Received January 25, 2021, accepted February 5, 2021, date of publication February 22, 2021, date of current version

March 2, 2021.
Digital Object Identifier 10.1109/ACCESS.2021.3060784

NOMA Based Cooperative Relaying Strategy for

Underwater Acoustic Sensor Networks Under
Imperfect SIC and Imperfect CSI:
A Comprehensive Analysis
VEERAPU GOUTHAM , (Member, IEEE), AND V. P. HARIGOVINDAN , (Senior Member, IEEE)
Department of Electronics and Communication Engineering, National Institute of Technology Puducherry, Karaikal 609609, India
Corresponding author: Veerapu Goutham (gouthamveerapu@gmail.com)
This work was supported by the Science Engineering Research Board, Department of Science and Technology, Government of India,
through the Mathematical Research Impact Centric Support scheme under Grant MTR/2019/001228.

ABSTRACT In this paper, we propose non-orthogonal multiple access based cooperative relaying strat-
egy (NOMA-CRS) for underwater acoustic sensor networks (UASNs). We analyse the performance of
NOMA-CRS for both shallow and deep water scenarios, under imperfect channel state information (I-CSI)
as well as imperfect successive interference cancellation (I-SIC). We derive mathematical expressions for
ergodic rate, outage probability as well as the energy efficiency of NOMA-CRS in UASNs by considering
the underwater specific characteristics, such as distance-dependent usable bandwidth, acoustic spreading,
propagation loss, and fading effects. We compare the performance of NOMA-CRS with the widely used
decode-and-forward based CRS in the UASNs. From the results, it is evident that NOMA-CRS can achieve
significant improvement in ergodic sum rate and energy efficiency. But the outage performance is slightly
degraded for the proposed scheme. Our results show that I-CSI and I-SIC have a significant impact on
the performance of the NOMA-CRS. We also investigate the impact of relay position, wind speed as well
as shipping activities on the performance of NOMA-CRS under the realistic underwater scenario. Results
show that high-speed winds and high shipping activities severely degrade the performance of ergodic sum
rate of the NOMA-CRS. Implementation of NOMA-CRS requires CSI at the transmitter. However, acquiring
perfect CSI at the transmitter is a challenging task in time-varying multi-path underwater acoustic channels.
As a solution, we also propose space-time block coded NOMA-CRS (STBC-NOMA-CRS) for UASNs,
which can be implemented without CSI at the transmitter. Extensive simulation studies are conducted to
corroborate the analytical findings.

INDEX TERMS Cooperative relaying strategy, ergodic rate, energy efficiency, outage probability,
power-domain non-orthogonal multiple access, space-time block codes, underwater acoustic sensor
networks.

I. INTRODUCTION of underwater things (IoUT) [1]. IoUT consists of multi-

Recent advances in underwater wireless sensor net- ple underwater objects, which are interconnected through a
works (UWSNs) lead to a variety of oceanic applications, wireless communication medium. In particular, IoUT allows
which include ocean exploration, underwater multimedia, applications to monitor the vast unexplored marine areas
military surveillance system, assisted navigation, pollution [1]–[3]. UWSNs may consist of both mobile and stationary
control, and much more. Nowadays, UWSNs also support the nodes that exchange information, such as control, telemetry,
Internet of Things, which has been described as the Internet speech, and video signals among themselves as well as to
a central node located on or offshore. Such diverse and
The associate editor coordinating the review of this manuscript and data-intensive underwater applications demand higher band-
approving it for publication was Giovanni Pau . width as well as data rate [4]. Innovative physical (PHY) layer

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
VOLUME 9, 2021 32857
 V. Goutham, V. P. Harigovindan: NOMA-CRS for UASNs Under I-SIC and I-CSI

schemes are therefore needed to develop spectral efficient, assumption in UASNs due to the underwater channel charac-
highly reliable, and energy-efficient UWSN to meet these teristics. This assumption does not provide clear insight into
challenges. Underwater wireless communication is achieved the performance of the practical UASN due to the existence
either by radio frequency (RF) or by optical or by acoustic of the channel estimation errors. It has been observed that the
signals. Although RF signals have a significant bandwidth, performance analysis of the NOMA scheme under I-CSI and
they are extremely prone to absorption losses in the under- I-SIC for UASNs has not yet reported in the literature.
water medium. Optical signals face extreme challenges due On the other hand, CRS is a powerful approach in
to absorption losses, dispersion caused by suspended particles UASNs, that enhances the reliability and energy efficiency.
and ambient light interference. This leads to a low transmis- In CRS, nearby relay nodes cooperate with the source node
sion range of a few meters with underwater optical commu- to improve the reliability by exploiting spatial diversity.
nication. The wireless technology with acoustic signals is Some of the existing research works have proposed the CRS
the preferred and most popular mode of communication in for UASNs to enhance throughput, reliability and energy
UWSNs due to its long-range data transmission. However, efficiency [4], [18]–[22]. Various types of CRS strategies
the use of acoustic signals in the PHY layer offers limited are proposed for UASNs, namely amplify-and-forward and
communication bandwidth, limited data rates and increased decode-and-forward [4], [23]. In amplify-and-forward CRS,
propagation delays compared to terrestrial wireless sensor the relay node amplifies the data received from the source
networks (WSNs). Additionally, the underwater communi- node, and retransmits the information to the destination
cation channel suffers from node mobility, multi-path fading node [24]. Whereas in the decode-and-forward CRS, the relay
and Doppler spread. All of these factors lead to variations node decodes the source node information, and retransmits
in the temporal and spatial characteristics of the underwater decoded information to the destination node. The authors
acoustic sensor network (UASN) channel. This makes the of [25] investigated the outage performance for cooper-
channel bandwidth dependent on both transmission distance ative orthogonal frequency division multiple access with
and signal frequency. Therefore, improving channel capacity both amplify-and-forward and decode-and-forward relaying
within the limited communication bandwidth is one of the strategies for underwater communication systems. Authors
major concerns in UASNs. of [26] proposed energy-efficient cooperative opportunistic
To achieve higher data rates, massive multiple input mul- routing protocol for UASNs. Here, the source node uses the
tiple output (MIMO), and non-orthogonal multiple access depth information of the underwater devices and residual
(NOMA), are introduced in terrestrial communication sys- energy of the relay nodes as constraints to determine the
tems [5]–[9]. However, practical implementation of massive route to the destination node. CRS mechanisms are also
MIMO over frequency-selective channels such as UASNs designed for UASNs to address the medium access control
is a challenging task considering the higher complexity of layer and network layers issues, which mainly focus on
the equalisers. Recently, NOMA is seen as a promising enhancing energy efficiency of UASNs [27]–[29]. Most of
technique in UASNs to achieve higher data rates within these works have focused primarily on improving energy
the limited communication bandwidth. Authors in [10]–[14] efficiency and reliability in bandwidth-limited UASNs with
proposed NOMA for UASNs to improve channel capacity. the cost of reduced ergodic rates. Differently from the existing
NOMA can be implemented by exploiting either power or works, in this research work, we aim to improve both ergodic
code domains. Power-domain NOMA can transmit multiple sum rate and energy efficiency of UASNs within the limited
user information within a single resource block by sharing bandwidth of the PHY layer by combining CRS with the
the transmission power among the users. A resource block NOMA. In addition, obtaining perfect CSI at the transmitter
can be a time slot or a frequency band or a code. The key is a complex task in UASNs due to the long propagation
techniques involved in power-domain NOMA are superposi- delays. We therefore propose NOMA-CRS scheme, which
tion coding and successive interference cancellation (SIC). can be implemented without CSI at the transmitter, by inte-
Further, authors in [15]–[17] combined the NOMA with a grating the proposed scheme with the STBC codes. STBC
cooperative relaying strategy (NOMA-CRS) to improve the codes are popular for the transmitter cooperation without CSI
channel capacity and reliability of the terrestrial wireless at the transmitter.
systems. The authors of [17] analysed the impact of imperfect To the best of authors’ knowledge, this is the first research
channel state information on the performance of NOMA with work to analyse NOMA based CRS for UASNs under imper-
cooperative relaying for terrestrial wireless communications fect CSI as well as imperfect SIC and to propose space-time
by considering perfect SIC. Although advanced SIC tech- block coded NOMA-CRS (STBC-NOMA-CRS) for UASNs.
niques can mitigate the impact of successive interference up The major contributions of this research work are as
to some extent, still near node will be affected by the residual follows:
successive interference from the higher power transmission 1) We propose non-orthogonal multiple access based
symbols, which increases proportionally with the distance cooperative relaying strategy (NOMA-CRS) for under-
between near node and the source node. Besides, all recently water acoustic sensor networks (UASNs).
published works on NOMA for UASNs have analysed the 2) We present a comprehensive study of NOMA-CRS
ergodic rate by assuming perfect CSI, which is too idealistic for UASNs under imperfect channel state information

