MAC Protocol for Cognitive Radio Networks

1

Course Report

Student: Le Thanh Tan (LETT25047901)Instructor: Prof. Le Bao LongSemester: Winter 2011Report Topic: MAC Protocol for Cognitive Radio Networks

I. I NTRODUCTION

Emerging broadband wireless applications have been demanding unprecedented increase in radio spec-trum resources. As a result, we have been facing a serious spectrum shortage problem. However, severalrecent measurements reveal very low spectrum utilization in most useful frequency bands [1]. To resolvethis spectrum shortage problem, the Federal CommunicationsCommission (FCC) has opened licensedbands for unlicensed users’ access. This important change in spectrum regulation has resulted in growingresearch interests on dynamic spectrum sharing and cognitive radio in both industry and academia. Inparticular, IEEE has established an IEEE 802.22 workgroup to build the standards of WRAN based onCR techniques [1].

Hierarchical spectrum sharing between primary networks and secondary networks is one of the mostpopular dynamic spectrum sharing paradigms. For this spectrum sharing paradigm, primary users typicallyhave strictly higher priority than secondary users in accessing the underlying spectrum. Therefore, aproper spectrum sensing mechanism is needed so that secondary users can search for and exploit availablespectrum holes (i.e., available frequency bands) [2]. There are several challenging technical issues related tothis spectrum discovery and exploitation problem. On one hand, secondary users should spend sufficientlylong time for spectrum sensing so that they do not interfere with active primary users. On the other hand,secondary users should efficiently exploit spectrum holes to transmit their data by using an appropriatespectrum sharing mechanism. Even though these aspects are tightly coupled with each other, they are nottreated thoroughly in the existing literature.

There is a rich literature on spectrum sensing for cognitiveradio networks (e.g., see [3] and referencestherein). Classical sensing based on, for example, energy detection techniques or advanced cooperativesensing strategies [4] where multiple secondary users collaborate with another to improve the sensingperformance have been investigated in the literature. In [2], optimization of sensing and throughput tradeoffunder a detection probability constraint was investigated. It was shown that the detection constraint is metwith equality at optimality. However, this optimization tradeoff was only investigated for a simple scenariowith one pair of secondary users. There are also a large number of papers considering MAC protocoldesign and analysis for cognitive radio networks [5]-[10].However, these existing works either assumedperfect spectrum sensing or did not explicitly model the sensing imperfection in their design and analysis.

In this paper, we make a further bold step in designing, analyzing, and optimizing a MAC protocolfor cognitive radio networks considering sensing performance captured in detection and false alarmprobabilities. Specifically, the contributions of this paper can be summarized as follows:i) we designa distributed synchronized MAC protocol for cognitive radio networks incorporating spectrum sensingoperation; ii ) we analyze saturation throughput of the proposed MAC protocols; iii ) we perform thethroughput maximization of the proposed MAC protocol against its key parameters, namely sensing timeand contention window;iv) we present numerical results to illustrate performance ofthe proposed MACprotocol and the throughput gain due to the optimal protocolconfiguration.

TEL352 Reading Course - Final Report

Fig. 1. Typical spectrum sharing model.

The remaining of this paper is organized as follows. SectionII describes a system model and thedesign of a synchronized MAC protocol. Throughput analysisand optimization are provided in SectionIII. Section IV demonstrates numerical results followed byconcluding remarks in Section V.

II. SYSTEM MODEL AND MAC PROTOCOLDESIGN

A. System Model

We consider a network setting whereN pairs of secondary users opportunistically exploit one particularradio channel (i.e., a frequency band) for their data transmission, which belongs a primary network.Extension to a more general case with multiple channels willbe discussed later. We assume that each pairof secondary users can overhear transmissions from other pairs of secondary users (i.e., it is a collocatednetwork). It is further assumed that transmission from eachindividual pair of secondary users affectsone different primary receiver. Note that it is straightforward to extend this assumption to the scenariowhere each pair of secondary users affects more than one primary receiver and/or each primary receiveris affected by more than one pair of secondary users. The network setting under investigation is shownin Fig. 1. In the following, we will refer to pairi of secondary users as secondary linki or flow iinterchangeably.

