PUs Departure rate of PUs Arrival rate of SUGs Departure rate of SUGs SNR Time-bandwidth Number of SUs SUG’s targeted false alarm probability

To examine the effect of SUG, we first analyze the SUG’s cooperation in Subsection A of Section III, and caculate the SUG’s detection probability Qd versus PU’s SNR γ with different δ as shown in Fig. 4, considering that the SUG’s false alarm probability is constant and equals 0.1. In Fig. 4, Qd will be increased with the increase of γ and δ. It is clear that Qd would be greater than 95% where γ = −15 and δ = 3. Assuming that ρ = ϕ = 0.5, the influence of the SUG’s detection and false alarm probability is described in Subsection B of Section III, and Fig. 5, Fig. 6 and Fig. 7, use the parameter γ = −15 and δ = 3. Fig. 5 shows the SUG’s blocking probability with different SAC threshold α versus SUGs’ arrival rate λs . The blocking

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2011 proceedings.

PB = Pr{l < α} =

M

M N −N x1

x2

x1 =0

x2 =0

x3 =0

Pr{N x1 + x2 − x3 ≥ M N − α}P (x1 , x2 , x3 )φ(x1 , x2 , x3 )

0.75

0.22 α=1 α=2 α=3 α=4 α=5

0.18

0.7 α=1 α=2 α=3 α=4 α=5

0.7

0.65

0.5

PF

PB

0.6 0.16

α=1 α=2 α=3 α=4 α=5

0.6

Throughput

0.2

(7)

0.55

0.4

0.3

0.14

0.12

0.1 0.1

0.2

0.3

0.4

0.5 λs

0.6

0.7

0.8

0.9

Fig. 5: Blocking probability PB versus SUGs’ arrival rate λs

0.5

0.2

0.45

0.1

0.4 0.1

0.2

0.3

0.4

0.5 λs

0.6

0.7

0.8

0.9

Fig. 6: Forced Termination probability PF versus SUGs’ arrival rate λs

probability, PB , increases when the SUGs’ arrival rate λs increases from 0.1 to 0.9 under a specific value of α. clearly, PB would increase rapidly when α increases from 1 to 5 under a given SUGs’ arrival rate λs . It is worth noting that SUGs’ blocking probability will increase with the SAC threshold α. This indicates that the spectrum admission control mechanism will effect the access of new arrival SUGs, and the SUG’s detection probability and false alarm probability also change the curve of PB . In Fig. 6, the SUGs’ forced termination probability PF decreases when SAC threshold α varies 1 to 5. In this case, the more channels reserved for spectrum handoff, the less SUGs would be forced terminated. It is a tradeoff between PB and PF to achieve greater throughput of SUGs. Fig. 7 shows that the SUGs’ throughput when the SAC threshold α varies 1 to 5. The throughput will increase more if we set the targeted SUG’s false alarm probability less than 0.1, and the parameter δ will also get greater to make the SUG’s detection probability exceed 0.95. It is clear that the SUGs’ throughput will increase with the increment of SUGs’ arrival rate λs . V. C ONCLUSION In this paper, we have proposed a novel spectrum handoff framework with ‘SUG’, considering the SUG’s cooperation and the spectrum admission control mechanism. We have obtained that the SUG’s detection probability would greater than 95% under a constant SUG’s false alarm probability with ‘OR’ rule when the SUG’s size δ no fewer than 3 for a given γ = −15 dB, and then the numerical results have shown the performance of the scheme with the appointed δ. Our proposed scheme is suitable for multi-user CRNs to achieve steady throughput. Future works including considering spectrum sensing overhead and spectrum handoff overhead. ACKNOWLEDGMENT This work was supported by The National Nature Science Foundation of China under Grants No. 60832009, 60872017,