32858 VOLUME 9, 2021

V. Goutham, V. P. Harigovindan: NOMA-CRS for UASNs Under I-SIC and I-CSI

(I-CSI) and imperfect successive interference cancella- by,

tion (I-SIC) in both shallow and deep water scenarios.
3) We derive mathematical expressions of ergodic SL = 10 log10 Is − 10 log10 Iref + 10 log10 ηea , (2)
rate, outage probability and energy efficiency for where, Is is the signal intensity transmitted from the source
NOMA-CRS for UASNs by considering the underwa- measured at 1 m distance, Iref is the reference intensity, and
ter specific characteristics such as distance-dependent ηea is the electrical to acoustic power conversion efficiency.
usable bandwidth and transmission power, acoustic Here, we consider both shallow and deep water scenarios. The
spreading, propagation loss, and fading effects. characteristics of shallow and deep water scenarios are given
4) The performance of NOMA-CRS is compared with in Table. 1.
widely used decode-and-forward based CRS in the
UASNs. The results show that the NOMA-CRS can TABLE 1. Shallow and deep water characteristics.
achieve improved ergodic sum rate, and energy effi-
ciency at the cost slightly reduced outage performance.
The impact of I-CSI and I-SIC on the performance of
NOMA-CRS is also thoroughly investigated.
5) We also examine the impacts of the relay position, and
environmental effects such as, wind speed and ship-
ping activities on the NOMA-CRS considering realis-
tic underwater scenario. Results show that high-speed
winds and high shipping activities severely degrade the
performance of ergodic sum rate of the NOMA-CRS.
6) Acquiring perfect CSI at the transmitter is one of the The transmitted signal intensity (Is ) is the power flow
major challenge in implementing NOMA in UASNs. per unit area. The propagation of acoustic waves encounters
As a solution to this, we propose space-time block the spherical spreading in deep water [30]. So, the signal
coded NOMA-CRS (STBC-NOMA-CRS) for UASNs, intensity transmitted from the source in deep water is given
which brings forth the opportunity to exploit the bene- by, Is = P4πtx , where Ptx is the transmitted signal power
fits of both spectral efficiency and transmit diversity. from the source. Whereas, the propagation of acoustic waves
The rest of the paper is organized as follows: Section II encounters cylindrical spreading in the case of shallow water
presents UASN model along with the proposed NOMA- due to the signals being bounded by the floor and surface of
CRS, and CRS schemes for UASNs. In this section, we also the ocean. Hence, the signal intensity transmitted from the
Ptx
derive mathematical expressions for ergodic rate, outage source in shallow water is given by, Is = 2πH , where H is the
probability and energy efficiency of NOMA-CRS scheme in water depth in meters. The term Iref is the reference intensity,
q2
UASNs. Section III presents STBC-NOMA-CRS for UASNs. which is given by ρc , where q = 1µ Pa rms is the root mean
Section IV provides a description of analytical and simulation square acoustic pressure, ρ the density of sea water, and c is
results. The conclusion is presented in Section V. the sound velocity in sea. c is mainly affected by temperature,
sea depth and salinity of sea water, and is given by [30], [32],
II. SYSTEM MODEL
This section presents (1) the physical layer modeling of c = 1448.96 + 4.591T − 0.05304T 2 + 0.0002374T 3
UASNs, (2) the proposed NOMA-CRS scheme for UASNs, + 1.340(χ − 35) + 0.0163H + 1.675 × 10−7 H 2
and (3) the mathematical modeling and derivation of accurate −0.01025T (χ − 35) − 7.139 × 10−13 TH 3 , (3)
expressions for the ergodic rate, the outage probability and
the energy efficiency of NOMA-CRS for UASNs. where T is the temperature, and χ is the salinity of sea water.
TL consists of absorption and spreading losses. TL between
A. PHYSICAL LAYER MODEL transceiving nodes in an underwater acoustic channel is given
In this subsection, we present the physical layer model for the by [33],
calculation of the SNR in UASNs [30], [31]. The signal-to- TL = k × 10 log r + r × 10−3 × 10 log a(H , T , f ) + A,(4)
noise ratio (SNR) at the receiver of the UASNs is provided
by the passive sonar equation. Using this equation, the SNR where k is the spreading factor, r is the distance between
at the receiver can be calculated from the transmitted signal transceiving nodes in km, A is transmission loss anomaly and
level (i.e., source level (SL)), by considering transmission a(H , T , f ) is the absorption coefficient in dB/km for f in kHz.
loss (TL), ambient noise level (NL), and directivity (DI) can The Francois & Garrison model proposed in [34],and [35] is
be calculated. The passive sonar equation is given by, an accurate and most commonly used model for calculating
the absorption coefficient in UASNs, which not only takes
0in dB = SL − TL − (NL − DI ), (1)
into account the effect of acoustic frequency but also intro-
where 0 is the expected SNR at the receiver in dB. SL mea- duces the effects of temperature (T ), depth (H ), pressure,
sured at 1 m with respect to the reference intensity is given acidity, and salinity (χ). In this research work, we use this

VOLUME 9, 2021 32859

 V. Goutham, V. P. Harigovindan: NOMA-CRS for UASNs Under I-SIC and I-CSI

accurate absorption coefficient model. The absorption coef- are primarily dependent on the signal frequency (f ) and
ficient a(H , T , f ) in dB/km (valid for 100 Hz < f < 1 MHz) the transmission distance (r). The achievable ergodic rate
is given by [34], [35], for frequency-dependent
hR noise is therefore
i defined as [36],
A1 P1 f1 f 2 A2 P2 f2 f 2 C = E B(r) log2 (1 + 0(r, f )) df , where B(r) represents
a(H , T , f ) = + + A3 P3 f 2 , (5) distance-dependent usable bandwidth, which is given by,
f 2 + f12 f 2 + f22
B(r) = Br −Q , and E[ ] is the expectation of the random
where f is the operating signal frequency, f1 , and f2 are the variable. Here B is bandwidth coefficient and Q is positive
relaxation frequencies in kHz for boric acid and magnesium bandwidth exponent.
sulphate. Ai i ∈ {1, 2, 3}, and Pi i ∈ {1, 2, 3} are the coef-
ficients, which considers the effects of temperature, salinity B. NOMA-CRS SCHEME
and depth on acoustic signal absorption. The expressions for We consider a simple three-terminal communication scenario
finding these coefficients are given by, as shown in Fig. 1, consisting of three half-duplex nodes. S is
 χ 0.5 h4− 1245 i the source node, G is the near node to S with good channel
f1 = 2.8 10 273+T conditions, and P is the node far from S with poor channel
35
h
1990
i conditions. The distance between the S-to-P and S-to-G links
8−
8.17 × 10 273+T are denoted as DSP , and DSG , respectively. The angle between
f2 = the S-to-P and S-to-G links is denoted as θ. Using cosine
1 + 18 × 10−4 (χ − 35)
rule,
q the distance between the G-to-P link is given by, DGP =
8.68
A1 = × 10(0.78 pH −5) D2SP + D2SG − 2 DSP DSG cos θ. The channel coefficient of
c
χ S-to-P link is expressed as hSP = |hSP |ejθSP , with an average
A2 = 21.44(1 + 0.025 T ) power of βSP , where |hSP | and θSP are the magnitude and
c
 phase of the S-to-P link, respectively. Similarly, the channel

 4.937×10 −2.59×10−5 T + 9.11 × 10−7 T 2
−4
coefficients of S-to-G, and G-to-P links are expressed as

hSG = |hSG |ejθSG , and hGP = |hGP |ejθGP , with average

+9.11×10−7 T 2 −1.50×10−8 T 3 for T ≤ 20◦ C

A3 = powers of βSG , and βGP , respectively. Channel coefficients
 3.964 × 10−4 −1.146×10−5 T +1.45 × 10−7 T 2


 hSP , hSG , and hGP are assumed to be independent Rayleigh
−6.5 × 10−10 T 3 for T > 20◦ C

fading coefficients due to the multi-path signal propagation in
UASNs. Accordingly, the channel powers |hSP |2 , |hSG |2 , and
P1 = 1
|hGP |2 are exponentially distributed random variables. It is
P2 = 1 − 1.37 × 10−4 H + 6.2 × 10−9 H 2 assumed that the average power of the channel coefficient
hSP is less than the average power of the channel coefficient
P3 = 1 − 3.83 × 10−5 H + 4.9 × 10−10 H 2 . (6)
hSG (|hSP |2 < |hSG |2 ), due to the high path loss of the S-to-P
Ambient noises in UASNs includes turbulence noise link compared to the S-to-G channel link. In UASNs, due to
(Nt (f )), shipping noise (Ns (f )), waves noise (Nw (f )) and channel estimation errors, it is extremely difficult to obtain
thermal noise (Nth (f )). The empirical formulas for the power the perfect CSI [37], [38]. Hence, we consider the estimation
spectral densities (expressed in dB re 1µ Pa per Hz) of these of hj is h̃j , j ∈ {SP, SG, GP}, and the channel coefficient can
noise components are given by, be modelled as h̃j = hj + κ ĥj , where ĥj represents channel
error vector which can be modelled as zero mean complex
10 log10 Nt (f ) = 17 − 30 log10 f , forf < 10 Hz normal distribution with variance of νj2 , and κ is the channel

40 + 20(s − 0.5) + 26 log10 f
 estimation error factor.
10 log10 Ns (f ) = −60 log10 (f + 0.03) , In NOMA-CRS, both the symbols X1 and X2 are intended


for 10 < f < 100 Hz to be transmitted to the node P as shown in Fig. 1. In NOMA-
 CRS scheme, the node S transmits two symbols X1 and X2
0.5
50 + 7.5w + 20 log10 f
 with distinct power levels α1 Pt and α2 Pt respectively, where
10 log10 Nw (f ) = −40 log10 (f + 0.4) , Pt is the total transmission power, α1 and α2 are the power
for 0.1 kHz < f < 100 kHz allocation coefficients (α1 + α2 = 1, α1 > α2 ). Here,


10 log10 Nth (f ) = −15+20 log10 f , for f > 100 kHz, (7) the node G operates as a half-duplex decode-and-forward
relay, which decodes the symbols using SIC technique. Then
where, w is the wind speed and s is the shipping activity. node G transmits the X2 symbol to node P with the power
The ambient noise level generally decreases with increase Pt in the next transmission time slot as shown in Fig. 1.
in frequency. The sum of these four noise components gives As a result, the node P can obtain the symbol X2 through
the total power spectral density (p.s.d) of ambient noise the S-to-G-to-P and S-to-P links. Subsequently, the node P
(N (f )) present in the ocean. The approximate p.s.d of ambient decodes the symbols received from the S and G nodes jointly
noise present in the ocean is given by, 10 log10 (N (f )) = by involving selection combining diversity and successive
50 − 18 log10 (f ) [36]. In UASNs, noise and path losses interference cancellation techniques. Accordingly, the signal

32860 VOLUME 9, 2021

V. Goutham, V. P. Harigovindan: NOMA-CRS for UASNs Under I-SIC and I-CSI

FIGURE 1. NOMA-CRS scheme for UASNs.