B. Spectrum Sensing

We assume that secondary links rely on a distributed synchronized MAC protocol to share the spectrumwhen it is available. Specifically, time is divided into fixed-size cycles and it is assumed that secondary linkscan perfectly synchronize with each other (i.e., there is nosynchronization error) [8], [11]. It is assumedthat each secondary link performs spectrum sensing at the beginning of each cycle and only proceeds tocontend with others to capture the channel if the sensing outcome indicates an available channel (i.e., itis not being used by nearby primary users). Detailed MAC protocol design will be elaborated in the nextsubsection.

Let H0 andH1 denote the events that a particular primary user is idle and active, respectively (i.e., theunderlying channel is available and busy, respectively). In addition, letP i (H0) andP i (H1) = 1−P i (H0)be the probabilities that secondary linki sees an idle and busy channel, respectively. We assume thatsecondary users employ an energy detection scheme and letfs be the sampling frequency used in thesensing period whose length isτ for all secondary links. There are two important performance measures,which are used to quantify the sensing performance, namely detection and false alarm probabilities. Inparticular, detection event occurs when a secondary link successfully senses a busy channel and false

2


…

One cycle

Sensing SYN Backoff

DATA ACK DATA ACK

Backoff

time

Fig. 2. Timing diagram of the proposed MAC protocol.

alarm represents the situation when a spectrum sensor returns a busy state for an idle channel (i.e., atransmission opportunity is overlooked).

Assume that transmission signals from primary users are complex-valued PSK signals while the noiseat the secondary links is independent and identically distributed circularly symmetric complex GaussianCN (0, N0) [2]. Then, the detection and false alarm probability for secondary link i can be calculated as[2]

P id

(

εi, τ)

= Q((

εi

N0

− γi − 1

)√

τfs2γi + 1

)

, (1)

P if

(

εi, τ)

= Q((

εi

N0

− 1

)

√

τfs

)

= Q(

√

2γi + 1Q−1(

P id

(

εi, τ))

+√

τfsγi

)

, (2)

where εi is the detection threshold for an energy detector,γi is the signal-to-noise ratio (SNR) of thePU’s signal at the secondary link,fs is the sampling frequency,N0 is the noise power,τ is the sensinginterval, andQ (.) is defined asQ (x) =

(

1/√2π)

∫∞x exp (−t2/2) dt. These probabilities will be used in

throughput analysis in the next section.

C. MAC Protocol Design

We describe our proposed synchronized MAC for dynamic spectrum sharing among secondary flowsin this subsection. We assume that each fixed-size cycle of lengthTcycle is divided into 3 phases, namelysensing phase, synchronization phase, and data transmission phase. During the sensing phase of lengthτ ,all secondary users perform spectrum sensing on the underlying channel. Then, only secondary links whosesensing results indicate an available channel proceed to the next phase (they will be called active secondaryusers/links in the following). In the synchronization phase, active secondary users broadcast beacon signalsfor synchronization purposes. Finally, active secondary users perform contention and transmit data in thedata transmission phase. The timing diagram of one particular cycle is illustrated in Fig. 2.

We assume that the length of each cycle is sufficiently large such that secondary links can transmitseveral packets during the data transmission phase. Indeed, the current 802.22 standard specifies thespectrum evacuation time upon the return of primary users is2 seconds, which is a relatively largeinterval. Therefore, our assumption would be valid for mostpractical cognitive systems. During the datatransmission phase, we assume that active secondary links perform a standard contention technique tocapture the channel similar to that employed by the CSMA/CA protocol. For simplicity, we assume afixed contention window equal toW is used for random backoff. Extension to the case where exponentialbackoff [12] is employed will be discussed later.

Specifically, each active secondary link starts the contention by choosing a random backoff timeuniformly distributed in the range[0,W − 1] and starts decrementing its backoff time counter whilecarrier sensing transmissions from other secondary links.Let σ denote a mini-slot interval, each of whichcorresponds one unit of the backoff time counter. Upon hearing a transmission from any secondarylink, each secondary link will “freeze” its backoff time counter and reactivate when the channel sensed

3


idle again. Otherwise, if the backoff time counter reaches zero, the underlying secondary link winsthe contention and transmits one data packet to its intendedreceiver. The transmitter then waits for anacknowledgment (ACK) from the receiver to indicate a successful reception of the packet. Standard smallintervals, namely DIFS and SIFS, are used before backoff time decrements and ACK packet transmissionas described in [12].

