Professional Documents
Culture Documents
Base Paper
Base Paper
Abstract
The natural acoustic system used by marine mammals and the artificial sonar system used by humans coexist in the
underwater cognitive sonar communication networks (CSCN). They share the spectrum when they are in the same
waters. The CSCN detects the natural acoustic signal depending on cooperative spectrum sensing of sonar nodes. In
order to improve spectrum sensing performance of CSCN, the optimization of cooperative spectrum sensing and data
transmission is investigated. We seek to obtain spectrum efficiency maximization (SEM) and energy efficiency
maximization (EEM) of CSCN through jointly optimizing sensing time, subchannel allocation, and transmission power.
We have formulated a class of optimization problems and obtained the optimal solutions by alternating direction
optimization and Dinkelbach’s optimization. The simulation results have indicated that SEM can achieve higher
spectrum efficiency while EEM may get higher energy efficiency.
Keywords: Cognitive radio, Cooperative spectrum sensing, Spectrum efficiency, Energy efficiency
© The Author(s). 2017 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and
reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the
Creative Commons license, and indicate if changes were made.
Liu and Jia EURASIP Journal on Wireless Communications and Networking (2017) 2017:171 Page 2 of 8
previous works focused on optimizing either cooperative Then, in the cooperative interaction slot, each SU uses
sensing or data transmission to maximize the spectrum one common channel to report the local sensing infor-
efficiency of the CSCN. For instance, the listening-before- mation to the fusion center, which makes a final decision
transmitting spectrum access is proposed, which can on the activation of the PU by the soft-decisional com-
maximize the spectrum efficiency through optimizing bination of the local sensing information. To avoid the
sensing time [9]; both [10] and [11] assumed that each mutual interference, the SUs cannot report from the com-
SU had a fixed transmission power, which ignored the mon channel simultaneously at the same time. Hence,
potential gain of dynamic resource optimization. The in order to save the bandwidth of the common channel,
water filling algorithm could maximize the spectrum effi- the SU adopts time division multiple access (TDMA) to
ciency of multichannel CSCN through optimizing sub- report the local sensing information. Finally, in the data
channel power [12]. Recently, green communications have transmission slot, the SUs access the licensed spectrum
attracted the attentions of the scholars. Energy efficiency to communicate according to the decision result of the
has been proposed to measure the effectiveness of energy fusion center.
utilization [13].
In this paper, we investigate the joint parameter opti- 2.1 Underwater acoustic channel model
mization of cooperative spectrum sensing and data trans- The fusion center maintains a collection of free channels
mission for multichannel CSCN. The contributions of the and the corresponding channel gain matrix in the under-
paper are listed as follows: water acoustic region by scanning and spectrum sensing.
Supposing that hn,l is the channel gain of SU n using sub-
• We seek to maximize spectrum efficiency and energy
channel l, the channel gain matrix H, can be given by
efficiency of CSCN, respectively, while considering
both spectrum sensing performance for detecting the ⎛ ⎞
nature acoustic system and the power constraint of h11 h12 · · · h1N
⎜ h h ··· h ⎟
each sonar node. ⎜ 21 22 2N ⎟
• We have formulated the cooperative spectrum H=⎜
⎜ .. .. . . . ⎟⎟ (1)
⎝ . . . .. ⎠
sensing and data transmission optimization as a class
of optimization problems about sensing time, hL1 aL2 · · · hLN
subchannel allocation, and transmission power. In
the optimization problem, the nature acoustic system Due to the reflection of the sea surface, submarine,
is fully protected, while the transmission performance and underwater medium, there are complicated multi-
of the CSCN is improved as much as possible. path propagation in shallow sea acoustic channel, which
• The joint optimization algorithm is proposed to can be considered as generalized uncorrelated scattering
obtain the solutions to the optimization problems, condition in general wireless channel. When the number
which is based on the alternating direction of multipaths is relatively large, the underwater acous-
optimization and the Dinkelbach optimization. tic channel obeys the Rayleigh distribution. Thus, the
channel gain distribution is given by
2 System modeling
1 hn,l
We consider L nature acoustic systems occupying L sub- fhn,l hn,l = exp − (2)
channels, which are seen as PUs and a CSCN constituting h̄n,l h̄n,l
of N sonar nodes, which are seen as SUs. We also suppose
that in an artificial acoustic system, there is an access point where h̄n,l is the statistical average of hn,l .