received at the node G in a given time slot i is expressed as, signals. Higher values of ξ indicates receiver with higher SIC
! inefficiency. The received signal at node P from S in a given
hSG + κ ĥSG hp p i
YSG (i) = √ a1 Pt X1 (i)+ a2 Pt X2 (i) +nSG (i), time slot i is given by,
TLSG (r, f ) !
hSP + κ ĥSP hp p i
(8) YSP (i) = √ a1 Pt X1 (i)+ a2 Pt X2 (i)
TLSP (r, f )
where nSG (i) denotes frequency dependent noise in the under-
+ nSP (i), (10)
water channel with variance of σSG2 (f ). Firstly, the node G

decodes symbol X1 . Then, it subtracts decoded symbol X1 where nSP (i) is the frequency dependent noise in the underwa-
from YSG (i) using SIC technique to obtain the symbol X2 . ter channel with the variance of σSP
2 (f ). Therefore, the signal-
X1
Therefore, the SINRs of both the symbols X1 and X2 at the to-interference noise ratio (SINR) for X1 symbol γSP (r, f ) at
node G are respectively expressed as, the node P is obtained by,

X1
a1 |hSG |2 TLSGP(r,f
t
) a1 |hSP |2 TLSPP(r,f
t
γSG = X1
γSP =
)
a2 |hSG |2 TLSGP(r,f ) + κ νSG TLSG (r,f ) + σSG (f )
2 2 Pt 2
a2 |hSP |2 TLSPP(r,f ) + κ νSP TLSP (r,f ) + σSP (f )
t t 2 2 Pt 2

a1 |hSG |2 0SG a1 |hSP |2 0SP

= , (9a) = , (11)
a2 |hSG |2 0SG + κ 2 νSG2 0
SG + 1 a2 |hSP |2 0SP + κ 2 νSP2 0 +1
SP
a2 |hSG |2 TLSGP(r,f
t
X2
γSG =
) where 0SP = Pt
2 (f ) . In the second transmission time
TLSP (r,f )σSP
ξ a1 |hSG TLSGP(r,f
|2 t
) + κ νSG TLSG (r,f ) + σSG (f )
2 2 Pt 2
slot, the node G transmits symbol X2 to the node P. Hence,
a2 |hSG |2 0SG the received signal at node P from G in a given time slot i is
= , (9b) given by,
ξ a1 |hSG |2 0SG + κ 2 νSG 2 0
SG + 1 !
hGP + κ ĥGP hp i
where 0SG = Pt
2 (f ) and the term ξ a1 |hSG | 0SG
2
YGP (i) = √ Pt X2 (i) + nP (i). (12)
TLSG (r,f )σSG TLGP (r, f )
indicates the residual successive interference after SIC at the
node G, and ξ denotes the SIC inefficiency. ξ varies in the Here, we consider selection diversity at the node P. The
range of [0, 1], ξ = 0 represents perfect SIC and other node P combines the X2 symbol from S and G using selec-
values of ξ represent I-SIC i.e., interference due to residual tion diversity technique. Therefore, the SINR for X2 symbol

VOLUME 9, 2021 32861

 V. Goutham, V. P. Harigovindan: NOMA-CRS for UASNs Under I-SIC and I-CSI

1
γPX2 (r, f ) at the node P is obtained by, where λSP = βSP 0SP
ψSP and λSG = βSG 0SG
ψSG . Let X =
( a2 W . Based on the linear random variable transformation,
a2 |hSP |2 0SP

X2 1 1
γP = max , the CDFs of X is obtained as, FX (x) = 1−e
−x
a2 λSP a2 λSG+
.
ξ a1 |hSP | 0SP + κ 2 νSP
2 2 0 +1
SP
) Using the random variables, W and X , (14) can be re-written
|hGP |2 0GP as,
, (13)
κ 2 νGP
2 0
GP + 1 1 fu ∞
Z Z
CX1 = log2 (1 + W ) fW (w) dw df
where 0GP = Pt
2 (f ) . 2 fl 0
TLGP (r,f )σGP
1 fu ∞
Z Z
1) ERGODIC RATE − log2 (1 + X ) fX (x) dx df . (18)
2 fl 0
Let E [Z ] , FZ (z) , and fZ (z) denote the expectation, cumu- R∞ R∞
lative distribution function (CDF) and probability density By applying 0 log2 (1 + W ) fW (w) dw = ln2 1
0
function of a random variable Z , respectively. The ergodic 1−FW (w) ∞ e−µx
= −e(µβ) Ei(−µβ) [39],
R
1+w dw, and 0 x+β dx
rate of the X1 symbol in NOMA-CRS scheme for UASNs is the ergodic rate for symbol X1 is obtained as,
given by, Z fu  1   
1 + 1 1 1
1 fu h −e λSP λSG Ei −
Z
CX1 = − df
 n oi
X1 X1
CX1 = E log2 1 + min γSP , γSG df . (14) 2 ln 2 fl λSP λSG
2 fl Z fu 1  1 1   
1 a2 λSP +λSG
1 1
where, fl and fu are the lower and upper cutoff fre- + e Ei − − df .
quencies, and γSPX1
and γSG X1
are the respective SINR 2 ln 2 fl a2 λSP a2 λSG
of symbol X1 at S-to-P and S-to-G links. From (14), (19)
we assume V1nis an arbitrary
o variable, considered as V1 =

X1 X1 Similarly, the ergodic rate for X2 symbol is given by,
log2 1 + min γSP , γSG . By substituting (11) and (9a),
1 fu h
Z  n oi
we obtain variable V1 as, CX2 = E log2 1 + min γPX2 , γSG X2
df . (20)
 
a1 |hSP |2 0SP 2 fl
V1 = log2 1 + min ,
a2 |hSP |2 0SP + ψSP From (20),
 V2 is assumed
n as an
oarbitrary variable given by
a1 |hSG |2 0SG X2 X2
V2 = log2 1 + min γP , γSG . By substituting (13) and

, (15)
a2 |hSG |2 0SG + ψSG (9b), we obtain variable V2 as,
where ψSP = κ 2 νSP
2 0
SP + 1 and ψSG = κ νSG 0SG + 1.
2 2   
a2 |hSP |2 0SP
Accordingly, V2 = log2 1 + min max ,
   ξ a1 |hSP |2 0SP + ψSP
|hGP |2 0GP a2 |hSG |2 0SG
 
 a1 a 1

V1 = log2 1+min , , , , (21)
a2 + ψSP a + ψSG
2 |h |2 0
 ψGP ξ a2 |hSG |2 0SG + ψSG
|hSP |2 0SP SG
SG
where ψGP κ 2 νGP
2 0

= GP + 1. We assume Y and
1
 a1 
Z are
n random2 variables, 2which o are denoted as, 2 Y =
= log2 1 + ,
 
1
a2 + 1 max ξ a a|h2 |h|SP2 0| 0+ψ
SP
, |hGPψ|GP0GP , and Z = ξ a |ha2 |hSG | 0SG
.
2 SG | 0SG +ψSG
2
  
|hSP |2 0SP |hSG |2 0SG 1 SP SP SP
min ψSP , ψ The corresponding CDFs of Y and Z are given by,
SG

|hSP |2 0SP |hSG |2 0SG

    
yψ
= log2 1 + min , 1−e
− β 0 (aSP−yξ a )
ψSP ψSG SP SP 2 1





a2 |hSP |2 0SP a2 |hSG |2 0SG
   yψ

FY (y) = − β GP (22)
− log2 1 + min , . 1−e GP 0 GP ; y < ξaa21
ψSP ψSG



y ≥ ξaa21 ,

(16) 1;

 zψSG

Let, U and V are random variables, which are denoted by, a
1 1
 1 − e− βSG 0SG (a2 −zξ a1 ) ; z < 2

U = |hSPψ|SP0SP , and V = |hSGψ|SG0SG , respectively. Accord- ξ a1 (23)
2 2
FZ (z) = a2
ingly, the CDF of random variable U is given by, FU (u) = 1;
 z≥ .
−β
u ψSP ξ a1
1−e SP 0SP and CDF of random variable V is given by,
v ψSG 1
−
FV (v) = 1 − e βSG 0SG . We consider W is a random variable Let, A = min {Y , Z }, and the CDF is obtained as,
1
denoted by, W = min {U , V }. Based on the random variable  a2
1 − (1 − FY (a)) (1 − FZ (a)) ; a <
 ,
transformations, the CDFs of W is obtained as,
FA (a) = ξ a1 (24)
a2
.
 