D. Throughput Maximization

Given the sensing model and proposed MAC protocol, we are interested in finding its optimal con-figuration to achieve the maximum throughput subject to protection constraints for primary receivers.Specifically, letNT (τ,W ) be the normalized total throughput, which is a function of sensing timeτ andcontention windowW . Suppose that each primary receiver requires that detection probability achieved byits conflicting primary linki to be at leastP

i

d. Then, the throughput maximization problem can be statedas follows:

maxτ,W

NT (τ,W )

s.t. P id (ε

i, τ) ≥ P̄ id, i = 1, 2, · · · , N

0 < τ ≤ Tcycle, 0 < W ≤ Wmax,

(3)

whereWmax is the maximum contention window and recall thatTcycle is the cycle interval. In fact, optimalsensingτ would allocate sufficient time to protect primary receiversand optimal contention window wouldbalance between reducing collisions among active secondary links and protocol overhead.

III. T HROUGHPUTANALYSIS AND OPTIMIZATION

We perform saturation throughput analysis and solve the optimization problem (3) in this section.Throughput analysis for the cognitive radio setting under investigation is more involved compared tostandard MAC protocol throughput analysis (e.g., see [11],[12]) because the number of active secondarylinks participating in contention in each cycle varies depending on the sensing outcomes. Suppose that allsecondary links have same packet length. LetPr (n = n0) andT (τ, φ |n = n0 ) be the probability thatn0

secondary links participating in contention and the conditional normalized throughput whenn0 secondarylinks join the channel contention, respectively. Then, thenormalized throughput can be calculated as

NT =N∑

n0=1

T (τ,W |n = n0 ) Pr (n = n0), (4)

where recall thatN is the number of secondary links,τ is the sensing time,W is the contention window.In the following, we show how to calculatePr (n = n0) andT (τ, φ |n = n0 ).

A. Calculation ofPr (n = n0)

It is noted that only secondary links whose sensing outcomesin the sensing phase indicate an availablechannel proceed to contention in the data transmission phase. There are two scenarios for which this canhappen for a particular secondary linki:

• The primary user is not active and no false alarm is generatedby the underlying secondary link.• The primary user is active and secondary linki mis-detects its presence.

Therefore, secondary linki joins contention in the data transmission phase with probability

P iidle =

[

1− P if

(

εi, τ)]

P i (H0) + P im

(

εi, τ)

P i (H1) , (5)

whereP im (εi, τ) = 1 − P i

d (εi, τ) is the mis-detection probability. Otherwise, it will be silent for the

whole cycle and waits until the next cycle. This occurs with probability P ibusy = 1 − P i

idle. We assumethat interference of active primary users to secondary users is negligible; therefore, a transmission from

4


any secondary link only fails when it collides with transmissions from other secondary links. Now, letSk denote one particular subset of all secondary links having exactly n0 secondary links. There areCn0

N = N !n0!(N−n0)!

such setsSk. The probability of the event thatn0 secondary links join contention in thedata transmission phase can be calculated as

Pr (n = n0) =

Cn0

N∑

k=1

∏

i∈Sk

P iidle

∏

j∈S\Sk

Pjbusy, (6)

whereS denotes the set of allN secondary links, andS\Sk is the set of remainingN − n0 secondarylinks. If all secondary links have the sameSNRp and the same probabilitiesP i (H0) andP i (H1), thenwe haveP i

idle = Pidle andP ibusy = Pbusy = 1− Pidle for all i. In this case, (6) becomes

Pr (n = n0) = Cn0

N (Pidle)n0(1− Pidle)

N−n0 , (7)

where all terms in the sum of (6) become the same.

