SPECIAL SECTION ON EMERGING TRENDS, ISSUES AND CHALLENGES

IN UNDERWATER ACOUSTIC SENSOR NETWORKS

Received May 31, 2019, accepted June 8, 2019, date of publication July 1, 2019, date of current version August 14, 2019.
Digital Object Identifier 10.1109/ACCESS.2019.2926158

A Load-Based Hybrid MAC Protocol for

Underwater Wireless Sensor Networks
ZIWEI ZHANG1 , WEI SHI1 , QIUNA NIU1 , YING GUO1 , JINGJING WANG 1, (Member, IEEE),
AND HANJIANG LUO 2 , (Member, IEEE)
1 School of Information Science and Technology, Qingdao University of Science and Technology, Qingdao 266000, China
2 School of Computer Science and Engineering, Shandong University of Science and Technology, Qingdao 266510, China
Corresponding authors: Wei Shi ([email protected]) and Jingjing Wang ([email protected])
This work was supported in part by the National Natural Science Foundation of China (NSFC) under Grant U1806201 and Grant
61671261, in part by the Key Research and Development Program of Shandong Province under Grant 2016GGX101007, in part by the
University Science and Technology Planning Project of Shandong Province under Grant J17KA058 and Grant J17KB154, in part by the
Natural Science Foundation of Shandong Province under Grant ZR2017BF036, and in part by the Applied Basic Research Projects of
Qingdao under Grant 17-1-1-4-jch.

ABSTRACT In underwater wireless sensor networks (UWSNs), media access control (MAC) is important
because it may have a significant impact on network performance; however, the complex and changeable
underwater communication environment is a great challenge for the MAC protocol. In flowing water,
the network nodes are constantly moving, and the number of competitors in the network also varies. The
existing hybrid MAC protocol neither can adapt to the dynamic network load nor can switch the access
control protocols with changing network loads, which may result in poor network performance. In order
to solve the above problems, this paper proposes a load-based time slot allocation (LBTSA) protocol.
The LBTSA selects the slot allocation scheme, from a set of possible schemes, according to the instantaneous
network load. Then, based on the relative priority of the nodes and the optimal number of backoff stages,
the host node selects the optimal access control protocol. This not only adapts well to changing network
loads but also maximizes network throughput. By assuming that the number of competitors obeys a universal
Poisson distribution, the LBSTA protocol and the HCR (a hybrid MAC protocol using channel reservation)
protocol are compared. As the results show, the throughput of the LBTSA is higher than that of the HCR
protocol, and the end-to-end delay of the LBTSA is lower than that of the HCR protocol.

INDEX TERMS MAC protocol, network load, time slot allocation, underwater wireless sensor networks.

I. INTRODUCTION nodes need to be precisely aligned [4]. However, the mis-

Human underwater activities in oceans have grown rapidly in alignment of optical communication nodes is unavoidable,
recent years; a huge number of sensors, actuators and various especially for mobile sensor nodes [5]. In order to pro-
types of vehicles have already been deployed underwater. vide electromagnetic communication using low frequency
Underwater equipment with communication functions can be signals, high transmission powers and large antennas are
used to construct the Internet of Underwater Things (IoUT) required, which are impractical in underwater environments.
[1]. Thus, underwater wireless networks have become an However, an underwater wireless communication (UWC)
increasingly popular research topic over the last decade [2]. system, which functions at low frequencies and provides
Currently, there are several underwater wireless acceptable signal attenuation is necessary [6]. Sound waves
communication technologies (e.g., acoustic, optical and elec- can travel long distances in seawater and experience low
tromagnetic wave), which can be applied to UWSNs [3]. media absorption rates; thus, they are suitable for remote
Underwater optical communication is affected by ocean underwater communications, therefore they are the pre-
turbidity, light scattering and background ambient light ferred wireless communication technology for developing
noise. Moreover, optical communication requires a line-of- reliable UWSNs [7].
sight link in point-to-point style, i.e., two communication In the past few decades, interest in UWSNs has grown
rapidly, as UWSNs can meet the communication needs of
The associate editor coordinating the review of this manuscript and a variety of underwater applications. For example, data
approving it for publication was Guangjie Han. collection is a significant application [8]. Various types of