−w λ 1 + λ 1 1; a≥
,

FW (w) = 1 − e SP SG (17) ξ a1
32862 VOLUME 9, 2021
V. Goutham, V. P. Harigovindan: NOMA-CRS for UASNs Under I-SIC and I-CSI

Using the random variable A, (20) can be re-written as, by,

1
Z fu Z ∞ L
log2 (1 + A) fA (a) da df 1X n  n o o
CX2 = PXout
1
= X1
Pr δfj log2 1 + min γSP X1
, γSG < R1
2 fl 0 L
Z fu Z ∞ j=1
1 1 − FA (a)
= da df L
2 ln 2 1+a 1 Xh n n
X1 X1
o oi
fl 0
a2
= 1 − Pr min γSP , γSG ≥ 81
1
Z fu Z ξ a1 (1−FY (a)) (1−FZ (a)) L
j=1
= da df . (25)
2 ln 2 1+a L 
ψSP 81
 
fl 0 1X
= 1 − Pr |hSP |2 <
L a1 0SP − a2 0SP 81
The ergodic sum rate in NOMA-CRS for UASNs is given j=1
ψSG 81
 
by, CX1 + CX2 .
Pr |hSG |2 <
a1 0SG − a2 0SG 81
2) OUTAGE PROBABILITY L   
ψ 81

1X − β 0 SP (a −a 8 )
In this subsection, the performance of NOMA-CRS for = 1 − 1 − e SP SP 1 2 1
UASNs is analysed in terms of outage probability. It is L
j=1
defined as the instance where the received power level is less ψSG 81
   
−
than the threshold power level, i.e. the receiver node is out of 1 − e βSG 0SG (a1 −a2 81 ) , (27)
the source node range. Let R1 and R2 are the desired rates of
X1 and X2 respectively. The mathematical expression for the 2R1
outage probability of X1 is given by, where 81 = 2 δfj −1. Similarly, the mathematical expression
  for average outage probability of X2 symbol over L flat fading
 1 Zfu
  n o 
 channels is given by, (28) as shown at the bottom of the page,
2R2
PXout
1
= Pr log2 X1 X1
1 + min γSP , γSG df < R1 . where 82 = 2 δfj − 1.
2
 

fl
(26) C. CRS SCHEME
Decode-and-forward based CRS scheme is widely used
However, obtaining the mathematical expression for out- scheme in UASNs for improving the energy efficiency and
age probability for frequency selective channels is tedious reliability [4], [21]. Here, the relay node is half-duplex, which
task. So, we formulate the problem as follows, we divide the completely decodes the source node information and retrans-
total bandwidth into L narrow sub-bands. Let jth sub-band mits it to the destination node in the next time slot. The
be centered around frequency fj with width δfj , where destination node combines the two symbols received from the
j = 1, 2, · · · , L. We assume each sub band is small enough relay and source nodes using selection combining diversity
such that the channel appears frequency-flat fading channel. technique. Accordingly, the average achieved ergodic rate by
As a result, the mathematical expression for average outage the decode-and-forward CRS is given by, (29) as shown at the
probability of X1 symbol over L flat fading channels is given bottom of the page,

 
 1 Zfu
  n o 

PXout
2
= Pr log2 X2 X2
1 + min γP , γSG df < R2
2
 

fl
L 
a2 |hSP |2 0SP |hGP |2 0GP a2 |hSG |2 0SG
     
1 X
= 1 − Pr max , ≥ 82 Pr ≥ 82
L ξ a1 |hSP |2 0SP + ψSP ψGP ξ a1 |hSG |2 0SG + ψSG
j=1
L    
8 ψSP
  
ψ 8
   
82 ψSG
 
1X − β 0 (a2 −ξ − β GP0 2 − β 0 (a
= 1 − 1 − 1 − e SP SP 2 1 2 a 8 )
1 − e GP GP 1 − e SG SG 2 1 2−ξ a 8 )
, (28)
L
j=1
" ( ( ))!#
1 fu 0SG |hSG |2 0SP |hSP |2 0GP |hGP |2
Z
CCRS = E log2 1 + min , max , df .
2 fl κ 2 νSG
2 0
SG + 1 κ 2 νSP
2 0 + 1 κ 2ν2 0
SP GP GP + 1
  xψ
 xψ
  xψ

− SG − SP − GP
Z fu Z ∞ 1 − 1 − e 0SG 1 − e 0SP 1 − e 0GP
1
= db df (29)
2 ln 2 fl 0 1+b

VOLUME 9, 2021 32863

 V. Goutham, V. P. Harigovindan: NOMA-CRS for UASNs Under I-SIC and I-CSI

Similarly, the outage probability for the decode-and-

forward CRS scheme is given by,
L    xψ
 xψ

1X − 0 SG − 0 SP
PCRS
out = 1 − 1 − 1 − e SG 1 − e SP
L
j=1
 xψ

− GP
1 − e 0GP . (30)

D. ENERGY EFFICIENCY
In this subsection, the performance of NOMA-CRS scheme FIGURE 2. An illustration of STBC-NOMA-CRS scheme for UASNs.
is analysed from the perspective of energy efficiency. The
expression for energy efficiency is given by [40], The protocol description of the STBC-NOMA-CRS
scheme for UASNs is as follows: the node S divides its
(1 − PXout
1
) CX1 + (1 − PXout
2
) CX2
η= . (31) transmission symbol stream X (t) into four blocks, X1 (t),
Total power consumption X2 (t), X3 (t) and X4 (t). Here, X1 (t) & X3 (t) are intended to
The total power consumed in the NOMA-CRS scheme is be transmitted to the node P and X2 (t) & X4 (t) are intended
given by, E =2 Pt + 2Prx , where Prx are the power con- to be transmitted to the node G simultaneously. As a result,
sumed for receiving the symbols. Whereas, the total power the S combines X1 (t) & X2 (t) to form √ a composite√ signal
consumed in the CRS scheme is given by, E =2 Pt +2 Prx . C1 (t), which is given by C1 (t) = a1 Pt X1 + a2 Pt X2 ,
where a1 & a2 are power allocation coefficients. It is assumed
III. STBC-NOMA-CRS SCHEME that the total transmit power, from the two antennas in the
Even though NOMA based UASNs are proposed in the lit- STBC-NOMA-CRS scheme, is same as the transmit power
erature to improve the performance of UASNs, researchers from the NOMA-CRS scheme. As a result, we assume a1 +
identified that the major challenge in the implementation of a2 = 0.5. Similarly, the S combines X3 (t) & X4 (t) to form
NOMA in UASNs is that, the transmitter needs the prior CSI
√ other composite
an √ signal C2 (t), which is given by C2 (t) =
status encountered by each receiver. However, acquiring per- a1 Pt X3 + a2 Pt X4 . At the first transmission time slot (T0 ),
fect CSI at the transmitter is a challenging job in time-varying the signal transmitted from the TX1 is C1 (t) and from TX2
multi-path underwater acoustic channels. In some cases, is C2 (t). During the next transmission time slot (T1 ), signal
the transmitter obtains CSI using a feedback signal from the −C2∗ (t) is transmitted from the TX1 and C1∗ (t) is transmitted
receiver in UASNs, where the receiver estimates CSI using from the TX2 , where ∗ represents conjugate of the complex
receiving a pilot signal from the transmitter. Many recent signal. Assuming that fading is constant across two consecu-
research works proposed NOMA for UASNs, by assuming tive symbols. The received signals at the G (i.e., RX2 ) and P
perfect or statistical CSI at the transmitter. But this assump- (i.e., RX1 ) nodes in a given time slot i can then be expressed
tion is too idealistic in UASNs. Differently, we propose as,
space-time block coded NOMA-CRS (STBC-NOMA-CRS)
for UASNs. To the best of authors’ knowledge, this is the first Yi (t)
!
work to propose, NOMA with STBC codes for UASNs. Since hi1 + κ ĥi1 hp p i
NOMA facilitates multiple transmissions simultaneously in = √ a1 Pt X1 (t)+ a2 Pt X2 (t)
TLi1 (r, f )
the resource block, its inclusion with the space-time block !
codes (STBC) brings forth the opportunity to exploit the hi2 + κ ĥi2 hp p i
+ √ a1 Pt X3 (t)+ a2 Pt X4 (t) +ni (t),
benefits of both spectral efficiency and transmit diversity. TLi2 (r, f )
To implement STBC with NOMA-CRS, we consider that, (32)
source node S is equipped with two transmitting antennas
(i.e., TX1 & TX2 ), node P equipped with one receiving antenna where i ∈ 1, 2. Here TLij (r, f ) denotes transmission loss
(i.e., RX1 ), and node G equipped with two transmitting anten- between jth transmitter and ith receiver and ni (t) denotes
nas and one receiving antenna (i.e., TX3 , TX4 & RX2 ) frequency dependent noise in the underwater channel. The
as shown in Fig. 2. Let h11 , h12 , h21 , h22 , h13 , and h14 received signals at the G and P nodes in the next transmission
denote the channel coefficients of TX1 − RX1 , TX2 − RX1 , time slot (i+1) can then be expressed as,
TX1 − RX2 , TX2 − RX2 , TX3 − RX1 , and TX4 − RX1 links
Yi+1 (t)
respectively. We model that all channels are independent and !
identically distributed Rayleigh fading channels because of hi1 + κ ĥi1 h p p i
= √ − a1 Pt X3∗ (t)− a2 Pt X4∗ (t)
the multi-path propagation of the acoustic signals in underwa- TLi1 (r, f )
ter communication. As a result, |h11 |2 , |h12 |2 , |h21 |2 , |h22 |2 , !
hi2 + κ ĥi2 hp i
|h13 |2 , and |h14 |2 are exponentially distributed random vari-
p
+ √ a1 Pt X1∗ (t)+ a2 Pt X2∗ (t) +ni (t).
ables with average power of β11 , β12 , β21 , β22 , β13 and β14 TLi2 (r, f )
respectively. (33)