B. Calculation of Conditional Throughput

The conditional throughput can be calculated by using the standard technique developed by Bianchiin [12] where we approximately assume a fixed transmission probability φ in a generic slot time. It wasshown in [12] that for a fixed contention windowW , we haveφ = 2/ (W + 1). Suppose there aren0

secondary links participating in contention in the third phase, the probability of the event that at least onesecondary link transmits is data packet can be written as

Pt = 1− (1− φ)n0 . (8)

However, the probability that a transmission occurring on the channel is successful given there is at leastone secondary link transmitting can be written as

Ps =n0φ(1− φ)n0−1

Pt

. (9)

The average duration of a generic slot time can be calculatedas

T̄sd = (1− Pt)Te + PtPsTs + Pt (1− Ps)Tc, (10)

whereTe = σ, Ts andTc represent the duration of an empty slot, the average time thechannel is sensedbusy due to a successful transmission, and the average time the channel is sensed busy due to a collision,respectively. These quantities can be calculated as [12]

Ts = T 1s = H + PS + SIFS + 2 ∗ PD+ACK+DIFS

Tc = T 1c = H + PS +DIFS + PD

H = HPHY +HMAC

, (11)

whereHPHY andHMAC are the packet headers for physical and MAC layers,PS is the packet size, whichis assumed to be fixed in this paper,PD is the propagation delay,SIFS is the length of a short interframespace,DIFS is the length of a distributed interframe space,ACK is the length of an acknowledgment.Based on these quantities, we can express the conditional normalized throughput as follows:

T (τ, φ |n = n0 ) =⌊

Tcycle − τ

T̄sd

⌋ PsPtPS

Tcycle

, (12)

where⌊.⌋ denotes the floor function and recall thatTcycle is the duration of a cycle. Note that⌊

Tcycle−τ

T̄sd

⌋

denotes the average number of generic slot times in one particular cycle excluding the sensing phase.Here, we omit the length of the synchronization phase, whichis assumed to be negligible.

5


C. Optimal Sensing and MAC Protocol Design

Now, we turn to solve the throughput maximization problem formulated in (3). Note that we cancalculate the normalized throughput given by (4) by usingPr (n = n0) calculated from (6) and theconditional throughput calculated from (12). It can be observed that the detection probabilityP i

d (εi, τ)

in the primary protection constraintsP id (ε

i, τ) ≥ P̄ id depends on both detection thresholdεi and the

optimization variableτ .We can show that by optimizing the normalized throughput over τ and W while fixing detection

thresholdsεi = εi0 where P id (ε

i0, τ) = P̄ i

d, i = 1, 2, · · · , N , we can achieve almost the maximumthroughput gain. The intuition behind this can be interpreted as follows. If we chooseεi < εi0 for agiven τ , then bothP i

d (εi, τ) and P i

f (εi, τ) increase compared to the caseεi = εi0. As a result,P i

idle

given in (5) decreases. Moreover, it can be verified that the decrease inP iidle will lead to the shift of the

probability distributionPr (n = n0) to the left. Specifically,Pr (n = n0) given in (6) increases for smalln0 and decreases for largen0 whenP i

idle decreases. Fortunately, with appropriate choice of contentionwindow W the conditional throughputT (τ,W |n = n0 ) given in (12) is quite flat for differentn0 (i.e., itonly decreases slightly whenn0 increases). Therefore, the normalized throughput given by(4) is almosta constant when we chooseεi < εi0.

In the following, we will optimize the normalized throughput over τ andW while choosing detectionthresholds such thatP i

d (εi0, τ) = P̄ i

d, i = 1, 2, · · · , N . From these equality constraints and (2) we have

P if = Q

(

αi +√

τfsγi

)

(13)

whereαi =√2γi + 1Q−1

(

P̄ id

)

. Hence, the optimization problem (3) becomes independent of all detectionthresholdsεi, i = 1, 2, · · · , N . Unfortunately, this optimization problem is still a mixedinteger program(note thatW takes integer values), which is difficult to solve. In fact, it can be verified even if we allowWto be a real number, the resulting optimization problem is still not convex because the objective functionis not concave [13]. Therefore, standard convex optimization techniques cannot be employed to find theoptimal solution for the optimization problem under investigation.

Therefore, we have to rely on numerical optimization [14] tofind the optimal configuration for theproposed MAC protocol. Specifically, for a given contentionwindow W we can find the correspondingoptimal sensing timeτ as follows:

max0<τ≤Tcycle

NT(τ,W ) =N∑

n0=1

T (τ,W |n = n0 )Pr (n = n0). (14)

This optimization problem is not convex because its objective function is not concave in general.However, there exists an optimal solutionτ ∗ for this problem because the objective function is boundedfrom above (becausePr (n = n0) < 1 and conditional throughput are all bounded) and its objectivefunction increases in smallτ and decreases in largeτ . This second property is formally stated in thefollowing proposition.

