active cooperation between primary users and cognitive radio users in

ACTIVE COOPERATION BETWEEN PRIMARY USERS AND COGNITIVE RADIO USERS IN COGNITIVE AD-HOC NETWORKS Weifeng Su∗ , John D. Matyjas† , and Stella Batalama∗ ∗ Dept. of Electrical Engineering, State University of New York at Buffalo, NY 14260 † Air Force Research Laboratory/RIGF, Rome, NY 13441 Emails: [email protected], [email protected], [email protected]

ABSTRACT In this work, a cognitive cooperative communication protocol is proposed for cognitive ad-hoc networks, in which primary users and cognitive radio (CR) users may cooperate for mutual benefit. The new cooperation protocol allows active cooperation between primary users and CR users in which CR users assist to relay primary users’ signals in exchange for some spectrum released from the primary users. While conventional cognitive radios do not guarantee continuous operation of CR users (they use the spectrum only when primary users do not), the protocol proposed in this work provides continuous service for CR users. The proposed cognitive cooperation protocol is optimized in terms of maximizing the primary user’s energy savings and the CR user’s own data transmission rate. It turns out that the primary users have significantly average energy savings from cooperation (e.g. up to 50% when compared to a non-cooperation case at the same transmission power level), which provides a good incentive for they to cooperate. Index Terms— Cooperative communications, cognitive radio, ad-hod networks. 1. BACKGROUND AND MOTIVATION Cognitive radio (CR) networks have attracted tremendous interest in recent years, due to a basic idea that if spectrum is not used by primary users, secondary users may use it based on cognitive radio technologies [1], [2], [3]. Network spectrum efficiency can be greatly improved as secondary users or CR users may sense and exploit “spectrum holes” whenever they are available. Primary users refer to licensed users who own the spectrum and secondary/CR users refer to non-licensed users, or in some applications, primary users refer to users with higher priority to use the spectrum and secondary/CR users have lower priority. Cooperative communications is an emerging concept where signal transmissions are optimized both at the physical layer and at the medium-access control (MAC) layer (see for example, [4]-[6] and the references therein). There are three ways to apply the cooperative communication concept in cognitive radio networks: i) cooperation among primary user peers; ii) cooperation among CR user peers; and iii) cooperation between primary users and CR users. The first category is trivial as its solution is reduced to the traditional

978-1-4244-4296-6/10/$25.00 ©2010 IEEE

3174

cooperative communications. For the second category, the cooperative communication concept can be applied straightforwardly among CR users with available dynamic spectrum [7], [8]. For the third category, since primary users and CR users have different priorities and they may have security concerns for their own data, cooperation between them becomes challenging. In [9], under the assumption that CR users know perfectly the data of primary users, it was shown that maximum rate can be achieved by simultaneous transmission of primary and CR user data over the same frequency, where CR user data are jointly encoded with primary user data via dirtypaper coding techniques. In [10], a more realistic scheme was proposed where CR users utilize spectrum holes only whenever available and help forward primary user data packets that have not been successfully received by an intended destination. The scheme was further generalized in [11] so that CR users while forwarding primary users’ unsuccessful packets, they embed and transmit their own data also based on dirty-paper coding techniques. In [12], it was proposed to deploy a “dumb” relay node to help relay primary or CR user signals to improve spectrum efficiency. In this work, we consider active cooperation between primary users and CR users. With respect to past literature, in our work we do not consider the assumption in [9] that primary user data is perfectly decoded and known to CR users, which is often not realistic due to security concerns. In addition, we propose an active cooperation scheme between primary users and CR users which allows continuous service/operation of the CR users in the networks. This is contrast to existing work in [10]-[12] where CR users transmit their own signals only if primary users are idle, resulting in CR data transmission that may not be continuous which may have catastrophic implications in time sensitive applications. Fig. 1 illustrates how transmission of primary user (PU) data can be achieved by the proposed protocol when a CR user/relay is available to assist (in aim to exchange some spectrum released from the primary user). Let T denote the time duration that the primary user is allowed to transmit data over bandwidth W Hz. If a CR user ia available to assist, then the primary user may decide to transmit in only a part of the time slot (e.g. T2 ) over a portion of the bandwidth, W1 < W , hoping that cooperation with a CR user and data relaying for the rest of the time slot may result in successful delivery of