32864 VOLUME 9, 2021

V. Goutham, V. P. Harigovindan: NOMA-CRS for UASNs Under I-SIC and I-CSI

For decoding the symbols, STBC decoding algorithm where i ∈ 1. The received signals at the node P in the next
can be applied to separate the composite signals (i.e., transmission time slot (i+3) can then be expressed as,
C1 (t) & C2 (t)) at both the nodes G and P. The STBC decoding !
hi3 + κ ĥi3 h p i
algorithm performs maximum-likelihood detection with lin- Yi+3 (t) = √ − 0.5 ∗ Pt X4∗ (t)
ear complexity. Subsequently, the node P does not perform TLi3 (r, f )
!
successive interference cancellation (SIC) because of poor hi4 +κ ĥi4 hp i
channel quality between TX1 − RX1 compared to TX1 − RX2 . + √ 0.5 ∗ Pt X2∗ (t) +ni (t). (38)
TLi4 (r, f )
For decoding the symbols X1 (t) & X3 (t), the node P treats
the symbols X2 (t) & X4 (t) as noise signals. As a result, The received SINRs at RX1 are given by,
the received signal-to-interference noise ratio (SINR) at RX1 2
!
|h1i | 01i
2
P
for symbol X1 can be given by,  a2
X2 i=1
2
! γP = max ! ,
2
|h1i |2 01i
P
a1 a1 ξ |h1i | 01i + ψ1
2
P
X1 i=1 i=1
γSP = ! , (34)
2 |h13 |2 013 + |h14 |2 014
 
|2 01i +ψ1 ,
P
a2 |h1i + (39)
i=1 2 ψ1
where, ψ1 = κ 2 ν13
2 0 + κ 2 ν 2 0 + 1. Due to the symmetry
13 14 14
where, ψ1 = κ 2 ν11
2 0 + κ 2 ν 2 0 + 1. Later, the near node
11 12 12 in transmission scheme, the SINR and ergodic rates achieved
decodes symbol X2 (t) & X4 (t) by performing the SIC. Firstly, by symbol X3 is exactly equal to X1 and X4 is equal to that
the node G decodes X1 (t) & X3 (t) symbols by considering of X2 .
X2 (t) & X4 (t) as noise signals. Later, it subtracts decoded
symbols X1 (t) & X3 (t) symbols from composite signals using A. ERGODIC RATE ANALYSIS
SIC technique to obtain X2 (t) & X4 (t) symbols. The received The ergodic rate of the X1 symbol in STBC-NOMA-CRS
SINRs at RX2 for symbols X1 & X2 are given by, scheme in UASNs is given by,
1 fu h
! Z  n oi
2 X1 X1
a1
P
|h2i | 02i
2 CX1 = E log2 1 + min γSP , γSG df , (40)
4 fl
X1 i=1
γSG = , (35) X1 X1
where γSP , and γSG
!
2 are the SINRs in UASNs at S and G
|2 02i + ψ2
P
a2 |h2i nodes calculated by using (34) and (35), respectively. From
i=1
! (40), we assume
 Ṽ1 nis an arbitrary
o variable, considered as
X1 X1
2
Ṽ1 = log2 1 + min γSP , γSG . By substituting (34) and
|h2i |2 02i
P
a2
X2 i=1 (35), we obtain variable Ṽ1 as,
γSG = ! , (36) !
2
 
2
a1 ξ |h2i |2 02i + ψ2
P
|h1i | 01i
2
 P
 a1



i=1  i=1
,

1 + min 
Ṽ1 = log2  !
2
where, ψ2 = κ 2 ν21 2 0 + κ 2 ν 2 0 + 1 and ξ denotes the
|h1i |2 01i +ψ1
  P
21 22 22  a2


SIC inefficiency. ξ varies in the range of [0, 1], ξ = 0 rep- i=1
resents perfect SIC and other values represent imperfect SIC
! 
2
with inefficiency of ξ . Due to the symmetry in transmission |h2i |2 02i
P 
a1 


X1 X X2 X4 
scheme, γSG = γSG3 & γSG = γSG . i=1 
! 
At the next transmission√time slot (T2 ), the signal trans- 2  
|h2i |2 02i + ψ2 
P  
mitted from the TX3 is 0.5Pt X2 (t) and from TX4 is a2 

√ i=1
0.5Pt X4√(t). During the next transmission time slot (T3 ),
2 2
( )!
signal − 0.5Pt X4∗ (t) is transmitted from the TX3 , and X |h1i |2 01i X |h2i |2 02i
√ ∗ = log2 1 + min ,
0.5Pt X2 (t) is transmitted from the TX4 . The received sig- ψ1 ψ2
i=1 i=1
nals at the node P in a given time slot i+2 can then be ( 2 2
)!
expressed as, X |h1i |2 01i X |h2i |2 02i
− log2 1 + a2 min , .
ψ1 ψ2
! i=1 i=1
hi3 + κ ĥi3 hp i
(41)
Yi+2 (t) = √ 0.5 ∗ Pt X2 (t)
TLi3 (r, f ) 1
! Let, Ũ and Ṽ are random variables denoted by, Ũ =
hi4 + κ ĥi4 hp i 2
|h1i |2 01i
2
1 P |h2i |2 02i
+ √ 0.5 ∗ Pt X4 (t) +ni (t), (37)
P
ψ1 , and Ṽ = ψ2 , respectively. Accordingly
TLi4 (r, f ) i=1 i=1

VOLUME 9, 2021 32865

 V. Goutham, V. P. Harigovindan: NOMA-CRS for UASNs Under I-SIC and I-CSI

!
2
the CDFs of random variable Ũ and Ṽ u are obtained as a2
P
|hji |2 0ji
− λu
FŨ (u) = 1 − λ11λ−λ λ12 1
−λ
,
11 11 + 12 , F (v) = 1 − i=1
12
e λ11 −λ12 e Ṽ Z̃j = ! j ∈ {1, 2}. The corresponding
2
λ21 − λu − λu βij 0ij
+ λ21λ−λ , where λij = a1 ξ |hji |2 0ji +ψj
P
22
λ21 −λ22 e e ψijn . We o
consider
21 22
22 i=1
1
W̃ is a random variable denoted by, W̃ = min Ũ , Ṽ . Based CDFs of Ỹ and Z̃j are given by,
on the random variable transformations, the CDFs of W̃ is λ̃13 −y
λ̃14 −y

obtained as, FỸ (y) = 1 − e λ̃13 + e λ̃14 . (47)

λ̃13 − λ̃14 λ̃13 − λ̃14
λ11 λ12 λ
   z
− λu − λu j2 −
FW̃ (w) = 1 − e 11 − e 12  1+ e a2 λj2 −ξ a1 λj2 z
λ − λ12 λ11 − λ12 λj1 − λj2


 11

λj1

z
λ21 λ22 a2

−

− u
e λ21 −
− u
e λ22 . (42) FZ̃j (z) = − e a2 λj1 −ξ a1 λj1 z ; z< (48)
λ21 − λ22 λ21 − λ22  λj1 − λj2 ξ a1
a2


,

1
1;
 z≥
Let X̃ = a2 W̃ . Based on the linear random variable trans- ξ a1
formation, the CDFs of X̃ is obtained as, FX̃ (x) = FW̃ ( ax2 ). where λ̃ij =
0ij βij
.
ψ̃i n
Using the random variables W̃ , and X̃ , (14) can be re-written 1
n o o
respectively as, Let, A = min max Ỹ , Z˜1 , Z˜2 , and the corresponding
CDF is obtained as,
1 fu ∞
Z Z  
CX1 = log2 1 + W̃ fW̃ (w) dw df    
4 fl 0 
 1 − 1 − FỸ (a)FZ˜1 (a)
a2

 
1 fu ∞ a< ,
Z Z  
FA (a) = 1 − FZ˜2 (a) ; (49)
− log2 1 + X̃ fX̃ (x) dx df . (43) ξ a1
4 fl 0  a2
.

1; a≥


The mathematical expression for ergodic rate for symbol ξ a1
X1 is obtained as, Using the random variable A, (45) can be re-written as,
Z fu
1 fu ∞
   
1 1
1 1
Z Z
CX1 = C1 −λ11 λ21 e λ11 λ21 Ei − + df CX2 = log2 (1 + A) fA (a) da df
fl λ11 λ21 4 fl 0
Z fu 
1 1
 
1 1
 Z fu Z a2
−λ11 λ22 e λ11 + λ22 1 ξ a1 1 − FA (a)
+ Ei −
λ11
+
λ22
df = da df . (50)
fl 4 ln 2 fl 0 1+a
Z fu    
1
+ 1 1 1 The overall
+ −λ12 λ21 e λ12 λ21 Ei − + df P ergodic rate achieved in STBC-NOMA-CRS is
fl λ12 λ21 given by, 4i=1 CXi .
Z fu     
1
+ 1 1 1
− −λ12 λ22 e λ12 λ22 Ei − + df , IV. NUMERICAL AND SIMULATION RESULTS
fl λ12 λ22
(44) In this section, analytical and Monte Carlo simulation results
are provided to describe our investigations on the impact
where C1 = 4 ln 2(λ11 −λ112 )(λ21 −λ22 ) . of I-CSI and I-SIC on ergodic rate, outage probability and
Similarly, the ergodic rate for X2 symbol is given by, energy efficiency of the NOMA-CRS scheme. We have used
Monte Carlo simulations to generate the Rayleigh fading
1 fu h
Z  n oi
CX2 = E log2 1 + min γPX2 , γSG X2
df . (45) coefficients for the channel links and MATLAB R
2018b with
4 fl TM
Phased Array System Toolbox to simulate the underwater
From (45), we assume Ṽ2nis an arbitrary channel. The results are compared with the conventional
 o variable, considered
as Ṽ2 = log2 1 + min γPX2 , γSG X2
. By substituting (39) decode-and-forward based CRS scheme in UASNs [4]. From
the system model, we recall that node P is far from node S
and (36), we obtain variable Ṽ2 as, (46) as shown at the bottom with poor channel conditions and node G is close to node
of the page. S with good channel conditions. Without loss of generality,
We assume Ỹ and Z̃ are random variables, which are we set, DSG = 0.1 DSP , θ = 60◦ , a1 = 0.95 and a2 = 0.05.
1 |h13 |2 013 +|h14 |2 014
 
denoted as, Ỹ = , and The remaining parameters used for simulation and analytical
˜ 2 ψ1

  ! ! 
2 2
|h1i | 01i
2 |h2i | 02i
2
 P P 

 a2 a2 

2 0 + |h |2 0
     

 i=1 |h13 | 13 14 14 i=1
, ,
 
Ṽ2 = log2 1 + min max  (46)
2 ψ̃1
! !
  2 2 
ξ 20 ψ ξ 20 ψ
  P P 


 a1 |h 1i | 1i + 2 a 1 |h 2i | 2i + 
2

i=1 i=1

32866 VOLUME 9, 2021

V. Goutham, V. P. Harigovindan: NOMA-CRS for UASNs Under I-SIC and I-CSI

TABLE 2. Parameters used for performance analysis.