Proposition 1: The objective functionNT(τ) of (14) satisfies the following properties

limτ→Tcycle

∂NT∂τ

< 0, limτ→0

∂NT∂τ

= +∞. (15)

Proof: Let us start the proof by defining the following quantities:ϕj := −(

αj+√

τfsγj

)

2

2and cn0

:=

6


PsPtPSTcycle

. Taking the derivative ofNT versusτ , we have

∂NT∂τ

=N∑

n0

cn0

CNn0∑

k=1

(

−1Tsd

)

∏

i∈Sk

P iidle

∏

j∈S\Sk

Pjbusy+

⌊

Tcycle−τ

Tsd

⌋√

fs8πτ

×

∑

i∈Sk

γi exp (ϕi)P i (H0)∏

l∈Sk\iP l

idle

∏

j∈S\Sk

Pjbusy

− ∑

j∈S\Sk

γj exp (ϕj)Pj (H0)∏

l∈S\Sk\jP l

busy

∏

i∈Sk

P iidle

. (16)

From this we have

limτ→Tcycle

∂NT∂τ

=N∑

n0

cn0

CNn0∑

k=1

(−1

Tsd

)

∏

i∈Sk

P iidle

∏

j∈S\Sk

Pjbusy < 0. (17)

Now, let us define the following quantity

Kτ:=N∑

n0

cn0

CNn0∑

k=1

∑

i∈Sk

γi exp (ϕi)P i(H0)∏

l∈Sk\iP l

idle

∏

j∈S\Sk

Pjbusy−

∑

j∈S\Sk

γj exp (ϕj)Pj(H0)∏

l∈S\Sk\jP l

busy

∏

i∈Sk

P iidle

. (18)

Then, it can be verified thatKτ > 0. Therefore, we have

limτ→0

∂NT∂τ

= +∞. (19)

Hence, we have completed the proof.In summary, one can find the globally optimal(W ∗, τ ∗) by finding optimalτ for eachW in its feasible

range[1,Wmax]. The procedure to find(W ∗, τ ∗) can be described in the following algorithm.

Algorithm 1 OPTIMIZATION OF COGNITIVE MAC PROTOCOL

1: For eachW ∈ [1,Wmax], find the optimalτ according to (14), i.e.,

τ(W ) = argmax0<τ≤Tcycle

NT(τ,W ) (20)

2: The globally optimal(W ∗, τ ∗) can then be found as

(W ∗, τ ∗) = argmaxW,τ(W )

NT(τ(W ),W ). (21)

Numerical studies reveal that this algorithm has quite low computation time for practical values ofWmax andTcycle.

Remark 1: The above analysis and optimization can be extended to the case where exponential backoffwith minimum contention windowWmin and maximum backoff stagem is employed. However, transmissionprobability φ needs to be found as a fixed point solution of two equations [12], which is used to calculateconditional throughput in Section III.B. In this case,Wmin and τ will be optimization variables.

Remark 2: We can extend the presented MAC protocol design to the case where there are multiplechannels. To exploit spectrum holes in this case, we assume that there is one control channel whichbelongs to the secondary network (i.e., it is always available) and M data channels whose states (i.e.,idle or busy) change over time. We further assume that each secondary user has two transceivers onewhich is always tuned to the control channel while the remainingone may be switched to different datachannels. We can design a multi-channel synchronized MAC protocol for this setting as follows. Eachcycle is still divided into 3 phases as before. However, in thesensing phase each secondary link needs to

7


8 16 32 64 128 256 512 1,0000.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Contention window (W)

Thro

ughput

(NT

)

Throughput vs contention window

N = 2

N = 5

N = 10

N = 15

N = 20

Fig. 3. Normalized throughput versus contention windowW for τ = 1ms and differentN .