ICASSP 2010

%DQGZLGWK

T 2

:

38

:

&5

T 2

&5

Fig. 1. Time-bandwidth allocation for PU and CR users. primary user data to the intended destination (to what extent and under what condition this can be achieved as well as what the overall benefits are will be studied in subsequent sections). In general, the above scheme leads to power and on-air time savings for the primary user. On the other hand, the CR user assists to relay primary user signal in exchange for some bandwidth, W2 Hz, released by the primary user for CR user own data transmission. 2. SYSTEM MODEL AND THE PROPOSED COGNITIVE COOPERATION PROTOCOL For convenience in presentation, we consider a basic fournode configuration consisting of a pair of primary users and a pair of CR users. We assume that the bandwidth owned by the primary users is W Hz. Foe each primary user, we also assume that the target transmission rate is RP U bits/s and the transmission power per unit frequency is P1 watts/Hz. If CR users are available to assist with relaying primary user data, primary users may release some spectrum to CR users for their own use. Let us assume that a primary user utilizes bandwidth W1 Hz and releases bandwidth W2 Hz to CR users (W1 + W2 = W ). Let x denote the signal transmitted by a primary user (source). The transmitted signal may be received by both the intended primary user (destination) and a nearby CR user (relay). The received signal√at the destination and at the relay √ is denoted by ys,d = P1 hs,d x + ηs,d , and ys,r = P1 hs,r x + ηs,r , respectively, where hs,d and hs,r denote the channels between the primary source and destination and between the source and the CR relay, correspondingly. The terms ηs,d and ηs,r represent additive Gaussian noise and are modeled as zero-mean complex Gaussian random variables, with an average power per unit frequency N0 = 4.0 × 10−21 watts/Hz, i.e. −174dBm. We propose to utilize amplify-and-forward (AF) type cooperation/relaying where CR users simply amplify and forward primary user signals since any decoding/encoding at the relay may compromise security of primary user data. Let us denote the CR relaying power as P2,1 watts/Hz (power per unit frequency), then the received signal at P2,1 βhr,d ys,r + ηr,d , where the destination is yr,d =  β = 1/ P1 |hs,r |2 + N0 is an amplification factor, hr,d is the channel between the CR user and the primary user’s destination, and ηr,d is additive Gaussian noise with variance N0 . The received signal at the destination is the aggregate

3175

of the signals received from the primary user directly and the CR relay which are jointly decoded by using maximumratio combining (MRC) [6]. The channels hs,d , hs,r and hr,d are assumed to be independent complex Gaussian random 2 2 2 , δs,r and δr,d , variables with mean zero and variance δs,d respectively. The channel variance depends on the distance 2 = ( 4πdλi,j )γ , of the channel link according to the rule δi,j where di,j is the distance of a link from i to j, λ is the carrier wavelength, and γ is the path loss coefficient. The capacity of the cooperative AF relay channel over bandwidth W1 Hz is [6]   P1 |hs,d |2 W1 log2 1+ +f (P1 |hs,r |2 , P2,1 |hr,d |2 ) , Ccoop = 2 N0 (1) P1 P2,1 |hs,r |2 |hr,d |2 1 2 2 where f (P1 |hs,r | , P2,1 |hr,d | ) = N0 P1 |hs,r |2+P2,1 |hr,d |2+N0 . We note that the factor 1/2 in (1) is due to the fact that the primary user only uses the first half of the time slot to transmit signals and the CR user utilizes the second half of the time slot to forward signals (both over bandwidth W1 Hz). To meet the primary user’s target transmission rate, Ccoop should not be less than RP U bits/s, i.e. Ccoop ≥ RP U , which implies that the bandwidth needed for the primary user should satisfy W1 ≥



2RP U

P |h |2 log2 1+ 1 Ns,d 0

 . (2) +f (P1 |hs,r |2 , P2,1 |hr,d |2 )