FIGURE 3. Impact of the CSI on the ergodic sum rate achieved by the
NOMA-CRS and CRS schemes with respect to distance between the S-P
link in shallow water: by setting H = 10 m, k = 1, DSG = 0.1 DSP , θ = 60◦ ,
ξ = 0 (perfect SIC), a1 = 0.95 and a2 = 0.05.

analysis are listed in Table. 2 and these values took from

Evologics R
practical underwater acoustic modem [41]. First,
we have analysed the impact of I-CSI considering the perfect
SIC. Then, assuming perfect CSI, the impact of I-SIC on the
ergodic sum rate, outage probability and energy efficiency is
analysed with respect to distance between the S-P link in both
shallow and deep water scenarios. Later, we have analysed
the impact of the relay position on the NOMA-CRS scheme.
Finally, the impact of wind speed and shipping activities on
the NOMA-CRS is analysed considering realistic underwater FIGURE 4. Impact of the CSI on the outage probability behaviors the
scenario. NOMA-CRS and CRS schemes with respect to distance between the S-P
link in shallow water: by setting H = 10 m, k = 1, DSG = 0.1 DSP , θ = 60◦ ,
ξ = 0 (perfect SIC), a1 = 0.95, a2 = 0.05 and R1 = 1.5, R2 = 1.75 bits per
A. SHALLOW WATER channel usage.
In this subsection, we analyse the impacts of the CSI and
SIC on the ergodic sum rate, outage probability and energy
efficiency with respect to the distance between the S-P link Figure 4 depicts the outage probability behaviours of
in shallow water scenario. NOMA-CRS (i.e. symbols X1 and X2 ) and CRS schemes
with respect to the distance between the S-P link in shallow
1) IMPACT OF CSI water by setting R1 = 1.5, R2 = 1.75 bits per channel
Figure 3 shows the impact of CSI on the ergodic sum-rate usage. As shown in Fig. 4, the outage behaviour of CRS is
achieved by the NOMA-CRS and CRS schemes with respect superior to that of NOMA-CRS for various distances between
to the distance between the S-P link in shallow water scenario. the S-P link. The reason behind the degradation of outage
In particular, it has been observed that the ergodic rate of behaviour in NOMA-CRS is superposition coding, in which
the NOMA-CRS and CRS decreases with the increase in the X1 symbol affects the outage behaviour of X2 symbol and
the distance between the S-P link. Because of the SNRs vice versa. Additionally, it can also be observed that, with the
dependence on the respective transmission distance. Another increase in values of channel estimation error factor, the out-
important fact from Fig. 3 is that the ergodic gains achieved age probability behaviours of both CRS and NOMA-CRS
by the NOMA-CRS compared to CRS in UASNs are reduced schemes are significantly degraded. As a result, the search
by increasing the values of channel estimation error factor for efficient techniques for perfect estimation of the CSIs in
(i.e., κ). This is because multiple symbols are transmitted in UASNs is crucial, which is open to the research community.
a specific time slot using superposition coding in the NOMA- Figure 5 plots the energy efficiency curves of NOMA-CRS
CRS. As a consequence, the increase in values of channel and CRS schemes with respect to the distance between the
estimation error factor has an impact on the performance of S-P link in shallow water. The energy efficiency performance
both X1 and X2 symbols in NOMA-CRS for UASNs. of NOMA-CRS is observed to be better than that of CRS for

VOLUME 9, 2021 32867

 V. Goutham, V. P. Harigovindan: NOMA-CRS for UASNs Under I-SIC and I-CSI

water. It can be noted that, from the mathematical modelling

presented in subsection II-B2, the impact of I-SIC only affects
the outage behaviour of X2 symbol in NOMA-CRS. It can
be observed that, with the increase in the values of SIC
inefficiency factor, the outage behaviour of X2 is deterio-
rating, as shown in Fig. 7. Further, the outage performance
of X2 symbol is approximately the same as the perfect SIC
NOMA-CRS scheme when a small amount of I-SIC occurs
at ξ = 1×10−6 . Figure 8 plots the energy efficiency curves of
NOMA-CRS and CRS schemes with respect to the distance
between the S-P link in shallow water, taking into account
I-SIC. It can be seen that NOMA-CRS is able to achieve better
energy efficiency compared to the CRS scheme for the case
FIGURE 5. Impact of the CSI on the energy efficiency performance of the
of perfect SIC and lower values of ξ ≤ 1 × 10−6 . For the case
NOMA-CRS and CRS schemes with respect to the distance between the of ξ > 1×10−4 , the performance of the NOMA-CRS scheme
S-P link in shallow water: by setting H = 10 m, k = 1, DSG = 0.1 DSP ,
θ = 60◦ , ξ = 0 (perfect SIC), a1 = 0.95, a2 = 0.05 and
is significantly degraded and is slightly lower than the CRS
R1 = 1.5, R2 = 1.75 bits per channel usage. scheme.

lower values of channel estimation error factor (κ ≈ 0.001).

For higher values of κ, it is observed that the energy efficiency
of NOMA-CRS is degraded, and is slightly lower than that of
the CRS scheme.

2) IMPACT OF SIC
Here, we analyse the performance of ergodic rate, outage
probability and energy efficiency of NOMA-CRS and CRS
schemes for the different levels of I-SIC (i.e., by varying
the SIC inefficiency factor (ξ )). It is noted from Fig. 6
that, we obtain the lowest ergodic rate performance of
NOMA-CRS scheme at a high level of I-SIC (for values
above ξ = 1 × 10−4 ), which is even lower than the CRS
regime. It is also noted that the performance gap between the
perfect SIC and the I-SIC is high at low distances between the FIGURE 7. Impact of the I-SIC on the outage probability behaviours the
NOMA-CRS and CRS schemes with respect to distance between the S-P
S-P link. Figure 7 depicts the outage probability behaviour of link in shallow water: by setting H = 10 m, k = 1, DSG = 0.1 DSP , θ = 60◦ ,
the NOMA-CRS (i.e. symbols X1 and X2 ) and CRS schemes κ = 0 (perfect CSI), a1 = 0.95, a2 = 0.05 and R1 = 1.5, = R2 = 1.75 bits
with respect to the distance between the S-P link in shallow per channel usage.

FIGURE 8. Impact of the I-SIC on the energy efficiency performance of

FIGURE 6. Impact of the I-SIC on the ergodic sum rate achieved by the the NOMA-CRS and CRS schemes with respect to the distance between
NOMA-CRS and CRS schemes with respect to the distance between the the S-P link in shallow water: by setting H = 10 m, k = 1, DSG = 0.1 DSP ,
S-P link in shallow water: by setting H = 10 m, k = 1, DSG = 0.1 DSP , θ = 60◦ , κ = 0 (perfect CSI), a1 = 0.95, a2 = 0.05 and
θ = 60◦ , κ = 0 (perfect CSI), a1 = 0.95 and a2 = 0.05. R1 = 1.5, = R2 = 1.75 bits per channel usage.

32868 VOLUME 9, 2021

V. Goutham, V. P. Harigovindan: NOMA-CRS for UASNs Under I-SIC and I-CSI

B. DEEP WATER
In this subsection, we analyse the impacts of the CSI and
SIC on the ergodic sum rate, outage probability and energy
efficiency with respect to the distance between the S-P link
in deep water scenario.

1) IMPACT OF CSI
Figure 9 shows the impact of CSI on the ergodic sum rate
with respect to the distance between S-P link in deep water
scenario. In Fig. 9, ergodic rate response in deep water
is observed to be similar to shallow water scenario. But,
the ergodic rate performance has deteriorated rapidly with
respect to distance when compared to shallow water sce-
nario. This is due to the fact that, acoustic signals experience FIGURE 10. Impact of the I-CSI on the outage probability behaviours the
NOMA-CRS and CRS schemes with respect to distance between the S-P
spherical spreading which results in high transmission losses link in deep water: by setting H = 1000 m, k = 2, DSG = 0.1 DSP , θ = 60◦ ,
in deep water. The performance of NOMA-CRS and CRS ξ = 0 (perfect SIC), a1 = 0.95, a2 = 0.05 and R1 = 1.5, R2 = 1.75 bits per
channel usage.
schemes have significantly deteriorated with the increase in
I-CSI in deep water, as shown in Fig. 9. Figure 10 depicts
the impact of CSI on the outage probability with respect to
the distance between S-P link in deep water scenario. Similar
to the ergodic rate performance, the outage behaviour also
showed a similar response in deep water as in shallow water.
However, the outage behaviour is rapidly conversing to a
value one with respect to distance between S-P link in deep
water. In Fig. 10, it can also be observed that the I-CSI
severely affecting the performance of outage probability of
NOMA-CRS and CRS schemes in deep water. Figure 11
shows the energy efficiency performance of the NOMA-CRS
and CRS schemes with respect to the distance between S-P
link. It can be seen that, NOMA-CRS is capable of achieving
better energy efficiency compared to the CRS scheme. Fur-
ther, a significant observation from Fig. 11 is that the energy
FIGURE 11. Impact of the I-CSI on the energy efficiency performance of
efficiency shows a better performance for d < 400 m, but the NOMA-CRS and CRS schemes with respect to the distance between
for d ≥ 400 m the energy efficiency of NOMA-CRS scheme the S-P link in deep water: by setting H = 1000 m, k = 2, DSG = 0.1 DSP ,
θ = 60◦ , ξ = 0 (perfect SIC), a1 = 0.95, a2 = 0.05 and R1 = 1.5, R2 = 1.75
significantly decreases due to spherical spreading. bits per channel usage.