0 0.1

0.2

0.3

0.4

0.5 λs

0.6

0.7

0.8

0.9

Fig. 7: SUGs’ throughput T hs versus SUGs’ arrival rate λs

60772100. R EFERENCES [1] I. F. Akyildiz, W. Y. Lee, M. C. Vuran, and S. Mohanty, “Next generation/dynamic spectrum access/cognitive radio wireless networks: A survey,” Computer Networks, vol. 50, no. 13, pp. 2127–2159, 2006. [2] X. Zhu, L. Shen, and T. S. P. Yum, “Analysis of cognitive radio spectrum access with optimal channel reservation,” IEEE Communications Letters, vol. 11, no. 4, pp. 304–306, 2007. [3] W. Ahmed, J. Gao, H. A. Suraweera, and M. Faulkner, “Comments on ”analysis of cognitive radio spectrum access with optimal channel reservation”,” IEEE Transactions on Wireless Communications, vol. 8, no. 9, pp. 4488–4491, 2009. [4] P. K. Tang, Y. H. Chew, L. C. Ong, and M. K. Haldar, “Performance of secondary radios in spectrum sharing with prioritized primary access,” in IEEE MILCOM 2006, Washington, DC, USA, October 2006, pp. 1–7. [5] J. Gambini, O. Simeone, U. Spagnolini, Y. Bar-Ness, and Y. Kim, “Cognitive radio with secondary packet-by-packet vertical handover,” in Proc. IEEE ICC 2008, Beijing, China, May 2008, pp. 1050–1054. [6] H. P. Shiang and M. van der Schaar, “Queuing-based dynamic channel selection for heterogeneous multimedia applications over cognitive radio networks,” IEEE Transactions on Multimedia, vol. 10, no. 5, pp. 896– 909, 2008. [7] L. C. Wang and C. W. Wang, “Spectrum handoff for cognitive radio networks: Reactive-sensing or proactive-sensing?” in IEEE IPCCC 2008, Austin, Texas, USA, December 2008, pp. 343–348. [8] M. Huang, R. Yu, and Y. Zhang, “Call admission control with soft-qos based spectrum handoff in cognitive radio networks,” in IWCMC 2009, Leipzig, Germany, June 2009, pp. 1067–1072. [9] D. Pacheco-Paramo, V. Pla, and J. Martinez-Bauset, “Optimal admission control in cognitive radio networks,” in 4th International Conference on CROWNCOM 2009, Hannover, Germany, June 2009, pp. 281–287. [10] B. Hamdaoui, “Adaptive spectrum assessment for opportunistic access in cognitive radio networks,” IEEE Transactions on Wireless Communications, vol. 8, no. 2, pp. 922–930, 2009. [11] J. Martinez-Bauset, V. Pla, M. J. Domenech-Benlloch, and D. PachecoParamo, “Admission control and interference management in dynamic spectrum access networks,” Eurasip Journal On Wireless Communications and Networking, 2010. [12] E. Peh and Y. C. Liang, “Optimization for cooperative sensing in cognitive radio networks,” in IEEE WCNC 2007, Hong Kong, March 2007, pp. 27–32. [13] A. Ghasemi and E. Sousa, “Collaborative spectrum sensing for opportunistic access in fading environments,” in First IEEE International Symposium on New Frontiers in DySPAN 2005, Baltimore Harbor, Maryland, USA, Nov. 2005, pp. 131–136.

A Novel Spectrum Handoff Scheme With Spectrum Admission Control in Cognitive Radio Networks Chengyu Wu, Chen He, Lingge Jiang

Yunfei Chen

Department of Electronic Engineering Shanghai Jiao Tong University, China Email: {jerry916, chenhe, lgjiang}@sjtu.edu.cn

School of Engineering University of Warwick Coventry, UK, CV4 7AL Email: [email protected]

Abstract—In this paper, we propose a new spectrum handoff scheme with spectrum admission control (SAC). In the proposed scheme, the secondary users (SUs) make up secondary user groups (SUGs) to achieve appointed detection probability of primary user signals, and perform spectrum handoff to an available spectrum when the primary users (PUs) reuse the spectrum in cognitive radio networks (CRNs). A simple Markov model is adopted to analyze the performance of spectrum handoff in terms of blocking probability, forced termination probability and throughput of the cognitive radio system. Numerical results show that the new handoff strategy is suitable for multi-user cognitive radio systems, and that spectrum admission control and cooperative sensing can effectively increase the efficiency of spectrum handoff.