that controls the channel usage of the SUs in the locat-
ing water, which can be seen as a fusion center. In order 2.2 Cooperative spectrum sensing
to avoid causing harmful interference to the PU, the SU The sensing signal of SU n for n = 1, 2, . . . , N in sub-
has to detect the utilization state of the licensed spectrum channel l for l = 1, 2, . . . , L, yn,l (m), is denoted as follows:
and decide the activation of the PU. Hence, the periodic
cooperative spectrum sensing based on “listening before
yn,l (m) = θl hn,l psl (m) + σl2 (m), m = 1, 2, . . . , M (3)
transmitting” is proposed in this paper, which divides the
communication time into several frames, each of which where θl = 0 and θl = 1 denote the absence and presence
constitutes of three time slots including local sensing slot, of the PU in subchannel l, respectively; hn,l denotes the
cooperative interaction slot, and data transmission slot, as subchannel gain from SU n to the PU; psl is the transmis-
shown in Fig. 1. In the local sensing slot, all the SUs sense sion power of the PU in subchannel l; σl2 is the power of
the activation of the PU in L subchannels and obtain the the noise in subchannel l; and M is the number of the sam-
local sensing information. plings. M is given by M = τ fs , where τ is the local sensing
Liu and Jia EURASIP Journal on Wireless Communications and Networking (2017) 2017:171 Page 3 of 8
time and fs is the sampling frequency. We can get the must be guaranteed. The sensing threshold λ can be
energy statistic of the sensing signal yn,l by accumulating obtained by fixing Qdl with some value, which is given as
the energy of M samplings as follows: follows:
⎡ ⎤
1
M (γ̄l + 1)2
ψn,l = yn,l (m)2 (4) λ = ⎣Q−1 Qdl + γ̄l + 1⎦ σl2 (7)
M Nτ fs
m=1
Each SU reports ψn,l to a fusion center, which combines The selection of sensing threshold is important to
ψn,l of N SUs to get an overall energy statistic of the PU the sensing performance. The detection probability will
signal in subchannel l as follows: improve with the increasing of sensing threshold; how-
ever, the false alarm probability will also increase, thus
1 1
N N M decreasing the spectrum access probability. Through
l = ψn,l = yn,l (m)2 (5) f
removing the threshold λ, Ql is denoted by Qdl as follows:
N MN
n=1 n=1 m=1
Ql = Q Q−1 Qdl (γ̄l + 1) + γ̄l Nτ fs
f
The overall energy statistic l is compared with a thresh- (8)
old λ. If l ≥ λ, the presence of the PU is detected;
otherwise, the absence of the PU is determined. When
M is large enough, l obeys the Gaussian distribution In underwater, since the channel between the signal
approximatively according to the central limit theorem. source and the sonar is often in severe fading, the received
Hence, the cooperative probabilities of false alarm and signal by one sonar may be too weak to be detected accu-
detection for subchannel l is given as follows: rately. However, with cooperative spectrum sensing, the
same signal can be received by multiple sonars from dif-
ferent paths. If the received signal by one sonar is too
f λ
Ql =Q σl2
−1 Nτ fs weak, the detection performance cannot be decreased
(6) through sharing the sensing information with the other
λ Nτ fs
Qdl = Q σl2
− γ̄l − 1 (γ̄ +1)2
sonars. Thus, the sensing diversity gain can be achieved to
l
improve the final detection performance through cooper-
ative spectrum sensing.
1 N psl h2n,l
where γ̄l = N n=1 σ 2 is the average sensing SNR
l 2.3 Spectrum efficiency maximization (SEM)
of N SUs in subchannel
l, and the function Q(x) =
1
+∞ z 2
Spectrum efficiency is an important indicator to eval-
√ exp − 2 dz. uate the spectrum sensing performance. In spectrum
2π x
In order to avoid causing harmful interference to other sensing, the decreasing of detection performance may
underwater acoustic systems, the detection probability decrease the spectrum access opportunity of the SU
Liu and Jia EURASIP Journal on Wireless Communications and Networking (2017) 2017:171 Page 4 of 8
because of the increasing false alarm probability; how- ωn,l gn,l
2
an,l pn,l . Then, we have Rn,l = an,l log 1 + an,l σl2
. From
ever, improving detection performance needs to increase
f
the sensing time, which decreases the data transmission (8) and (9), ηSE improves with the decreasing of Ql while
time. Hence, we need to optimize the sensing parameters f
Ql reduces with the decreasing of Qdl . Hence, ηSE may
to maximize the spectrum efficiency. The SU can trans-
reach the maximum only when Qdl = Qmind . The optimiza-
mit data effectively only when the absence of the PU is
tion problem is rewritten as follows:
detected accurately. The accurate idle detection proba-
f
bility of subchannel l is Pr (θl = 0) 1 − Ql , and the T − τ − Nε
N L
SU n. Supposing the frame duration is T and the length of s.t. ωn,l ≤ p̂max
n , n = 1, 2, . . . , N (11b)
each cooperative interaction slot is ε, the overall spectrum l=1
efficiency of N SUs over L subchannels is given as follows: N
an,l = 1, an,l ∈[ 0, 1] , l = 1, 2, . . . , L (11c)
n=1
T − τ − Nε
N L
0 ≤ τ ≤ T − Nε (11d)
ηSE =
T
n=1 l=1 ωn,l ≥ 0, n = 1, 2, . . . , N, l = 1, 2, . . . , L (11e)
2
pn,l gn,l
f
× Pr (θl = 0) 1 − Ql an,l log 1 + where κ = Q−1 (Qmind )(γ̄l +1) and p̂n
max = pmax −p . Then,
n c
σl2 we will give the following theorem.