2) IMPACT OF SIC
Figure 12 shows the effect of SIC on the ergodic sum rate with
respect to the distance between the S-P link in deep water scenario. The ergodic rates of NOMA-CRS have gradually
decreased with the gradual increase in the impact of SIC in
deep water similar to the shallow water scenario. Figure 13
depicts the impact of SIC on the outage probability with
respect to the distance between the S-P link in deep water
scenario. The increase in the SIC imperfections has a sub-
stantial decrease in the outage probability of X2 symbol of
NOMA-CRS. Figure 14 shows the impact of SIC on energy
efficiency with respect to the distance between the S-P link
in deep water scenario. It can be seen that NOMA-CRS is
capable of achieving better energy efficiency compared to
the CRS scheme for the lower values of the SIC inefficiency
factor up to ξ = 1 × 10−6 . Further, it can be observed from
Fig. 14 that the energy efficiency performance is better for
d < 400 m, but for d ≥ 400 m the energy efficiency of
FIGURE 9. Impact of the I-CSI on the ergodic sum rate achieved by the
NOMA-CRS and CRS schemes with respect to distance between the S-P
NOMA-CRS scheme significantly decreases due to spherical
link in deep water: by setting H = 1000 m, k = 2, DSG = 0.1 DSP , θ = 60◦ , spreading. Finally, it can be observed that the SIC also has an
ξ = 0 (perfect SIC), a1 = 0.95 and a2 = 0.05. impact similar to shallow water scenario on the ergodic rate,

VOLUME 9, 2021 32869

 V. Goutham, V. P. Harigovindan: NOMA-CRS for UASNs Under I-SIC and I-CSI

FIGURE 12. Impact of the I-SIC on the ergodic sum rate achieved by the FIGURE 14. Impact of the I-SIC on the energy efficiency performance of
NOMA-CRS and CRS schemes with respect to the distance between the the NOMA-CRS and CRS schemes with respect to the distance between
S-P link in deep water: by setting H = 1000 m, k = 2, DSG = 0.1 DSP , the S-P link in deep water: by setting H = 1000 m, k = 2, DSG = 0.1 DSP ,
θ = 60◦ , κ = 0 (perfect CSI), a1 = 0.95 and a2 = 0.05. θ = 60◦ , κ = 0 (perfect CSI), a1 = 0.95, a2 = 0.05 and
R1 = 1.5, R2 = 1.75 bits per channel usage.

FIGURE 13. Impact of the I-SIC on the outage probability behaviours the
NOMA-CRS and CRS schemes with respect to distance between the S-P FIGURE 15. Impact of the position of the relay node on the ergodic sum
link in deep water: by setting H = 1000 m, k = 2, DSG = 0.1 DSP , θ = 60◦ , rate achieved by the NOMA-CRS and CRS schemes with respect to the
κ = 0 (perfect CSI), a1 = 0.95, a2 = 0.05 and R1 = 1.5, R2 = 1.75 bits per distance between the S-P link in underwater: by setting k = 1.5, θ = 60◦ ,
channel usage. (ξ, κ) = (1 × 10−6 , 1 × 10−3 ), a1 = 0.95, and a2 = 0.05.

outage probability, and energy efficiency of NOMA-CRS in

deep water. DSG = 0.1 DSP (node G is existed near to the node S
It can be concluded from both the shallow and deep water compared to P), and DSG = 0.9 DSP (node G is existed near to
scenarios that, the performance of NOMA-CRS is signifi- the node P compared to S) by setting θ = 60◦ . It is noted from
cantly affected by the CSI estimation error factor compared Fig. 15 that, the ergodic rate performance of NOMA-CRS
to the CSR scheme. As a result, there exists a trade-off approaching to ergodic rate performance of CRS, when the
between the ergodic rate and the energy efficiency because relay node is moves away from the source node. It is also
of outage probability dependence on CSI estimation error evident that NOMA-CRS is capable of achieving higher
factor. As the technological advances in the CSI estimation ergodic rates even in the practical scenario in which impacts
techniques rapidly evolving will guarantee the performance of both I-CSI and I-SIC are considered.
enhancement of the NOMA-CRS scheme. Further, there is a
need for efficient SIC techniques are required to get higher 4) IMPACT OF ENVIRONMENTAL EFFECTS NAMELY, WIND
performance gains of NOMA-CRS scheme compared to the SPEED AND SHIPPING ACTIVITY
CRS scheme. In this subsection, we discussed the performance analysis of
NOMA-CRS in terms of ergodic sum rate by considering
3) IMPACT OF RELAY POSITION the effects of environmental parameters, namely shipping
In this subsection, the ergodic rate performance of noises, and wind speed (surface motion). The wind speed
NOMA-CRS and CRS schemes are analysed with the change (w) mainly varies approximately in the range of 1 − 60 m/s.
in the position of the relay node in a practical underwater The shipping noise depends on shipping activity factor (s),
scenario. Here, the relay node position is changed using which varies in the range of [0, 1], where 0 and 1 represent

32870 VOLUME 9, 2021

V. Goutham, V. P. Harigovindan: NOMA-CRS for UASNs Under I-SIC and I-CSI

TABLE 3. Ergodic rate (in kbps) analysis of STBC-NOMA-CRS and NOMA-CRS.

ergodic rate, outage probability and energy efficiency for

NOMA-CRS were derived by considering underwater spe-
cific channel characteristics. Impact of imperfect channel
state information (I-CSI) and imperfect successive interfer-
ence cancellation (I-SIC) on the performance of NOMA-CRS
is thoroughly investigated. From the analytical and simulation
results, it can be observed that NOMA-CRS can achieve
significant improvement in ergodic sum rate and energy
efficiency, at the cost of a slight degradation in the outage
performance. I-CSI and I-SIC have considerable influence
on the performance of NOMA-CRS. The performance of
NOMA-CRS in UASNs is better when the relay node is
placed near to the source node. Finally, we have analysed
FIGURE 16. Impact of environmental effects namely, wind speed and
the impact of environmental effects, namely wind speed and
shipping activity on the ergodic sum rate achieved by the NOMA-CRS and shipping activities on the NOMA-CRS. From the results,
CRS schemes with respect to distance between the S-P link: by setting
it is observed that that high-speed winds and high shipping
k = 1.5, DSG = 0.1 DSP , θ = 60◦ , (ξ, κ) = (1 × 10−6 , 1 × 10−3 ), a1 = 0.95,
and a2 = 0.05. activities can severely degrade the performance of ergodic
sum rate and energy efficiency of the NOMA-CRS. Finally,
we have proposed STBC-NOMA-CRS for UASNs. From the
low and high shipping activities, respectively [36]. As a results, it is evident that this scheme can provide performance
result, we formed three different pairs of wind velocities and enhancement same as NOMA-CRS ascheme for UASNs,
shipping activities as shown in Fig. 16. It is evident from without transmitter CSI.
the results that high-speed winds and high shipping activities
severely degrade the performance of ergodic sum rate of the REFERENCES
NOMA-CRS. [1] M. Jouhari, K. Ibrahimi, H. Tembine, and J. Ben-Othman, ‘‘Underwater
wireless sensor networks: A survey on enabling technologies, localiza-
tion protocols, and Internet of underwater things,’’ IEEE Access, vol. 7,
5) COMPARATIVE ANALYSIS OF STBC-NOMA-CRS AND pp. 96879–96899, 2019.
NOMA-CRS [2] C.-C. Kao, Y.-S. Lin, G.-D. Wu, and C.-J. Huang, ‘‘A comprehensive study
on the Internet of underwater things: Applications, challenges, and channel
In this subsection, we compare the performance of STBC- models,’’ Sensors, vol. 17, no. 7, p. 1477, Jun. 2017. [Online]. Available:
NOMA-CRS and NOMA-CRS schemes, as shown in Table 3. https://www.mdpi.com/1424-8220/17/7/1477
In NOMA-CRS, we assume that perfect CSI is available at [3] M. C. Domingo, ‘‘An overview of the Internet of underwater things,’’
J. Netw. Comput. Appl., vol. 35, no. 6, pp. 1879–1890, 2012.
the transmitter. From the results, it is evident that the STBC- [Online]. Available: http://www.sciencedirect.com/science/article/
NOMA-CRS is able to provide the same performance of pii/S1084804512001646
NOMA-CRS scheme, even without the CSI at the transmitter. [4] S. Al-Dharrab, M. Uysal, and T. M. Duman, ‘‘Cooperative underwater
acoustic communications [accepted from open call],’’ IEEE Commun.
A slight improvement in the ergodic rate performance is also Mag., vol. 51, no. 7, pp. 146–153, Jul. 2013.
observed as shown in Table 3. It is due to the benefits of both [5] L. Dai, B. Wang, Y. Yuan, S. Han, I. Chih-Lin, and Z. Wang, ‘‘Non-
spectral efficiency and transmit diversity, which makes this orthogonal multiple access for 5G: Solutions, challenges, opportuni-
ties, and future research trends,’’ IEEE Commun. Mag., vol. 53, no. 9,
scheme perfectly suitable for practical UASNs. pp. 74–81, Sep. 2015.
[6] Y. Liu, Z. Qin, M. Elkashlan, Z. Ding, A. Nallanathan, and L. Hanzo,
V. CONCLUSION ‘‘Nonorthogonal multiple access for 5G and beyond,’’ Proc. IEEE,
vol. 105, no. 12, pp. 2347–2381, Dec. 2017.
The objective of this research work is to improve the perfor- [7] E. G. Larsson, O. Edfors, F. Tufvesson, and T. L. Marzetta, ‘‘Massive
mance of cooperative relaying strategy in UASNs by incor- MIMO for next generation wireless systems,’’ IEEE Commun. Mag.,
porating NOMA. We have proposed a NOMA-CRS scheme vol. 52, no. 2, pp. 186–195, Feb. 2014.
[8] L. Lu, G. Y. Li, A. L. Swindlehurst, A. Ashikhmin, and R. Zhang, ‘‘An
for enhancing the performance of the bandwidth-limited and overview of massive MIMO: Benefits and challenges,’’ IEEE J. Sel. Topics
energy-constrained UASNs. Mathematical expressions for Signal Process., vol. 8, no. 5, pp. 742–758, Oct. 2014.