104542 This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see http://creativecommons.org/licenses/by/4.0/ VOLUME 7, 2019
Z. Zhang et al.: Load-Based Hybrid MAC Protocol for UWSNs

data (temperature, humidity, image) can be collected by In addition, to save some energy during the listening mode,
deploying a large number of nodes that can sense environ- the node also estimates channel quality by transmitting a
mental data at locations of interest [9]. It then becomes nec- beacon to adjust its competition window. Although the power
essary to transmit the collected sensing data to the target node consumption of p-BORE is small, the network throughput of
safely and reliably. p-BORE is still low. For data collection in UWSNs, the data-
The objective of a MAC protocol is to allow multiple collection-oriented MAC (DCO-MAC) protocol has been
users to fairly and efficiently share a common medium, and proposed in [19]. In DCO-MAC, the network is divided into
is achieved if and only if communication signal reception two subnets according to the traffic load. The contention-
is adequate [2]. Reliable data transfer techniques or pro- based MAC protocol is used in subnets with lighter traffic
tocols attempt to guarantee that the destination node can loads, while the contention-free MAC protocol is used in
successfully receive what the sending node has sent to it, subnets with heavier traffic loads. Although the DCO-MAC
which is the basis of many underwater applications related to protocol is characterized by low end-to-end packet delay
tactical surveillance, coastline defense, off-shore production, and small energy overheads, interference between different
ecological monitoring and scientific exploration as well as subnets and subnet nodes may affect data transmission, which
disaster prevention, etc. [10]. Over the past few decades, in turn reduces throughput.
a variety of MAC protocols have been proposed for ter- Existing network protocols cannot adapt well to dynam-
restrial wireless sensor networks. Because long propaga- ically changing underwater network environments, where
tion delays, high energy consumption and dynamic network underwater network nodes are constantly moving, and ulti-
topologies are characteristic of UWSNs, the existing proto- mately causing instability to the network topology [20].
cols used in terrestrial sensor networks cannot be directly Therefore, the number of competitors, that is, the number
applied to underwater communication. Moreover, due to the of nodes that need to send data often changes, which leads
peculiar features of underwater acoustic channels, such as to a continuously changing network load. Existing protocols
slow signal propagation speeds (about 1.5 km/s in seawa- neither can adapt to varying network loads, nor can they
ter), very small channel capacity, low channel quality and switch the access mode at any time according to load changes.
highly varying channel quality dynamics, MAC protocol Therefore, this paper proposes a load-based slot allocation
design for underwater acoustic networks faces many novel (LBTSA) MAC protocol, which combines the merits of time
challenges [11]–[13]. division multiple access (TDMA) and carrier sense multi-
As stated, the propagation delay in underwater acoustic ple access with collision avoidance (CSMA/CA), and can
communication environments is high and the communication dynamically adapt to network load changes. Considering that
distance may be long, so the sender cannot accurately judge time slot allocation is an important factor affecting system
the dynamic network load; meanwhile, due to the relatively performance, a load-based time slot allocation scheme is used
low bandwidth and high bit error rate, the transmitted packet in this paper to improve channel utilization.
length in the UWSNs is intentionally limited. Therefore, The rest is organized as follows. In the second section,
the key point to solve in these situations is how the receiver the network model and frame structure are presented; in the
allocates time slots according to dynamic load in order to third section, the design method of a hybrid MAC protocol
reduce network conflicts and improve throughput. based on load slot allocation for underwater wireless sensor
The performances of contention-based and contention-free networks is given; the fourth section introduces the underwa-
MAC protocols are poor when the traffic load in the network ter acoustic channel propagation loss model and the setting
changes dynamically. In order to obtain good performance of simulation parameters used in the simulation; the fifth
in these cases, several hybrid MAC protocols suitable for section discusses the simulation results; and the sixth section
UWSNs have been proposed. In [14], a hybrid MAC protocol summarizes the full text.
Preamble-MAC (P-MAC) is introduced, which consists of a
contention-free protocol and a slot multiple access collision II. NETWORK MODEL AND FRAME STRUCTURE
avoidance (MACA) protocol. Although P-MAC overcomes The network of interest can be considered a star cen-
the shortcoming of low accuracy of time synchroniza- tralized network consisting of N + 1 nodes (i.e., a host
tion, it cannot dynamically capture real-time changes node and N slave nodes), where the N slave nodes
in the network environment. In some efforts [15]–[17], are randomly distributed. The network model is shown
a distributed code division multiple access (CDMA)-based in in Fig. 1.
energy-efficient protocol with ALOHA is given. The pro- The slave node mainly collects monitoring data (such as
tocol uses ALOHA to send signaling and CDMA to send temperature, salinity and depth), images. A slave node may be
data packets. When the network load is heavy, the ALOHA fixed or moving. The host node collects data from slave nodes
protocol generates a large number of collisions, resulting in and stores them. All packets must be sent at the beginning of
unsuccessfully transmitted signals and reduced throughput. the slot, the slot duration is the packet transmission time plus
In [18], a MAC protocol based on the prioritized beacon the maximum propagation delay. Further, the link is treated as
repetition and competition window selection (p-BORE) has one-way and is established between the host node and other
been presented to solve the beacon transmission failure issue. slave nodes.

VOLUME 7, 2019 104543

 Z. Zhang et al.: Load-Based Hybrid MAC Protocol for UWSNs

FIGURE 2. Hybrid MAC protocol frame structure.