Then, the remaining bandwidth, (W − W1 )Hz, may be released to CR users. Without any CR relaying help, the primary user original transmission power P0(watts per unit frequency) must satisfy  P |h |2 = RP U , i.e. W log2 1 + 0 Ns,d 0  N0  RP U W 2 − 1 , (3) P0 = |hs,d |2 and for total transmission time T , the primary user original energy consumption is P0 W T joules. On the other hand, when CR relaying takes place, the primary user transmits only in the first half of the time slot, so the energy consumption is P1 W1 T2 , and the rate of primary user energy savings is Φ=

P0 W T − P1 W1 T2 P1 W1 =1− . P0 W T 2P0 W

(4)

At the same time, the CR user can use the released bandwidth W2 = W −W1 Hz for its own data transmission. Let us assume that the CR user has an energy budget equal to ECR joules, and let us denote by P2,2 watts/Hz the transmission power level of the CR user dedicated to its own data transmissions. Then, P2,2 should satisfy P2,1 W1 T2 + P2,2 W2 T = ECR , or equivalently, 1 P2,1 W1 + P2,2 W2 = PCR , (5) 2 where PCR is the average transmission power of the CR user. Thus, the CR (own) data transmission rate is

 RCR = W2 log2

 P2,2 2 1+ |hCR | , N0

(6)

dCR = 500m, α opt = 0.61

C R own da ta ra te (Mbits/s)

1

dCR = 1Km, α opt = 0.56 0 .8

dCR = 2Km, α opt = 0.50

0 .6

0 .4

0 .2

3. PROTOCOL OPTIMIZATION In this section, first we determine the optimum allocation scheme for the CR user to allocate relaying power based on its own power budget, and then we address primary user’s choice in cooperation. 3.1. CR User Relaying Power Optimization For any given CR power budget PCR , let α denote the ratio of the CR user power allocated to assist relaying primary user signals over the CR power budget, i.e. 1 P2,1 W1 = αPCR , and P2,2 W2 = (1 − α)PCR . (7) 2 The CR user may choose any value (from 0 to 1) for the power ratio α to allocate relay power. However, if α is too small, the allocated relaying power may not be enough to trigger cooperation. On the other hand, if α is large, the remaining power may be not enough for CR user’s own data. For any ratio α ∈ [0, 1], based on (2) and the fact that 1 P W1 = αPCR , the corresponding relaying power P2,1 2,1 2 can be determined by the following equation P2,1 RP U

 = αPCR . + f (P1 |hs,r |2 , P2,1 |hr,d |2 ) (8) We note that the left-hand side of the equation is increasing in terms of P2,1 . When P2,1 is zero, the left-hand side of the equation is zero, and when P2,1 goes to infinity, the lefthand side of the equation also goes to infinity. Therefore, there exists a unique solution P2,1 for the equation (8). With P2,1 determined, according to (2), a corresponding minimum bandwidth W1 can be determined (and so can W2 ). Then, from (7), the power P2,2 that is used for CR user’s own data CR . Then, transmission can be determined as P2,2 = (1−α)P W2 (6) implies that the CR user’s own data rate is given by   (1 − α)PCR |hCR |2 (9) RCR = (W − W1 ) log2 1 + (W − W1 )N0 log2 1 +

CR

1 .2

where hCR is the channel between the CR user and its own destination. We observe that the more of its power the CR user allocates to the relaying process, the more bandwidth gets released to the CR user who, now, has less remaining power for its own data transmission. So, the CR user has to determine how much power should be allocated to the relaying process in order to maximize its own data rate.



PU tota l ba ndwidth W = 1 MH z, C R powe r budge t P = 1 0 dBm CR

1 .4

P1 |hs,d |2 N0

which can be optimized with respect to α. Fig. 2 illustrates the rate of the CR user for the transmission of its own data in terms of varying power ratio α (0 ≤ α ≤ 1). We assume that the primary user has total bandwidth W = 1 MHz, and the CR user has a power budget PCR =10dBm. In this example, we consider a scenario where