VOLUME 9, 2021 32871

 V. Goutham, V. P. Harigovindan: NOMA-CRS for UASNs Under I-SIC and I-CSI

[9] S. M. R. Islam, N. Avazov, O. A. Dobre, and K.-S. Kwak, ‘‘Power-domain [31] R. J. Urick, Principles Of Underwater Sound, 3rd ed. Los Altos, CA, USA:
non-orthogonal multiple access (NOMA) in 5G systems: Potentials and Peninsula Publishing, 1983.
challenges,’’ IEEE Commun. Surveys Tuts., vol. 19, no. 2, pp. 721–742, [32] K. V. Mackenzie, ‘‘Nine-term equation for sound speed in the oceans,’’
2nd Quart., 2017. J. Acoust. Soc. Amer., vol. 70, no. 3, pp. 807–812, Sep. 1981.
[10] L. Ma, S. Zhou, G. Qiao, S. Liu, and F. Zhou, ‘‘Superposition coding [33] K. S. Geethu and A. V. Babu, ‘‘A hybrid ARQ scheme combining erasure
for downlink underwater acoustic OFDM,’’ IEEE J. Ocean. Eng., vol. 42, codes and selective retransmissions for reliable data transfer in under-
no. 1, pp. 175–187, Jan. 2017. water acoustic sensor networks,’’ EURASIP J. Wireless Commun. Netw.,
[11] Ahmad and Chang, ‘‘Downlink power allocation strategy for next- vol. 2017, no. 1, p. 32, Dec. 2017.
generation underwater acoustic communications networks,’’ Electronics, [34] R. E. Francois and G. R. Garrison, ‘‘Sound absorption based on ocean
vol. 8, no. 11, p. 1297, Nov. 2019. measurements: Part I: Pure water and magnesium sulfate contributions,’’
[12] M. Jain, N. Sharma, A. Gupta, D. Rawal, and P. Garg, ‘‘Performance anal- J. Acoust. Soc. Amer., vol. 72, no. 3, pp. 896–907, Sep. 1982, doi:
ysis of NOMA assisted underwater visible light communication system,’’ 10.1121/1.388170.
IEEE Wireless Commun. Lett., vol. 9, no. 8, pp. 1291–1294, Aug. 2020. [35] R. E. Francois and G. R. Garrison, ‘‘Sound absorption based on ocean mea-
[13] M. J. Bocus, D. Agrafiotis, and A. Doufexi, ‘‘Non-orthogonal multiple surements. Part II: Boric acid contribution and equation for total absorp-
access (NOMA) for underwater acoustic communication,’’ in Proc. IEEE tion,’’ J. Acoust. Soc. Amer., vol. 72, no. 6, pp. 1879–1890, Dec. 1982, doi:
88th Veh. Technol. Conf. (VTC-Fall), Aug. 2018, pp. 1–5. 10.1121/1.388673.
[14] E. A. Makled, A. Yadav, O. A. Dobre, and R. D. Haynes, ‘‘Hierarchical [36] M. Stojanovic, ‘‘On the relationship between capacity and distance in
full-duplex underwater acoustic network: A NOMA approach,’’ in Proc. an underwater acoustic communication channel,’’ in Proc. 1st ACM Int.
OCEANS MTS/IEEE Charleston, Oct. 2018, pp. 1–6. workshop Underwater Netw. (WUWNet), New York, NY, USA, 2006,
[15] J.-B. Kim and I.-H. Lee, ‘‘Capacity analysis of cooperative relaying sys- pp. 41–47.
tems using non-orthogonal multiple access,’’ IEEE Commun. Lett., vol. 19, [37] Y. Zhang, R. Venkatesan, O. A. Dobre, and C. Li, ‘‘Efficient estimation and
no. 11, pp. 1949–1952, Nov. 2015. prediction for sparse time-varying underwater acoustic channels,’’ IEEE J.
[16] Z. Ding, M. Peng, and H. V. Poor, ‘‘Cooperative non-orthogonal multiple Ocean. Eng., vol. 45, no. 3, pp. 1112–1125, Jul. 2019.
access in 5G systems,’’ IEEE Commun. Lett., vol. 19, no. 8, pp. 1462–1465, [38] Y. Zhou and F. Tong, ‘‘Channel estimation based equalizer for underwater
Aug. 2015. acoustic Multiple-Input-Multiple-Output communication,’’ IEEE Access,
[17] Y. Zhang, Z. Yang, Y. Feng, and S. Yan, ‘‘Performance analysis of coopera- vol. 7, pp. 79005–79016, 2019.
tive relaying systems with power-domain non-orthogonal multiple access,’’ [39] D. Zwillinger, V. Moll, I. Gradshteyn, and I. Ryzhik, Eds., ‘‘3-4—Definite
IEEE Access, vol. 6, pp. 39839–39848, Jul. 2018. integrals of elementary functions,’’ in Table of Integrals, Series, and Prod-
[18] Y. Chen, X. Jin, L. Wan, X. Zhang, and X. Xu, ‘‘Selective dynamic coded ucts, 8th ed. Boston, MA, USA: Academic, 2015, pp. 249–519.
cooperative communications for multi-hop underwater acoustic sensor [40] X. Yue, Y. Liu, S. Kang, A. Nallanathan, and Y. Chen, ‘‘Modeling and
networks,’’ IEEE Access, vol. 7, pp. 70552–70563, 2019. analysis of two-way relay non-orthogonal multiple access systems,’’ IEEE
[19] Z. Liu, F. Bai, and L. Song, ‘‘Underwater decode-interleave-forward coop- Trans. Commun., vol. 66, no. 9, pp. 3784–3796, Sep. 2018.
erative strategy for underwater acoustic communication,’’ IEEE Access, [41] Underwater Acoustic Modems. Accessed: Nov. 17, 2020. [Online]. Avail-
vol. 7, pp. 19698–19708, 2019. able: https://evologics.de/acoustic-modems
[20] A. Ghosh, J.-W. Lee, and H.-S. Cho, ‘‘Throughput and energy efficiency
of a cooperative hybrid ARQ protocol for underwater acoustic sensor net-
works,’’ Sensors, vol. 13, no. 11, pp. 15385–15408, Nov. 2013. [Online].
Available: https://www.mdpi.com/1424-8220/13/11/15385
[21] V. Goutham and V. P. Harigovindan, ‘‘Modeling and analysis of hybrid
ARQ scheme for incremental cooperative communication in underwater VEERAPU GOUTHAM (Member, IEEE) received
acoustic sensor networks,’’ Iranian J. Sci. Technol., Trans. Electr. Eng., the bachelor’s degree in electronics and commu-
vol. 45, no. 1, pp. 279–294, Mar. 2021. nication engineering from Sri Krishnadevaraya
[22] R. Wang, A. Yadav, E. A. Makled, O. A. Dobre, R. Zhao, and University, Anantapur, and the master’s degree in
P. K. Varshney, ‘‘Optimal power allocation for full-duplex underwa- digital communications from the National Insti-
ter relay networks with energy harvesting: A reinforcement learning tute of Technology Bhopal, India. He is currently
approach,’’ IEEE Wireless Commun. Lett., vol. 9, no. 2, pp. 223–227, pursuing the Ph.D. degree with the Department
Feb. 2020. of Electronics and Communication Engineer-
[23] D. D. Tan and D.-S. Kim, ‘‘Cooperative transmission scheme for multi- ing, National Institute of Technology Puducherry,
hop underwater acoustic sensor networks,’’ Int. J. Commun. Netw. Dis- India. His research interests include wireless net-
trib. Syst., vol. 14, no. 1, pp. 1–18, Nov. 2015, doi: 10.1504/IJC- works, wireless communications, and underwater acoustic sensor networks.
NDS.2015.065998.
[24] R. Cao, F. Qu, and L. Yang, ‘‘Asynchronous amplify-and-forward relay
communications for underwater acoustic networks,’’ IET Commun.,
vol. 10, no. 6, pp. 677–684, Apr. 2016.
[25] S. Al-Dharrab, A. Muqaibel, and M. Uysal, ‘‘Performance of multicarrier
cooperative communication systems over underwater acoustic channels,’’ V. P. HARIGOVINDAN (Senior Member, IEEE)
IET Commun., vol. 11, no. 12, pp. 1941–1951, Aug. 2017. received the B.Tech. degree in electronics and
[26] M. A. Rahman, Y. Lee, and I. Koo, ‘‘EECOR: An energy-efficient coop- communication engineering from the University
erative opportunistic routing protocol for underwater acoustic sensor net- of Calicut, the M.Tech. degree (with First Rank)
works,’’ IEEE Access, vol. 5, pp. 14119–14132, 2017. in digital electronics and communication systems
[27] A. Yahya, S. U. Islam, M. Zahid, G. Ahmed, M. Raza, H. Pervaiz, and from Visvesvaraya Technological University, and
F. Yang, ‘‘Cooperative routing for energy efficient underwater wireless the Ph.D. degree from the National Institute of
sensor networks,’’ IEEE Access, vol. 7, pp. 141888–141899, 2019. Technology Calicut, in 2013. He is currently an
[28] S. Ghoreyshi, A. Shahrabi, and T. Boutaleb, ‘‘A novel cooperative oppor-
Assistant Professor with the Department of Elec-
tunistic routing scheme for underwater sensor networks,’’ Sensors, vol. 16,
tronics and Communication Engineering and an
no. 3, p. 297, Feb. 2016. [Online]. Available: https://www.mdpi.com/1424-
8220/16/3/297
Associate Dean (Faculty Welfare) with the National Institute of Technology
[29] H. Tran-Dang and D.-S. Kim, ‘‘Channel-aware energy-efficient two-hop Puducherry (under Ministry of Education, Government of India). He has
cooperative routing protocol for underwater acoustic sensor networks,’’ more than 25 International publications to his credit. He is currently serving
IEEE Access, vol. 7, pp. 63181–63194, 2019. as the Principal Investigator for two funded projects one from the Department
[30] M. C. Domingo and R. Prior, ‘‘Energy analysis of routing pro- of Science and Technology, Government of India, and one from the Science
tocols for underwater wireless sensor networks,’’ Comput. Com- and Engineering Research Board. His research interests include wireless
mun., vol. 31, no. 6, pp. 1227–1238, Apr. 2008. [Online]. Available: networks and wireless communications.
http://www.sciencedirect.com/science/article/pii/S0140366407004689

32872 VOLUME 9, 2021