Joint Sensing Time and Power Allocation in Cooperatively Cognitive Networks-77K

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 3

IEEE COMMUNICATIONS LETTERS, VOL. 14, NO.

2, FEBRUARY 2010 163

Joint Sensing Time and Power Allocation in


Cooperatively Cognitive Networks
Chengshi Zhao, Student Member, IEEE, and Kyungsup Kwak, Member, IEEE

Abstract—The jointly optimal allocation of sensing time and Node 1 T1


power for a two-user amplify-and-forward overlay cognitive R1 T1 h10
network is developed by maximizing the averaged aggregate Node 0
throughput of the secondary network. In particular, observing h21 h12 R0
that the sensing duration lies within a strict interval, the jointly
optimal strategy of sensing time and power allocation is proved T2 R2
to be tractable by sequential optimization. h20
Node 2
Index Terms—Cognitive radio (CR), spectrum sensing, power T2
allocation, cooperative communications, optimization.
Fig. 1. Model for cooperative sensing and transmission.
I. I NTRODUCTION
Phase 0 Phase 1 Phase 2
N overlay cognitive radio (CR) networks [1], the pre-
I requisite of the access of secondary users (SUs) is no
interference to primary users (PUs), which makes spectrum
Node 1
Node 2
Sensing
Sensing
P1t
P2t
P1c
P2c
sensing a key role. In [2], authors studied the problem of τ ( T −τ ) / 2 ( T −τ ) / 2
designing the sensing duration to maximize the achievable
throughput for the secondary network under the constraint Fig. 2. Frame structure and power allocation of the considered cooperation.
that PUs are sufficiently protected. In [3], instead of selecting
fixed sensing time for the SU’s frames, the authors studied
spectrum-sensing policies that determine the spectrum sensing AF-cooperative cognitive network, without causing harmful
duration in order to optimize the SU’s performance, in which interference to the PUs and exceeding the power budget of
the energy cost of spectrum sensing was taken into account. each SU node.
A multiband joint detection framework was introduced in [4],
which jointly detects the primary signals over multiple bands II. S YSTEM M ODEL AND F ORMULATION
rather than over one band at a time. We consider a system in which two SUs (nodes 1 and 2)
On the other hand, the growing demand for reliable wish to cooperate in transmitting data to another SU (node 0)
spectrally-efficient wireless communication has led to the in- as shown in Fig. 1. Similar to the implementations proposed
terests in systems where nodes cooperate during the transmis- in [5][6][8], nodes 1 and 2 adopt orthogonal channels (e.g., in
sions [5]. The achievable rate regions of cooperative systems frequency domain) to transmit and relay the data. Therefore
were shown to be larger than those of conventional systems as depicted in Fig. 1, in the view of node 1 or node 2, each
without cooperation between the source nodes [6]. In [7], of them needs three channels (two channels for transmission
power allocation problem was investigated in an amplify-and- and one channel for receiving) to carry out the proposed
forward (AF) cooperative network considering the bit error cooperative transmission; while four channels are needed in
rate (BER) performances and the work regions of two coop- the view of the whole network.
erative users. In [8], a half-duplex AF scheme was proposed, As shown in Fig. 2, the same frame structure is adopted in
which was proved to be able to provide a larger achievable nodes 1 and 2, where phase 0 is for the cooperative sensing of
rate region than that provided by the original AF scheme. PUs’ states while phase 1 and phase 2 are for the cooperative
In this letter, our objective is to design the jointly opti- transmission of SUs’ data. � is the frame duration and � is
mal strategy of sensing time and power allocation so as to the sensing duration. In particular, in phase 1, nodes 1 and
maximize the averaged aggregate throughput of a two-SUs 2 transmit their self-data directly to the terminal node 0 with
Manuscript received October 25, 2009. The associate editor coordinating power �1� and �2� , respectively, and the data is also received
the review of this letter and approving it for publication was F. Jondral. by the partner; in phase 2, nodes 1 and 2 help their partner to
This research was supported by the MKE (The Ministry of Knowledge forward the data received in phase 1 to terminal node 0 with
Economy), Korea, under the ITRC (Information Technology Research Center)
support program supervised by the IITA (Institute for Information Technology power �1� and �2� , respectively.
Advancement)(IITA-2009-C1090-0902-0019).
The authors are with the Graduate School of IT and Telecom., Inha Univer-
sity, Rep. of Korea (e-mail: zhaochengshi@gmail.com; kskwak@inha.ac.kr). A. Cooperative Sensing
C. Zhao is also with the Wireless Network Lab, School of Information and The spectrum is sensed by SUs in phase 0 as illustrated in
Communications Engineering, Beijing University of Posts and Telecommuni-
cations, Beijing 100876, China. Fig 2. Following [4], we assume that the wideband channel
Digital Object Identifier 10.1109/LCOMM.2010.02.092102 is divided into � non-overlapping narrow subbands. In order
c 2010 IEEE
1089-7798/10$25.00 ⃝