I. I NTRODUCTION Cognitive radio (CR) improves the usage efficiency of the under-utilized radio spectrum, and this has motivated researchers to study CR. One of the most important functionalities in CR is spectrum mobility [1]. To achieve spectrum mobility, spectrum handoff is required. In a cognitive radio network with both primary users (PUs) and secondary users (SUs), the PUs do not always occupy their frequency bands. Consequently, the SUs may temporarily occupy the frequency bands which are not occupied by the PUs. However, the SUs must vacate these frequency bands when the PUs come back and reuse them. In this case, the communication links of the SUs established over the PUs’ frequency bands will be lost and therefore, the SUs must find another idle frequency bands to resume their communication links. This is called spectrum handoff [2]. Spectrum handoff is less investigated in the literature compared with spectrum sensing. Most of the literature research the spectrum handoff is applying a Markov chain model to analyze the performance. In [2], [3], a Markov chain model is proposed to analyze the performance of spectrum access, and then a channel reservation scheme for spectrum handoff is designed to make tradeoff between the forced termination probability and the blocking probability. A three-dimensional Markov chain model is developed in [4] to analyze and study the cognitive radio performance given certain primary traffic, and also a channel reservation scheme for the primary system is considered to reduce the forced termination probability. And some models for spectrum handoff is based on queuing theory. A queuing system is designed to maximize the system

throughput under maximum-delay quality of service (QoS) constraints in the primary activity considered in [5]. In [6], the authors propose a novel priority virtual queue interface to analyze the multimedia users’ interactions and determine the required information exchange according to the priority queuing analysis. A preemptive resume priority (PRP) M/G/1 queuing network is developed in [7] to analyze the reactiveor proactive-sensing spectrum handoff considering the sensing time. In order to maintain a given QoS requirement for SUs, a proper spectrum admission control (SAC) scheme, which is known as an effective functionality of the media access control (MAC) layer in ensuring the QoS of wireless networks, should be integrated into the MAC layer. Since the number of available frequency bands always changes dynamically due to the random variations of the PU traffic, SAC is an indispensable part of a CR network to improve the flexibility and to satisfy the QoS of the SUs’. In [8], a soft-QoSbased spectrum handoff mechanism is integrated into SAC by using the Markov chain model to describe the procedures of SAC and spectrum handoff. A nonlinear optimization is formulated to minimize the dropping probability with the constraint of blocking probability. In [9], a fractional guard channel reservation scheme is applied to tradeoff between the blocking probability of SUs and the forced termination probability. In this paper, a new spectrum handoff scheme with SAC is proposed for cognitive radio networks (CRNs) by introducing cooperation among SUs into sensing to reduce the probability of missed detection and the probability of false alarm. Also, we develop secondary user group (SUG) as a unit to access the available spectrum, the members of which can cooperate with each other to prevent hidden-terminal problem and to increase the detection probability. A suitable approach for SUGs’ spectrum access is also developed. Based on this, a Markov model is proposed to analyze the performance of the spectrum handoff in terms of blocking probability, forced termination probability and system throughput. Numerical results show that the new spectrum handoff scheme is ideal for multi-user cognitive radio systems, the approach of SAC for SUGs can effectively improve the system performance. The remainder of the paper is organized as follows. In Section II, the system model is described along with the

978-1-4244-9268-8/11/$26.00 ©2011 IEEE

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2011 proceedings.

1 SU1 SU7 SUG1

Others

2

RCTS Handoff Sense Handoff

SU 1 RRTS

3

SU 4

SU4

SUG2

SU6

ĂĂ

SU2

λp

CCC

MN-3

SU 1 DATA

SU 4

MN-1 MN

ACK

RCTS Handoff Sense Handoff

OR Handoff RRTS

DATAPeriodic

Handoff

Sensing

Handoff

Handoff Sense Handoff OR CRTS Handoff RTS

RCTS Handoff Sense Handoff CCC

Channel l

Handoff CCC

Fig. 2: SUG’s Shaping and Spectrum Handoff

•

•

•

II. S YSTEM M ODEL

As shown in Fig. 1, we assume that all members of each SUG are single radio devices, and they have the same outband common control channel (CCC) to communicate with each other. An SUG may consist of two or more SUs, which may join or leave the group at any time. In general, an SUG is shown as Fig. 2(1) and performs as follows:

DATA Channel j

CCC

(2) SUG's Spectrum Handoff for PU's activity with Periodic Sensing

Fig. 1: System Model

A. SUGs’ Shaping

Periodic Sensing

Handoff

Channel j

SU9

In the considered CR system, the PUs do not need to know anything about the SUs, so that the primary system should not be modified due to the activity of secondary users. We assume that a channel is a basic spectrum unit, and that the spectrum consists of M parallel primary bands with each band divided into N channels. Each band is allocated for a PU, each channel is allocated for an SU, and PUs have priority to use the channels over SUs. We assume the arrival of PUs is Poisson process with arrival rate λp , and the corresponding service time is exponentially distributed with rate μp . And we define the packets arrival of SUs as a distributed Poisson process with rate λs , the duration time as exponentially distribution with parameter μs , and the traffic load perform as η = λs /μs . The observation range of one single SU is small and typically less than its transmission range. So the SUs may cause interferences to the PUs inside their transmission ranges even though the SUs consider the channels available. In this case, we denote a group of SUs as an SUG [10] which could be cooperative with each other to prevent hidden-terminal problem and to increase the sensing accuracy.

CTS

CACK

(1) SUG's Shaping

MN-2

considered procedures and implementation schemes. Subsequently, in Section III, the cooperation spectrum sensing scheme is briefly introduced along with the main hypothesis and considerations, and the Markov model is illustrated and analyzed with a number of relevant performance metrics. The numerical results and discussions are given in Section IV. Finally, Section VI concludes the paper with some final remarks and future considerations.

DATA

PU Active Periodic Sensing

Others

SUG3 SU8

Handoff

Channel j

PU 1

SU5

SU3

Handoff Sense Handoff OR CRTS Handoff RTS

RCTS Handoff Sense Handoff

4 SUG Base

Handoff

•

•

•

•

Step 1: Initialization, the SUs sense channels and report the sensing results to the SUG base station, and SUG base station initializes the channels usage statistics. Step 2: Polling, SU 1 that want to communicate with SU 4 sends broadcast RRTS (Ready Request to Send, including the channel number j to be sensed) messages in the CCC. Step 3: Reporting, other SUs (including the specified SU 4), which receive the RRTS messages, sense the appointed channel and send RCTS (Ready Clear to Send, including the sensing result) messages to SU 1, SU 1 would not send ‘Reject’ messages to the certain SUs until the number of received RCTS at SU 1 up to δ, which denotes the optimal number of members in an SUG. Step 4: Decision, SU 1 merges the results with ‘OR’ rule to decide to access the channel or not, and report the results to the SUG base station. If not, go to Step 2. Step 5: Handoff , SU 1 sends CRTS (Control Request to Send) messages in CCC, and all the members of the SUG will receive CRTS and perform spectrum handoff to channel j, and also SU 1 switch to channel j. Step 6: Transmitting, SU 1 sends RTS (Request to Send) messages on channel j, SU 4 will receive RTS and send CTS(Clear to Send) back to establish communication with SU 1. The other SUs of the SUG continue to sense inband on channel j periodically. At the end, SU 4 sends ACK (ACKnowledge Character) if it completely receives the data. Step 7: Idle, SU 1 receives ACK , broadcasts CACK (Complete ACKnowledge Character) on channel j, and all the members of the SUG perform spectrum handoff to CCC after receiving CACK. Finally, go to Step 2.