0

0

0 .2 0 .4 0 .8 0 .6 R a tio of re la ying powe r ove r C R powe r budge t, α

1

Fig. 2. Maximization of CR user (own) data rate in terms of ratio of relaying power over CR total power budget. the primary user maintains the same transmission power level (i.e. P1 = P0 ) based on (3). We assume that the distance between the primary user and its destination is 1Km, the distance between the primary user and the relaying CR user is 500m, and the distance between the CR user and the destination is 1Km. We plot the CR user own data rate for three different values of the distance between the CR user and its own destination, namely to be 500m, 1Km and 2Km. The optimum ratio of the relaying power over the total power budget is 61%, 56% and 50%, respectively. The corresponding energy savings for the primary user is 58%, 57.8% and 57.3%, correspondingly. 3.2. Primary User’s Choices in Cooperation The primary user, as spectrum owner, is the one who decides how much spectrum to release and what level of its own transmission power to use. Let the transmission power that the primary user decides to use be equal  to P1 = ζP0 RP U N0 W watts/Hz, where P0 = |hs,d |2 2 − 1 watts/Hz is its original transmission power without cooperation and ζ is a parameter to be determined. In the following, we determine the range of the parameter ζ and examine its impact on the primary user energy savings. First, we determine a minimum meaningful value that parameter ζ can take. We observe that below a certain transmission power, the target data rate of the primary user cannot be guaranteed no matter how much the CR user efforts is in relaying. Since the total bandwidth is W Hz, based on (2), we 2RP U . have W ≥ W1 ≥ ζP0 |hs,d |2 log2 [1+

N0

+f (ζP0 |hs,r |2 ,P2,1 |hr,d |2 )]

Thus, we have a constraint on ζ as follows

2RP U ζP0 |hs,d |2 + f (ζP0 |hs,r |2 , P2,1 |hr,d |2 ) ≥ 2 W − 1. (10) N0

ζP |h

|2

|h

|2

Since f (ζP0 |hs,r |2 , P2,1 |hr,d |2 ) < min{ 0N0s,r , 2,1N0r,d } and f (ζP0 |hs,r |2 , P2,1 |hr,d |2 ) is an increasing function in terms of P2,1 , so when the CR relaying power P2,1 is high enough, f (ζP0 |hs,r |2 , P2,1 |hr,d |2 ) converges to

3176

P

ζP0 |hs,r |2 . N0

ζ ≥ ζmin

≥2

2RP U W

− 1. So,

 RP U  |hs,d |2 W 2  + 1 . |hs,d |2 + |hs,r |2

(11)

Therefore, the primary user transmission power should not be less than ζmin P0 Watts/Hz. Next, we evaluate the maximum value of the parameter ζ that the primary user may choose to determine its transmission power. Beyond that value, the primary user cannot save any energy from cooperation, and for this reason there is no incentive to cooperate with the CR user. Since the average transmission power of the primary user over the bandwidth W1 is 12 P1 W1 which should not be larger than P0 W , ζP0 RP U so ≤ P0 W . Since ζP0 |hs,d |2 log2 [1+

N0

+f (ζP0 |hs,r |2 ,P2,1 |hr,d |2 )]

ζP |h

|2

f (ζP0 |hs,r |2 , P2,1 |hr,d |2 ) is upper bounded by 0N0s,r , we ζ E ≤ W and substitution of D have ζP RP U log2 1+ N 0 (|hs,d |2 +|hs,r |2 )  0R  PU 0 P0 = |hN 2 W − 1 leads to another constraint for ζ: 2 s,d |   log2 1 + ζ 1 +

ζ |hs,r |2 |hs,d |2



2

RP U W

W  ≤ . R PU −1

(12)

Note that the left-hand side of the above inequality is an increasing function of ζ. There exists a unique solution for the parameter if we consider equality in (12). Let us denote the solution as ζmax , which is the maximum parameter that the primary user may choose. So the maximum transmission power that the primary user can use without compromising any power savings gained by cooperation is ζmax P0 watts/Hz. In Fig. 3, we plot the primary user energy savings in terms of the power parameter ζ (ζmin ≤ ζ ≤ ζmax ). We assume that the distance between the primary user and the CR user is 500m, and that the other channels have distance of 1Km. We consider the 10GHz frequency band and W = 1MHz. We observe that the primary user energy saving can be up to 70%. The smaller the power parameter, the more the energy savings of the primary user and the less the benefits for the CR user. On the other hand, the larger the parameter, the less the power savings of the primary user and the more the benefits for the CR user. We also plot the maximum rate of the CR user’s own data as a function of the power parameter ζ. We can see that the CR user’s own data rate increases when the primary user transmission power level is increased. 4. CONCLUSIONS In this work, we proposed a cognitive cooperation protocol for cognitive ad-hoc networks, which allows active cooperation between primary and CR users that can provide continuous service/operation for the CR users. From the primary user perspective, as long as its data is successfully delivered, it doesn’t matter it uses all of its bandwidth or only partial bandwidth. We optimized the cognitive cooperation protocol