Authorized licensd use limted to: IE Xplore. Downlade on May 13,20 at 1:53 UTC from IE Xplore. Restricon aply.
164 IEEE COMMUNICATIONS LETTERS, VOL. 14, NO. 2, FEBRUARY 2010

for the SUs to exploit the spectrum opportunities over the � total power) is used by both node 1 and node 2. Therefore,
subbands, wideband spectrum sensing is required to sense the the corresponding power allocation is ��� = �� and ��� =
presence of the PUs’ signal on each subband simultaneously. (1 − �)� . By applying maximal ratio combining (MRC), the
To decide whether the �th subband is occupied or not, the SNR achieved at terminal node 0 is the sum of the SNRs at
following binary hypotheses are tested: each phase. Motivated by the work in [7][8], the achievable
rate regions of node 1 and node 2 transmitting to node 0 are
�0,� : �� (�) = �� (�), (1a)
expressed as
�1,� : �� (�) = ℎ� �� (�) + �� (�) (1b) ( )
1 ��3 (1 − �)�2
where �0,� and �1,� represent the hypotheses that the PU is �1 (�) = log2 1 + ��1 + ,
2 1 + ��3 + (1 − �)�2
absent and present on subband �, respectively; �� (�) is the (5a)
�th received sample on subband �; ℎ� represents the channel 1
(
��4 (1 − �)�1
)
gain on the �th subband, which is assumed to undergo flat �2 (�) = log2 1 + ��2 +
2 1 + ��4 + (1 − �)�1
fading and is constant during the sensing period; �� (�) is (5b)
the �th symbol transmitted on subband �; and �� (�) is the
noise. where �1 = � ℎ10 /�02 , �2 = � ℎ20 /�02 , �3 = � ℎ12 /�22 and
Assume that all PUs’ signals are complex-valued PSK �4 = � ℎ21 /�12 . We further assume the flat fading channels
signal; the noise is the circularly symmetric complex Gaus- in cooperation intervals; the fading gains do not change and
sian noise. For a large number of samples � , by central are independent with each other and are independent for
limit theorem, the test statistic using energy detector can be each cooperation interval, thus ℎ�� is replaceable by ℎ̄�� .
approximated by normal distributions. For a target detection Considering the scenarios where full channel state information
probability ��� of subband �, the probability of false alarm (CSI) is available, the optimal power allocation in cooperative
can be expressed as [2] transmission is to find the desirable value of �.
(√ √ )
��� (� ) = � 2�� + 1�−1 (��� ) + � �� �� (2)
C. Joint Sensing Time and Power Allocation
∫ +∞ 2
where �(⋅) is defined as �(�) = √12� � � ��; �� is − �2 As mentioned above, in the view of node 1 or node 2,
the received signal-to-noise ratio (SNR) of the PUs’ signal each of them needs three available channels to carry out the
measured at the secondary receiver of the interest, under proposed cooperative transmission, while four channels are
hypothesis �1,� ; and �� is the sampling rate. needed in the view of the whole network. To make sure that
In the opportunistic spectrum access model, there are two there is no collision at terminal node 0, we request that nodes 1
scenarios for which a SU can operate at the PUs’ frequency and 2 have the same vision of four available channels. The four
band [2]: 1) Under hypothesis �0,� , when PUs are not present channels are allocated to these two nodes through the higher
and no false alarm is generated by the SU; 2) Under hypothesis layer protocol. Thus, it is requested that at least four common
�1,� , when the band is occupied by a PU but it is not correctly channels are available for both node 1 and node 2. In the
detected by the SU. The probability that subband � will be case of � channels/subbands in the cognitive network, there
�!
accessible for the SU is thus given by are � = (�−4)!4! possible scenarios for the SUs to use the
channels. By cooperative sensing, the probability that scenario
��� (� ) = � �(�0,� )(1 − ��� (� )) + � �(�1,� )(1 − ��� ) (3) � occurs in the views of nodes 1 and 2 is expressed as
where � �(�0,� ) and � �(�1,� ) represent the probabilities that