B. Spectrum Handoff with Spectrum Admission Control The SUG should perform spectrum handoff to guarantee the QoS of PU in case a PU starts to transmit, so periodic sensing is necessary. As shown in Fig. 2(2), all the members of the SUG do spectrum handoff to CCC and sense channels to achieve an available channel when the SUG detects PU is active with periodic sensing. The sensing period determines

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2011 proceedings.

the maximum time during which the SUG will be unaware of a reappearing PU and hence may harmfully interfere with it. Designing an admission control scheme for SUGs is able to limit both the forced termination probability of SUGs and the interference probability upon PUs from SUGs. We define α ∈ [0, M N ] as the admission control threshold, considering the SUGs which perform spectrum handoff have priority over the new arrival SUGs to access the reservation channels. An admission decision is taken as follows: >α accept (1) M N − b(x + e2 ) ≤α reject

′+ (x 1

x2 μ s

μ 1)

p

ρ Qd p (x, κ )λ p

x1′, x2 , x3

ϕ (1-Q f )λs

) λp

Q

x1′ + 1, x2 − κ , x3

ϕ(

fo

rm

ρQ

d

p

=1

,κ (x

f

)

ϕ (1-Q f ) λ s

Qf 1-

)λ s

x1′, x2 − m, x3 − m

p(

x,

)

λ

And the detection probability of the individual SUs in the SUG is given as √ 1 · Q−1 (Pf ) − N 1 γ) Pd = Q( √ 2γ + 1 (5) √ 1 δ ˜ f ) − N 1 γ) = Q( √ · Q−1 (1 − 1 − Q 2γ + 1 where γ is the PU’s signal to noise ratio (SNR) received at the cooperative SU, Q(·) is the complementary cumulative distribution function of the standard Gaussian, and N1 = τs B is the time-bandwidth, where τs is the sensing duration and B is the bandwidth of channel j. The total Qd of the SUG is derived by substituting (5) into (2). The optimum number δ (less than the number of all SUs which denotes as n0 ) of SUs in an SUG to obtain the accredited Qd is computed from (4), (5) and (2).

ρ(

d

κ

Ideally, we would like to have Qd = 1.0 and Qf = 0.0, but these will be difficult to achieve in the practical situation. It is more reasonable to allow Qd = 0.9 ∼ 0.95 and Qf = 0.01 ∼ 0.1 [12]. In the ‘OR’ fusion rule, to achieve a targeted ˜ f , the false alarm probability of the false alarm probability Q individual SUs in the SUG is given as δ ˜f Pf = 1 − 1 − Q (4)

( x2 + 1) μ s

Q 1−

ϕ

(3)

x1′, x2 + 1, x3 + 1

x1′ + 1, x2 , x3 + m

( x2 + 1) μ s

Qf = 1 − (1 − Pf )

ϕ (1-Q f ) ⋅ λ s

(2)

δ

p x′1μ

Qd = 1 − (1 − Pd )δ

λp

A. SUG’s Cooperation Assumed that the detection probability for a channel of an SU is Pd , and the optimal number of cooperative SUs in an SUG is δ. The target detection probability and false alarm probability of the SUG for cooperative sensing using the OR fusion rule with energy detector is given as [12], [13]:

d

III. A NALYSIS

x1′ − 1, x2 , x3 − m ρQ

We define x = (x1 , x2 ) as the system vector, where x1 denotes the number of PUs that SUGs observed and x2 stands for the number of in-service SUGs. Considering the presence of sensing error(miss detection or false alarm), the actual number of PUs may be x1 = x1 + 1 or x1 = x1 − 1 [11]. So we adopt x = (x1 , x2 , x3 ) as the actual system vector, x3 denotes the number of interfered channels [9]. Let b(x) = N x1 +x2 +x3 denote the number of occupied channels at state x, and ei be a triple dimensional vector with position i set to 1 and other positions set to 0.

x1′, x2 − 1, x3

p

x1′, x2 − κ , x3

x1′, x2 + 1, x3 Fig. 3: Markov Model

B. Analysis with Markov Model To analyze the performance of spectrum handoff with SAC for SUGs, we develop a simple Markov model shown as Fig. 3. P r(H0 ) and P r(H1 ) denote the probability of the hypotheses H0 (noise only, denoting PU’s signal absence) and H1 (where the PU’s signal is present) respectively, and we denote ρ = P r(H1 ), ϕ = P r(H0 ), and ρ+ϕ = 1. Considering the detection probability and the false alarm probability of an SUG, we classify the event that trigger the transmission of the system as follows: 1) PU arrival, there are three situations for a PU arrival: • The SUG senses the arrival of PU, and the PU is actually there. The PU arrival will terminate κ SUGs, and trigger state(x1 , x2 , x3 ) ⇒ (x1 + 1, x2 − κ, x3 ) shown as Fig. 3, κ ∈ [0, min(x2 , N )] and the transmission rate of a new PU arrival with termination κ SUGs is ρQd · p(x, κ)λp , where p(x, κ) is given by N (M −x −1)N p(x, κ) =