(9)
Theorem 1 Problem (11) is a convex optimization
Our goal is to maximize the spectrum efficiency of the SU problem.
by jointly optimizing sensing time, subchannel allocation,
and transmission power, subject to the constraints that the Proof We often let Qlf ≤ 0.5 and then have κ +
detection probability is above the lower limit Qmin
d , the
γ̄l Nτ fs ≥ 0. Fixing Rn,l , We can get the secondary
total power of SU n is below the maximal power pmax
n , and
derivative of ηSE in τ as follows:
one subchannel is only allocated to one SU. Supposing the
Pr (θl = 0) Nfs
spectrum sensing power is pc , the optimization problem N L
∂ 2 ηSE
of SEM is given as follows: = − √
∂ 2τ T 2πτ n=1 l=1
⎡ ⎛ 2 ⎞⎤
max ηSE (10a) κ + γ̄l Nτ fs
τ ,{an,l },{pn,l } ⎢ ⎜ ⎟⎥
× ⎣γ̄l Rn,l exp ⎝− ⎠⎦
2
s.t. Qdl ≥ Qmin
d , l = 1, 2, . . . , L (10b)
L
T − τ − Nε
N L
an,l pn,l + pc ≤ pmax
n , n = 1, 2, . . . , N − √ ×
l=1 4Tτ 2π n=1 l=1
(10c) ⎡ ⎛ 2 ⎞
κ + γ̄l Nτ fs
N ⎢ ⎜ ⎟
× ⎣Pr (θl = 0)Rn,l exp ⎝− ⎠
an,l = 1, an,l = {0, 1}, l = 1, 2, . . . , L 2
n=1
(10d) ⎤
1 ⎥
0 ≤ τ ≤ T − Nε; (10e) × κ + γ̄l Nτ fs γ̄l2 Nfs + γ̄l Nfs τ 2 ⎦ < 0
pn,l ≥ 0, n = 1, 2, . . . , N, l = 1, 2, . . . , L
(10f) (12)
For simplifying the optimization problem (10), we relax which indicates that ηSE is convex in τ . Moreover, Rn,l
the integer an,l with any value within [0,1] and let ωn,l = is obviously convex in (an,l , ωn,l ). Since ηSE is the non-
Liu and Jia EURASIP Journal on Wireless Communications and Networking (2017) 2017:171 Page 5 of 8
negative linear combination of Rn,l , ηSE is also convex in the optimal τ through iteratively updating τ {k} until it is
({an,l }, {ωn,l }). Hence, ηSE is a convex optimization prob- convergent, as follows:
lem about (τ , {an,l }, {ωn,l }).