3177

PU e ne rgy sa ving, %

|2 +|hs,r |2 )

Ma ximum ra te for C R da ta (Mbits/s)

ζP (|h

Thus, (10) implies that 0 s,dN0 the parameter ζ is lower bounded as

80 60 40

PCR = 10dBm

20 0 0 .5

1

PCR = 20dBm

1 .5 2 2 .5 3 3 .5 4 .5 4 Prim a ry us e r tra ns . powe r le ve l, ζ ( ζm in ≤ ζ ≤ ζm a x )

5

5 4

PCR = 20dBm

3 2

PCR = 10dBm

1 0 0 .5

1

1 .5 2 2 .5 3 3 .5 4 .5 4 Prim a ry us e r tra ns . powe r le ve l, ζ ( ζm in ≤ ζ ≤ ζm a x )

5

Fig. 3. Primary user energy savings and corresponding maximum rate for CR user’s own data in terms of ζ. in terms of maximizing the primary user energy savings and the CR user’s own data transmission rate. We developed an optimum allocation scheme for the CR user to allocate relaying power based on its own power budget and the relationship between relaying power needed and the corresponding spectrum to be released, in order to maximize the CR user’s own data transmission rate. Numerical results show that the primary user enjoys significant average energy savings as a result of the cooperation with CR users. 5. REFERENCES [1] Federal Communications Commission, “Notice of Proposed Rule Making and Order,” ET Docket No. 03-322, Dec. 2003. [2] J. Mitola III and G. Q. Maguire, Jr., “Cognitive radio: making software radio more personal,” IEEE Pers. Comm., pp.13-18, Aug. 1999. [3] F. K. Jondral, “Cognitive radio: a communications engineering view,” IEEE Wireless Communications, pp.28-33, August 2007. [4] J. N. Laneman, D. N. C. Tse, and G. W. Wornell, “Cooperative diversity in wireless networks: efficient protocols and outage behavior,” IEEE Trans. Inform. Theory, vol. 50, no. 12, pp.3062-3080, Dec. 2004. [5] A. Sendonaris, E. Erkip, and B. Aazhang, “User cooperation diversityParts I and II,” IEEE Trans. Comm., vol. 51, pp.1927-1948, Nov. 2003. [6] K. J. R. Liu, A. Sadek, W. Su, and A. Kwasinski, Cooperative Communications and Networking, Cambridge Univ. Press, New York, 2009. [7] J. Jia, J. Zhang, and Q. Zhang, “Cooperative relay for cognitive radio networks,” in Proc. IEEE INFOCOM, pp.2304-2312, Apr. 2009. [8] X. Gong, W. Yuan, W. Liu, W. Cheng, and S. Wang, “A cooperative relay scheme for secondary communication in cognitive radio networks,” in Proc. IEEE GLOBECOM, pp.1-6, Nov. 2008. [9] N. Devroye, P. Mitran, and V. Tarokh, “Achievable rates in cognitive radio channels,” IEEE Info. Theory, vol. 52, pp.1813-1827, May 2006. [10] O. Simeone, Y. Bar-Ness, and U. Spagnolini, “Stable throughput of cognitive radios with and without relaying capacity,” IEEE Trans. Comm., vol. 55, no. 12, pp.2351-2360, Dec. 2007. [11] Y. Chen, H. Huang, Z. Zhang, P. Qiu, and V. K. N. Lau, “Cooperative spectrum access for cognitive radio network employing rateless code,” in Proc. IEEE ICC, pp.326-331, May 2008. [12] I. Krikidis, Z. Wei, J. N. Laneman, and J. Thompson, “Cognitive legacy networks via cooperative diversity,” IEEE Comm. Letters, vol. 13, no. 2, pp.106-108, Feb. 2009.