κ

1

x −κ

(M −x2 )N x2

1

(6)

The SUG detects the arrival of PU, but the PU is not there. This is a false alarm, the results also would terminate κ SUGs, state (x1 , x2 , x3 ) ⇒ (x1 , x2 − κ, x3 ) with transmission rate ϕQf · p(x, κ)λp . • The SUG does not detect the arrival of PU, but the PU is there. This is a missed detection, state (x1 , x2 , x3 ) ⇒ (x1 +1, x2 , x3 +m) with transmission rate ρ(1−Qd )·λp , where m ∈ [0, min(x3 , N )]. 2) PU departure, there are two situations for a PU departure: • State (x1 , x2 , x3 ) ⇒ (x1 − 1, x2 , x3 ) with transmission rate x1 μp , denotes that the channels before the PU departs are only occupied by the PU. • State (x1 , x2 , x3 ) ⇒ (x1 − 1, x2 , x3 − m) with trans mission rate x1 μp , denotes that m channels occupied •

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2011 proceedings.

M

M N −N x1

x2

x1 =0

x2 =0

x3 =0

P (x1 , x2 , x3 )φ(x1 , x2 , x3 ) = 1

(8)

An SUG that is terminated by the new arrival of PUs, would be forced termination if there is no channel for spectrum handoff, and the forced termination probability of SUG PF can be derived as PF = 1 − Pr{l > 0} min(x2 ,N )

≈1−

κ=1

λp (ρQd + ϕQf )p(x, κ) Pr{α − κ < 0} μp (9)

where α is the remaining number of reservation channels, and α ≤ α because of partial reservation channels are occupied by PUs. Assuming that all the SUGs have the same data transmission rate (normalized 1 bit per unit time), the throughput of SUG can be expressed as the product of connection completion rate and the average duration of the successfully completed connections. The connection completion rate is (1 − PB )(1 − PF )λs ,

1 0.9 0.8

d

0.7

Q

by the PU and SUG before the PU departs, and m ∈ [0, min(x3 , N )]. 3) SUG arrival, there are three situations for an SUG arrival: • State (x1 , x2 , x3 ) ⇒ (x1 , x2 + 1, x3 ) with transmission rate ϕ(1 − Qf )λs , denotes that the new SUG accesses the available channel that not occupied by a PU. • State (x1 , x2 , x3 ) ⇒ (x1 , x2 , x3 ), denotes that the new SUG is blocked, and the transmission rate is 1. The new arrival of SUG would be blocked if the number of available channels below the threshold of SAC, so the blocking probability of SUGs with SAC scheme can be obtained as (7), where l denotes the number of channels which are available for SUGs, P (x1 , x2 , x3 ) is the state probability. Let Φ as the set of feasible state in the Markov model, and φ(x1 , x2 , x3 ) denote an indicator of Φ: φ(x1 , x2 , x3 ) = 1 if x = (x1 , x2 , x3 ) ∈ Φ, and 0 otherwise. • State (x1 , x2 , x3 ) ⇒ (x1 , x2 +1, x3 +1) with transmission rate ϕ(1−Qf )λs , denotes that the new SUG accesses the channel that also is occupied by a PU because of the false alarm. 4) SUG departure, there are two situations for an SUG departure: • State (x1 , x2 , x3 ) ⇒ (x1 , x2 − 1, x3 ) with transmission rate x2 μs , denotes that the SUG departs the channel that not occupied by a PU. • State (x1 , x2 , x3 ) ⇒ (x1 , x2 − m, x3 − m) with transmission rate ρQd p(x, κ), denotes that the SUG departs the interfered channel that also is occupied by a PU, or performs spectrum handoff for detecting the presence of a PU, and m ∈ [0, min(x3 , N )]. As shown in figure 3, we could get the balance equations, and