∂ηSE
∂τ (k)
τ (k+1) = τ (k) − (17)
We use the alternating direction optimization (ADO) to ∂ 2 ηSE
get the solution to the optimization problem. Firstly, fixing ∂ 2 τ (k)
τ , we optimize (an,l , ωn,l ). Using the Lagrange multiplier, where k ≥ 0 is the iteration index. As mentioned above,
the optimization function is given by we use ADO to optimize (an,l , ωn,l ) and τ alternatively
until an,l , ωn,l , and τ are all convergent, as shown in
Algorithm 1. Then, if an,l = 0, the transmission power
N
L
ωn,l gn,l
2 ω
pn,l = 0; otherwise, pn,l = a n,l . Since ηSE is convex in τ , a
(an,l , ωn,l ) = ρl an,l log 1 + n,l
n=1 l=1
an,l σl2 and ω, ηSE is non-decreasing during each iteration, which
is described as follows:
N
L
− μn ωn,l − p̂max
n
n=1 l=1 Algorithm 1 Joint optimization algorithm based on ADO
N
L 1: Initialize k = 0 and τ (k) with any value within
− νl an,l − 1 (k) = {0} and {ω }(k) = {0};
[0, T − Nε], {a n,l } n,l
l=1 n=1
2: while ηSE τ , {an,l }(k) , {ωn,l }(k) is not convergent
(k)
(13)
do
3: With given τ (k) , obtain the Lagrange multiplier
f function (13);
where ρl = Pr (θl = 0) 1 − Ql is constant and μn for
4: Set t = 0, {μn (t)}, {νl (t)}, ζ1 (t) and ζ2 (t) with the
n = 1, . . . , N and νl for l = 1, . . . , L are the Lagrange
∂ ∂ presettled values;
multipliers. Through calculating ∂a = 0 and ∂ω = 0,
n,l n,l 5: while {νl (t)} and {μn (t)} are not convergent do
we have
6: Update {an,l }∗ and {ωn,l }∗ with {μn (t)} and {νl (t)}
by (14) and (15);
+
1 σ2 7: Update {μn (t + 1)} and {νl (t + 1)} with updated
ωn,l = ρl an,l − 2l (14) {an,l }∗ and {ωn,l }∗ ;
μn gn,l
8: Let t = t + 1;
0, νl ≤ n,l (μn ) 9: end while
an,l = (15)
1, νl > n,l (μn ) 10: Set {an,l }(k+1) = {an,l }∗ and {ωn,l }(k+1) = {ωn,l }∗ ;
11: With {an,l }(k+1) and {ωn,l }(k+1) , calculate τ (k+1) by
2 (17);
g μn σl2
where n,l (μn ) = ρl log μ n,lσ 2 + 2 − 1 . Then, 12: Let k = k + 1;
n l gn,l
13: end while
the Lagrange multipliers μn and νl can be obtained by
14: : Output τ (k) , {an,l }(k) and {ωn,l }(k) .
using the gradient method that leads to the following
update equations
L
ηSE τ (k),{an,l }(k),{ωn,l }(k) ≤ ηSE τ (k+1) , {an,l }(k) , {ωn,l }(k)
μn (t + 1) = μn (t) + ζ1 (t) × ωn,l − p̂max
n
l=1 ≤ ηSE τ (k+1) , {an,l }(k+1),{ωn,l }(k)
N
νl (t + 1) = νl (t) + ζ2 (t) × an,l − 1 ≤ ηSE τ (k+1),{an,l }(k+1),{ωn,l }(k+1)
n=1
(18)
(16)
Hence, the convergence of ηSE can be obtained after
some iterations.
where t ≥ 0 is the iteration index and ζ1 (t) and ζ2 (t) are
Supposing the estimation error is δ, ηSE will be conver-
both the positive step sizes. Then, the updated Lagrange
gent when τ (k) , {an,l }(k) , and {ωn,l }(k) are all convergent.
multipliers in (16) is used for updating the power allo-
Thus, the iterative complexity
of the joint optimization
cation in (14). Secondly, we fix Rn,l with the optimized
(an,l , ωn,l ). Using the Newton iteration method, we can get algorithm is given by O δ13 .
Liu and Jia EURASIP Journal on Wireless Communications and Networking (2017) 2017:171 Page 6 of 8
η (τ ,{a },{p })
2.4 Energy efficiency maximization (EEM) Setting q = ESE n,l n,l
T (τ ,{an,l },{pn,l })
, the optimization problem
Compared to the traditional communication system, the (21) can be rewritten as follows:
SU may consume more energy due to spectrum sens- % & % &
ing power. Thus, we use energy efficiency to evaluate max ηSE τ , {an,l }, {pn,l } − qET τ , an,l , pn,l
τ ,{an,l },{pn,l },q
the energy consumption of cooperative spectrum sensing. (22a)
We define the energy efficiency as spectrum efficiency to
energy consumption ratio. The consumed total energy for s.t.(21b) − (21f ) (22b)
spectrum sensing and data transmission within unit time Supposing the feasible region of the solutions to (21) is
is given by denoted as S, the joint optimization algorithm based on
the Dinkelbach optimization is described in Algorithm 2.