0 0.5 1 1.5 2 2.5 3 3.5

x 10−3

0.4

0.45

0.5

0.55

0.6

0.65

0.7

0.75

0.8

0.85

0.9

Sensing time (τ )

Thro

ughput

(NT

)

Throughput vs sensing time

N = 30

N = 20

N = 2

N = 5

N = 10

Fig. 4. Normalized throughput versus the sensing timeτ for W = 32 and differentN .

sequentially perform spectrum sensing on all data channels. Therefore, by the end of the first phase eachsecondary linki has a list of available channelCi.

In the data transmission phase, active secondary links who find at least one available channel contendon the control channel to choose a data channel for their transmissions. Again, contention is performed byusing the CSMA-based random backoff mechanism and secondarylinks winning the contention exchangeRTS/CTS which contains information about one chosen data channel for data transmission as well asthe packet length to be transmitted (for other secondary users to update their network allocation vector(NAV)). Note that because all secondary users have one transceiver tuned to the control channel, theyalways know which channels are being used by other secondary users. Due to the space constraint,analysis and optimization of this multi-channel MAC protocol are not pursued further in this paper.

IV. N UMERICAL RESULTS

We present numerical results to illustrate throughput performance of the proposed cognitive MACprotocol. We take key parameters for the MAC protocol from Table II in [12]. Other parameters are chosenas follows: cycle time isTcycle = 100ms; mini-slot (i.e., generic empty slot time) isσ = 20µs; samplingfrequency for spectrum sensing isfs = 6MHz; bandwidth of PUs’ QPSK signals is6MHz. The signal-to-noise ratio of PU signals at secondary linksSNRi

p are chosen randomly in the range[−15,−20]dB.The target detection probability for secondary links and the probabilitiesP i (H0) are chosen randomly inthe intervals[0.7, 0.9] and [0.7, 0.8], respectively.

In Fig. 3, we show normalized throughputNT versus contention windowW for different values ofN when the sensing time is fixed atτ = 1ms for one particular realization of system parameters. Themaximum throughput on each curve is indicated by a star symbol. This figure indicates that the maximumthroughput is achieved at largerW for larger N . This is expected because larger contention windowcan alleviate collisions among active secondary for largernumber of secondary links. It is interesting to

8


00.5

11.5

x 10−3

0

0.02

0.04

0.060

0.2

0.4

0.6

0.8

1

Sensing time (τ )

Throughput vs transmission probability and sensing time

Transmissionprobability (φ)

Thro

ughput

(NT

)

0.2

0.3

0.4

0.5

0.6

0.7

0.8NT opt(0.0003, 0.0125) = 0.8542

Fig. 5. Normalized throughput versus sensing timeτ and transmission probabilityφ for N = 15.

02

46

8

x 10−4

0

0.02

0.04

0.060

0.2

0.4

0.6

0.8

1

Sensing time (τ )

Throughput vs transmission probability and sensing time

Transmissionprobability (φ)

Thro

ughput

(NT

)

0.2

0.3

0.4

0.5

0.6

0.7

0.8NT opt(0.0002, 0.0091) = 0.8546

Fig. 6. Normalized throughput versus sensing timeτ and transmission probabilityφ for N = 20.

observe that the maximum throughput can be larger than 0.8 althoughP i (H0) are chosen in the range[0.7, 0.8]. This is due to a multiuser gain because secondary links are in conflict with difference primaryreceivers.

In Fig. 4 we present the normalized throughputNT versus sensing timeτ for a fixed contention windowW = 32 and different number of secondary linksN . Again, the maximum throughput is indicated by astar symbol on each curve. This figure confirms the fact that the normalized throughputNT increaseswhenτ is small and decreases with largeτ as being proved in Proposition 2. In addition, the normalizedthroughputNT is not a concave function ofτ for largeN . Moreover, for a fixed contention windowoptimal sensing time indeed decreases with the number of secondary linksN . Finally, the multi-userdiversity gain can also be observed in this figure.

To illustrate the joint effect of contention windowW and sensing timeτ , we show the normalizedthroughputNT versusτ and transmission probabilityφ for N = 15 andN = 20 in Fig. 5 and Fig. 6,respectively. Recall that the transmission probability canbe calculated from contention windowW asφ = 2/(W + 1). We show the globally optimal parameters(φ∗, τ ∗) which maximize the normalizedthroughputNT of the proposed cognitive MAC protocol by a star symbol in these two figures. Thesefigures reveal that the performance gain due to optimal configuration of the proposed MAC protocolis very significant. Specifically, while the normalized throughputNT tends to be less sensitive to thetransmission probabilityφ, therefore the contention windowW , it decreases significantly when the sensingtime τ is deviated from the optimal valueτ ∗. Therefore, the proposed optimization approach would be

9


very useful in achieving the largest throughput performance for the secondary network.