�Ψ� (� ) = {��� ,1 (� ) ⋅ ��� ,2 (� )} (6)
the PU is absent and present on subband �, respectively. �∈Ψ�

where � is the set of active channels in scenario �; ��� ,1


B. Cooperative Transmission and ��� ,2 , which correspond to (3), reflect the accessible
In phase 2, node 1 and node 2 amplify and forward the data probabilities of subband � for node 1 and node 2, respectively.
received in phase 1 by the factors In this letter, we are interested in maximizing the averaged
√ √ aggregate throughput of cooperative SUs for a prescribed PUs’
�1� �2� detection probability and the available power of each SU node.
�1 = , � 2 = (4)
�2� ⋅ ℎ21 + �12 �1� ⋅ ℎ12 + �22 Therefore, the problem can be formulated as
respectively to ensure that all the available relaying power is 1 ∑� −�
efficiently used [6], where ℎ�� captures the effects of path-loss, max �(�, �, Ψ) = �Ψ� (� )�� ; (7)
�,�,Ψ � �

shadowing and frequency nonselective fading between nodes
� ∈ {1, 2} and � ∈ {0, 1, 2}; ��2 is the variance of zero-mean s. t. ��� = ��� , � ∈ {1, 2, ..., �}; (8a)
Gaussian noise at node � ∈ {0, 1, 2}. We assume that ℎ�� �� = ��1 (�� ) + (1 − �)�2 (�� ); (8b)
√ distribution with mean ℎ̄�� ; therefore, the
follows exponential � ∈ (0, � ); (8c)
amplitude gain ℎ�� follows Rayleigh distribution.
�� ∈ [0, 1] (8d)
For simplicity, we set that the total power is fixed for each
user, i.e., ��� + ��� = � , where � is the total available power where �� is the cooperative ratio in scenario �; � = {�� },
for each user; and the same cooperative ratio � (which implies Ψ = {Ψ� }, � ∈ {1, 2, ..., �}; ��� is the target probability of
the ratio of the power used for self-data transmission over the detection on subband �; �� is the achievable throughput in

Authorized licensd use limted to: IE Xplore. Downlade on May 13,20 at 1:53 UTC from IE Xplore. Restricon aply.
ZHAO and KWAK: JOINT SENSING TIME AND POWER ALLOCATION IN COOPERATIVELY COGNITIVE NETWORKS 165

scenario �; � ∈ [0, 1] is the weighting factor to make the two

Ergodic throughput of the secondary network [bps/Hz]


SUs achieve the desirable rates. It should be noted that in a 1
Global Optimal (0.0141, 0.5435, 1.1189)
scenario, four available channels are allocated for the coop- 0.8
0.6
erative transmission, thus �� is achieved based on (8b). The 0.4
0.2
corresponding fading gains are substituted into the expressions 0

of �1 (�� ) and �2 (�� ) according to (5a) and (5b), respectively. 1


Local Optimal
0.8
Note that the throughput given in (7) takes into account 0.6

the statistics of the PUs’ activities and the imperfection of β 0.4


spectrum sensing (e.g., false alarm, miss detection). Basically, 0.2
0.08 0.1
0.06
it presents the throughput that the cooperation of two SUs can 0
0 0.02 0.04
τ [s]