0.6 0.5 δ=1 δ=3 δ=50 δ=200

0.4 0.3 0.2 −20

−19

−18

−17

−16 γ (dB)

−15

−14

−13

−12

Fig. 4: SUG’s detection probability versus PU’s SNR γ

the average duration of the successful completed connection μs )−1 , and the throughput of SUG is given by [3] is ( 1−P F T hs = (1 − PB )(1 − PF )λs (

μs )−1 1 − PF

(10)

IV. N UMERICAL R ESULTS In this section, we present the numerical results of blocking probability, forced termination probability, and the throughput to explore the optimal and suboptimal performance of the system. The parameters setting for the system is shown in Table I. To maximize the throughput of SUs, we develop this optimization considering the effect of SUG and SAC respectively with parameters α and δ. TABLE I: Simulation Parameters Parameter

Value

Description

M N λp μp λs μs γ N1 n0 ˜f Q

3 6 0.2 0.3 0.1 ∼ 0.9 0.6 -20 ∼ -12 dB 5 200 0.1

Number of spectrum bands Number of channels in each band Arrival rate of PUs Departure rate of PUs Arrival rate of SUGs Departure rate of SUGs SNR Time-bandwidth Number of SUs SUG’s targeted false alarm probability

To examine the effect of SUG, we first analyze the SUG’s cooperation in Subsection A of Section III, and caculate the SUG’s detection probability Qd versus PU’s SNR γ with different δ as shown in Fig. 4, considering that the SUG’s false alarm probability is constant and equals 0.1. In Fig. 4, Qd will be increased with the increase of γ and δ. It is clear that Qd would be greater than 95% where γ = −15 and δ = 3. Assuming that ρ = ϕ = 0.5, the influence of the SUG’s detection and false alarm probability is described in Subsection B of Section III, and Fig. 5, Fig. 6 and Fig. 7, use the parameter γ = −15 and δ = 3. Fig. 5 shows the SUG’s blocking probability with different SAC threshold α versus SUGs’ arrival rate λs . The blocking

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2011 proceedings.

PB = Pr{l < α} =

M

M N −N x1

x2

x1 =0

x2 =0

x3 =0

Pr{N x1 + x2 − x3 ≥ M N − α}P (x1 , x2 , x3 )φ(x1 , x2 , x3 )

0.75

0.22 α=1 α=2 α=3 α=4 α=5

0.18

0.7 α=1 α=2 α=3 α=4 α=5

0.7

0.65

0.5

PF

PB

0.6 0.16

α=1 α=2 α=3 α=4 α=5

0.6

Throughput

0.2

(7)

0.55

0.4

0.3

0.14

0.12

0.1 0.1

0.2

0.3

0.4

0.5 λs

0.6

0.7

0.8

0.9

Fig. 5: Blocking probability PB versus SUGs’ arrival rate λs

0.5

0.2

0.45

0.1

0.4 0.1

0.2

0.3

0.4

0.5 λs

0.6

0.7

0.8

0.9

Fig. 6: Forced Termination probability PF versus SUGs’ arrival rate λs

probability, PB , increases when the SUGs’ arrival rate λs increases from 0.1 to 0.9 under a specific value of α. clearly, PB would increase rapidly when α increases from 1 to 5 under a given SUGs’ arrival rate λs . It is worth noting that SUGs’ blocking probability will increase with the SAC threshold α. This indicates that the spectrum admission control mechanism will effect the access of new arrival SUGs, and the SUG’s detection probability and false alarm probability also change the curve of PB . In Fig. 6, the SUGs’ forced termination probability PF decreases when SAC threshold α varies 1 to 5. In this case, the more channels reserved for spectrum handoff, the less SUGs would be forced terminated. It is a tradeoff between PB and PF to achieve greater throughput of SUGs. Fig. 7 shows that the SUGs’ throughput when the SAC threshold α varies 1 to 5. The throughput will increase more if we set the targeted SUG’s false alarm probability less than 0.1, and the parameter δ will also get greater to make the SUG’s detection probability exceed 0.95. It is clear that the SUGs’ throughput will increase with the increment of SUGs’ arrival rate λs . V. C ONCLUSION In this paper, we have proposed a novel spectrum handoff framework with ‘SUG’, considering the SUG’s cooperation and the spectrum admission control mechanism. We have obtained that the SUG’s detection probability would greater than 95% under a constant SUG’s false alarm probability with ‘OR’ rule when the SUG’s size δ no fewer than 3 for a given γ = −15 dB, and then the numerical results have shown the performance of the scheme with the appointed δ. Our proposed scheme is suitable for multi-user CRNs to achieve steady throughput. Future works including considering spectrum sensing overhead and spectrum handoff overhead. ACKNOWLEDGMENT This work was supported by The National Nature Science Foundation of China under Grants No. 60832009, 60872017,