V. CONCLUSION

In this paper, we have designed, analyzed, and optimized a MAC protocol for cognitive radio networksthat explicitly take into account spectrum sensing performance. Specifically, we have derived the normal-ized throughput of the proposed MAC protocol as a function ofsensing time, contention window andother system parameters. Then, we have showed how to choose detection thresholds for energy spectrumsensors, sensing time, and contention window to maximize the normalized throughput subject to protectionconstraints for primary receivers. In addition, we have presented numerical results to confirm importanttheoretical findings in the paper and show the significant performance gain achieved by the optimalconfiguration for proposed MAC protocol. Finally, several potential extensions including consideration ofexponential backoff and multi-channel scenario have been discussed.

REFERENCES

[1] C. Stevenson, G. Chouinard, L. Zhongding, H. Wendong, S. Shellhammer, W. Caldwell, “IEEE 802.22: The first cognitive radio wirelessregional area network standard,”IEEE Commun. Mag.,vol. 47, no. 1, pp. 130–138, Jan. 2009.

[2] L. Ying-Chang, Z. Yonghong, E. C. Y. Peh and H. Anh Tuan, “Sensing-throughput tradeoff for cognitive radio networks”,IEEE Trans.Wireless Commun., vol. 7, no. 4, pp. 1326-1337, April 2008.

[3] T. Yucek and H. Arslan, “A survey of spectrum sensing algorithmsfor cognitive radio applications,”IEE Commun. Surveys andTutorials, vol. 11, no. 1, pp. 116–130, 2009.

[4] J. Unnikrishnan and V. V. Veeravalli, “Cooperative sensing for primary detection in cognitive radio,”IEEE J. Sel. Areas SignalProcessing,vol. 2, no. 1, pp. 18–27, Feb. 2008.

[5] H. Kim and K. G. Shin, “Efficient discovery of spectrum opportunities with MAC-layer sensing in cognitive radio networks,”IEEETrans. Mobile Computing,vol. 7, no. 5, pp. 533-545, May 2008.

[6] H. Nan, T.-I. Hyon, and S.-J. Yoo, “Distributed coordinated spectrum sharing MAC protocol for cognitive radio,” inIEEE DySPAN’2007.[7] C. Cordeiro, and K. Challapali, “ C-MAC: A cognitive MAC protocol for multi-channel wireless networks,” inIEEE DySPAN’2007.[8] Y.R. Kondareddy, and P. Agrawal, “Synchronized MAC protocol for multi-hop cognitive radio networks,” inProc. IEEE ICC’2008.[9] A. C.-C. Hsu, D.S.L. Weit, and C.-C.J. Kuo, “A cognitive MAC protocol using statistical channel allocation for wireless ad-hoc

networks,” inProc. IEEE WCNC’2007.[10] H. Su, and X. Zhang, “Cross-layer based opportunistic MAC protocols for QoS provisionings over cognitive radio wireless networks,”

IEEE J. Sel. Areas Commun., vol. 26, no. 1, pp. 118–129, Jan. 2008.[11] J. Shi, E. Aryafar, T. Salonidis and E. W. Knightly, “Synchronized CSMA contention: Model, implementation and evaluation”, inIEEE

INFOCOM, pp. 2052-2060, 2009.[12] G. Bianchi, “Performance analysis of the ieee 802.11 distributed coordination function”,IEEE J. Sel. Areas Commun., vol. 18, no. 3,

pp. 535-547 Mar. 2000.[13] S. P. Boyd and L. Vandenberghe, Convex optimization, Cambridge University Press, Cambridge, UK ; New York, 2004.[14] S. Kameshwaran and Y. Narahari, “Nonconvex piecewise linearknapsack problems,”European J. Operational Research,vol. 192, no.

1, pp. 56-68, 2009.

10

MAC Protocol for Cognitive Radio Networks

Technology

Transcript of MAC Protocol for Cognitive Radio Networks