achieve on average. We are not only interested in identifying


the optimal sensing duration � (which is independent of the Fig. 3. Achievable throughput for the cooperative secondary network.
variety of the scenarios since the sensing time, which is
correlated with the frame structure, is not easy to be changed
in a implemented secondary network), but also designing the series of sensing time within the interval (0, � ) and a series of
optimal power allocation strategies �� for each scenario �. cooperation ratio within the interval [0, 1]. It is seen from the
For any given � , the above optimization is equivalent to figure that the throughput is a concave function of the sensing
∑ time and the cooperation ratio �. A uniquely optimal sensing
max �(�) = �Ψ� (� ){��1 (�� ) + (1 − �)�2 (�� )}; time can be obtained to achieve the maximum throughput. The
�,Ψ
� unique cooperation ratio is also achieved in this figure because
s. t. ��� = ��� , � ∈ {1, 2, ..., �}; (9a) of our special set as listed above; if the SNR received at the
�� ∈ [0, 1]. (9b) SUs’ receivers are correlated to the channels’ states in diverse
scenarios, the different cooperation ratio �� will be achieved
Furthermore, once the scenario � has been determined, the in various scenarios.
optimal power allocation strategy (�� ) for the given scenario is The sensing time is a fixed parameter for the secondary
independent of the activities of the other scenarios. Hence, the network and it is observed in Fig. 3 that the throughput is a
optimization problem is equivalent to the following problem unimodal function of the sensing time, thus it can be computed
max �(�� ) = ��1 (�� ) + (1 − �)�2 (�� ); (10a) offline. We may use a more efficient search method to obtain
�� the optimal sensing time, e.g., bisection search, instead of
s. t. �� ∈ [0, 1]. (10b) using exhaustive search.
The result of the proposed cooperative method in secondary
It is easy to verify that �(�� ) is a concave function of �� , network is based on a simplified power allocation scheme
and the optimal value of �� can be archived by zeroing the during the cooperation. However, through the above analysis
first derivative of �(�� ) with respect to �� . Therefore, we can where the optimal sensing time is achievable by the exhaustive
conclude that the optimal power allocation vector � can be search and is relatively independent of the power allocation
achieved by resolving the optimization problem shown in (10) procedure; we may draw the conclusion that various power
for various scenarios; and the optimal sensing time can then allocation schemes adopted in the literatures [6]–[8] are all
be found by exhaustive search over (0, � ). compatible with and can be introduced into the proposed joint
optimization with minor appropriate modifications.
III. N UMERICAL R ESULTS AND C ONCLUSIONS
R EFERENCES
We use a fixed frame of � = 100ms for the secondary
network and the target probability of detection is set to [1] R. Menon, R. M. Buehrer, and J. H. Reed, “Outage probability based
comparison of underlay and overlay spectrum sharing techniques,” in
��� = 0.9 for all subbands. The PU is assumed to be a Proc. IEEE DYSPAN, pp. 101-109, Nov. 2005.
QPSK modulated signal with the bandwidth of 6MHz. The [2] Y.-C. Liang, Y. Zeng, E. C. Y. Peh, and A. T. Hoang, “Sensing-
sampling frequency is the same as the bandwidth of the PU. throughput tradeoff for cognitive radio networks,” IEEE Trans. Wireless
Commun., vol. 7, no. 4, pp. 1326-1337, Apr. 2008.
The weighting factor � is set to be 0.6. � = 10 subbands [3] A. T. Hoang, Y.-C. Liang, D. Wong, Y. Zeng, and R. Zhang, “Oppor-
are considered. There is a PU potentially using each channel tunistic spectrum access for energy-constrained cognitive radios,” IEEE
with the probability of � �(�1,� ) = 0.2. Correspondingly, the Trans. Wireless Commun., vol. 8, no. 3, pp. 1206-1211, Mar. 2009.
[4] Z. Quan, S. Cui, A. H. Sayed, and H. V. Poor, “Optimal multiband
received SNR of the PUs’ signal measured at node 1 and joint detection for spectrum sensing in cognitive radio networks,” IEEE
node 2 are set as [−20, −19, ..., −11] and [−11, −12, ..., −20], Trans. Signal Process., vol. 57, no. 3, pp. 1128-1140, Mar. 2009.
respectively. Without loss of generality, we assume that the [5] A. Sendonaris, E. Erkip, and B. Aazhang, “User cooperation diversity—
part I: system description,” IEEE Trans. Commun., vol. 51, no. 11, pp.
SNR received at the SUs’ receivers are only decided by the 1927-1938, Nov. 2003.
corresponding transmission power while independent of the [6] J. N. Laneman, D. N. C. Tse, and G. W. Wornell, “Cooperative diversity
channels selected in a scenario. Thus we set �1 = 6dB, in wireless networks: efficient protocols and outage behavior,” IEEE
Trans. Inf. Theory, vol. 50, no. 12, pp. 3062-3080, Dec. 2004.
�2 = 12dB, �3 = 20dB and �4 = 24dB (This leads the [7] L. Zhao and Z. Liao, “Power allocation for amplify-and-forward coop-
invariability of �� over various scenarios, however it does not erative transmission over rayleigh-fading channels,” J. Commun., vol. 3,
violate the correctness of the proposed optimization). no. 3, pp. 33-42, July 2008.
[8] W. Mesbah and T. Davidson, “Joint power and channel resource allo-
The simulation result is shown in Fig. 3.The curved surface cation for two-user orthogonal amplify-and-forward cooperation,” IEEE
is obtained by calculating the throughput numerically for a Trans. Wireless Commun., vol. 7, no. 11, pp. 4681-4691, Nov. 2008.

Authorized licensd use limted to: IE Xplore. Downlade on May 13,20 at 1:53 UTC from IE Xplore. Restricon aply.

You might also like