1
N
L
! "
ET = Npc (τ + ε) + (T − τ − Nε) an,l pn,l Algorithm 2 Joint optimization algorithm based on
T Dinkelbach’s optimization
n=1 l=1
(19) 1: Initialize q(k) = 0 where the iteration index k = 1, and
the estimation error δ;
2: With
given q(k) , use the ADO to obtain the solution
(k)
% &(k) % &(k)
Hence, the energy efficiency of cooperative spectrum τ , an,l , pn,l to the following equivalent
sensing is given by
optimization problem:
' % & % & % &
F(q(k) ) = max ηSE τ , an,l , pn,l −q(k) ET τ, an,l ,
ηSE
ηEE = % & % & % & (
ET
#
$ pn,l | τ , an,l , pn,l ∈ S ;
N L pn,l g 2
f
(T −τ −Nε) n=1 Pr (θl = 0) 1−Ql an,l log 1+ σ 2n,l
l=1
= L ! "
l
3: If F(q(k) ) ≤ δ, go to
step (5), otherwise, go to step (4);
Npc (τ +ε)+(T −τ −Nε) N n=1 l=1 a n,l pn,l (k)
ηSE τ (k) ,{an,l } ,{pn,l }
(k)
n=1
(21d)
0 ≤ τ ≤ T − Nε; (21e)
Qmin
d
=0.6
pn,l ≥ 0, n = 1, 2, . . . , N, l = 1, 2, . . . , L 0.5
Qmin=0.7
d
(21f) min
Qd =0.8
min
Q =0.9
d
25 25
Sensing SNR=−14dB SEM
Sensing SNR=−12dB EEM
Sensing SNR=−10dB
20 Sensing SNR=−8dB 20
(bit/Hz/mJ)
(bit/Hz/mJ)
15 15
EE
10
EE
10
η
η
5 5
0 0
0 1 2 3 4 5 6 7 8 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1
min
τ (ms) Qd
Fig. 3 Energy efficiency with sensing time Fig. 5 Energy efficiency comparison
Rayleigh distribution with the mean −10 dB, the absence and sensing SNR increases. Figures 4 and 5 compare spec-
probability of the PU is Pr (θl = 0) = 0.5, the frame trum efficiency and energy efficiency of SEM and EEM,
duration is T = 100 ms, the length of cooperative inter- respectively. We can see that SEM can obtain higher spec-
action slot is ε = 1 ms, the sampling frequency is fs = trum efficiency while EEM may achieve higher energy
100 KHz, the maximal power of each SU is pmaxn = 10 mW, efficiency, due to the different optimization objectives of
the sensing power pc = 1 mW, and the noise power is SEM and EEM.
σl2 = 0.01 mW. Then, we use power utilization ratio ρ to describe the
Figures 2 and 3 show spectrum efficiency ηSE and energy usage of power in CSCN, which represents the power
efficiency ηEE varying with sensing time τ , respectively. consumption required to accomplish the desired commu-
We can see that there exists an optimal τ to maximize nication task when the maximum power is limited. ρ can
ηSE or ηEE . When τ is smaller, the detection performance be formulated as the ratio of the total consumed power to
will degrade to decrease the spectrum access; however, the total transmission power, as follows:
when τ is larger, both transmission time and transmission N L
energy will decrease. Thus, there is a tradeoff between n=1 l=1 an,l pn,l + Npc
ρ= N (23)
spectrum sensing and data transmission. We also see that max
n=1 pn
ηSE and ηEE improve as detection probability decreases
10 0
1.5
SEM
EEM
1
ηSE (bps/Hz)
10 -1
ρ
0.5
EEM
SEM
0
0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1 10 -2
0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1
Qmin Qmin
d d
Fig. 4 Spectrum efficiency comparison Fig. 6 Power utilization ratio comparison with detection probability
Liu and Jia EURASIP Journal on Wireless Communications and Networking (2017) 2017:171 Page 8 of 8
Authors’ contributions
XL conceived and designed the study. MJ performed the simulation
0.9
experiments. XL wrote the paper. MJ reviewed and edited the manuscript.
EEM
0.8 SEM Both authors read and approved the final manuscript.
0.7
Competing interests
0.6 The authors declare that they have no competing interests.
Author details
1 School of Information and Communication Engineering, Dalian University of
Acknowledgements
This work was supported by the National Natural Science Foundations of
China under Grant Nos. 61601221, 91438205, and 61671183; the Natural
Science Foundations of Jiangsu Province under Grant No. BK20140828; the
China Postdoctoral Science Foundations under Grant No. 2015M580425; the
Fundamental Research Funds for the Central Universities under Grant No.
DUT16RC(3)045; and the Open Research Fund of State Key Laboratory of
Space-Ground Integrated Information Technology under Grant No.
2015_SGIIT_KFJJ_TX_02.