0 0.1

0.2

0.3

0.4

0.5 λs

0.6

0.7

0.8

0.9

Fig. 7: SUGs’ throughput T hs versus SUGs’ arrival rate λs

60772100. R EFERENCES [1] I. F. Akyildiz, W. Y. Lee, M. C. Vuran, and S. Mohanty, “Next generation/dynamic spectrum access/cognitive radio wireless networks: A survey,” Computer Networks, vol. 50, no. 13, pp. 2127–2159, 2006. [2] X. Zhu, L. Shen, and T. S. P. Yum, “Analysis of cognitive radio spectrum access with optimal channel reservation,” IEEE Communications Letters, vol. 11, no. 4, pp. 304–306, 2007. [3] W. Ahmed, J. Gao, H. A. Suraweera, and M. Faulkner, “Comments on ”analysis of cognitive radio spectrum access with optimal channel reservation”,” IEEE Transactions on Wireless Communications, vol. 8, no. 9, pp. 4488–4491, 2009. [4] P. K. Tang, Y. H. Chew, L. C. Ong, and M. K. Haldar, “Performance of secondary radios in spectrum sharing with prioritized primary access,” in IEEE MILCOM 2006, Washington, DC, USA, October 2006, pp. 1–7. [5] J. Gambini, O. Simeone, U. Spagnolini, Y. Bar-Ness, and Y. Kim, “Cognitive radio with secondary packet-by-packet vertical handover,” in Proc. IEEE ICC 2008, Beijing, China, May 2008, pp. 1050–1054. [6] H. P. Shiang and M. van der Schaar, “Queuing-based dynamic channel selection for heterogeneous multimedia applications over cognitive radio networks,” IEEE Transactions on Multimedia, vol. 10, no. 5, pp. 896– 909, 2008. [7] L. C. Wang and C. W. Wang, “Spectrum handoff for cognitive radio networks: Reactive-sensing or proactive-sensing?” in IEEE IPCCC 2008, Austin, Texas, USA, December 2008, pp. 343–348. [8] M. Huang, R. Yu, and Y. Zhang, “Call admission control with soft-qos based spectrum handoff in cognitive radio networks,” in IWCMC 2009, Leipzig, Germany, June 2009, pp. 1067–1072. [9] D. Pacheco-Paramo, V. Pla, and J. Martinez-Bauset, “Optimal admission control in cognitive radio networks,” in 4th International Conference on CROWNCOM 2009, Hannover, Germany, June 2009, pp. 281–287. [10] B. Hamdaoui, “Adaptive spectrum assessment for opportunistic access in cognitive radio networks,” IEEE Transactions on Wireless Communications, vol. 8, no. 2, pp. 922–930, 2009. [11] J. Martinez-Bauset, V. Pla, M. J. Domenech-Benlloch, and D. PachecoParamo, “Admission control and interference management in dynamic spectrum access networks,” Eurasip Journal On Wireless Communications and Networking, 2010. [12] E. Peh and Y. C. Liang, “Optimization for cooperative sensing in cognitive radio networks,” in IEEE WCNC 2007, Hong Kong, March 2007, pp. 27–32. [13] A. Ghasemi and E. Sousa, “Collaborative spectrum sensing for opportunistic access in fading environments,” in First IEEE International Symposium on New Frontiers in DySPAN 2005, Baltimore Harbor, Maryland, USA, Nov. 2005, pp. 131–136.