CSMA/CA protocol; when the load is heavy, all slave nodes

first use the CSMA/CA protocol. In CSMA/CA, slave nodes
competing for a channel causes a collision. Due to the colli-
FIGURE 1. Network model diagram.
sion caused by the CSMA/CA protocol, when the number of
backoff stages is greater than the optimal number of backoff
stages, TDMA protocol is chosen by the host node to realize
At any point in time, it is entirely expected that there the self-adaptation to the network load The key goal of the
may be multiple nodes that need to send data simultaneously protocol is the method by which network load size is eval-
within a network. The maximum transmission range of all uated so as to efficiently and dynamically switch the access
nodes is considered the same, and the propagation speed mode, that is to say, the determination of the critical value is
of underwater sound is fixed. In underwater wireless sensor the critical point to be solved. The proposed protocol focuses
networks, data communication is mainly divided into three on how to determine the optimal number of backoff stages.
stages. In the first stage, the slave nodes have data to transmit, Since the larger number of backoff stages will cause larger
and the request to send (RTS) is sent for channel reservation. delay, the network throughput does not increase indefinitely
If a collision happens, the slave node randomly backs off for a as the number of backoff stages increases, so we need to seek
period of time before attempting to compete again. In the sec- an optimal backoff stages to ensure getting better throughput
ond stage, the host node judges the number of competitors at lower latency.
according to the number of RTS in order to determine the size The chosen time slot allocation scheme and employed
of the network load. Depending on the network load, the slot access mode together determine the overall performance of
allocation scheme is determined, and then the optimal access the system. The existing MAC protocol cannot adapt to the
control protocol is selected according to the relative priority changing load of the network environment, so LBTSA is
of the slave nodes and the optimal number of backoff stages; proposed to solve the above problems.
finally, the host node broadcasts the information to all slave
nodes. In the third stage, the slave nodes transmit the packets III. LBTSA MAC PROTOCOL DESIGN
according to the information broadcast by the host node, A. PROTOCOL IDEA
and the host node receives the packets and responds with an The proposed LBTSA MAC protocol is a star centralized net-
acknowledgement (ACK) to confirm the correct reception of work protocol. Firstly, the slot allocation is selected according
the transmitted data. to the load of the current network, and then the optimal access
The frame structure of the hybrid MAC protocol is shown mode is chosen according to the relative priority of the slave
in Fig. 2. The proposed protocol schema consists of alter- nodes and the optimal number of backoff stages. The main
nating active and dormant periods, with each active period ideas of LBTSA are shown in Algorithm 1.
including a beacon period, a competition cycle and a trans- According to the characteristics of nodes moving by the
mission cycle. The host node dynamically adjusts the ratio influence of ocean currents, a low-energy energy efficiency
between competition cycles and transmission cycles based on distributed time synchronization (E2 DTS) algorithm was pro-
the network load to achieve load adaptation. posed in the literature [21] using moving nodes as time ref-
In Fig. 2, the yellow blocks represent the beacon period, erence nodes. LBTSA uses the E2 DTS time synchronization
the red blocks represent the competition cycle, the white scheme. A single-transmission interaction mode is adopted
blocks represent the transmission cycle, and the blue blocks by E2 DTS. The moving node first broadcasts a plurality of
represent the dormant period. During the beacon period, time beacon packets to the slave nodes to be synchronized.
the host node evaluates the network load by counting the The slave nodes to be synchronized use the reception time
number of competitors. The host node determines the pri- difference between consecutive beacon packets and the abso-
ority of slave nodes based on the data type. The measure- lute time difference in the beacon packet for linear fitting,
ment information such as temperature, salinity, and pressure thereby realizing synchronization of nodes in the network.
information have high priority, and the image data have If a slave node has data to send, the RTS is sent for channel
low priority. Then, the host node broadcasts the payload reservation. If there is a collision with another slave node,
size, the access mode, the priority of slave nodes, and the the slave node performs a random backoff for a period of
slot allocation scheme. When the load is light, the slave time and then attempts to compete again. The host node
nodes with high priority use TDMA protocol, the others use counts the number of competitors according to the number

104544 VOLUME 7, 2019

Z. Zhang et al.: Load-Based Hybrid MAC Protocol for UWSNs

Algorithm 1 LBTSA Pseudocode TDMA to transmit data packets, the slave nodes with low
1: the slave nodes listen channel priority use CSMA/CA.
2: while the channel is busy do Under CSMA/CA, if a channel is idle and remains dis-
3: backoff; tributed inter-frame spacing (DIFS) time, the slave node
4: end while sends data immediately; otherwise, when the slave node starts
5: send RTS; listening for the channel state "busy" indicator from the host
6: the host node counts the number of RTS; node, the slave node continues to listen until the channel is
7: if the number of RTS > the threshold (N ) then idle at which point it holds the DIFS time for backoff in
8: the host node brosdcasts into continuous allocation of order to avoid conflicts with other slave nodes. If the channel
time slots, CSMA/CA; is busy, the value of the network allocation vector (NAV)
9: while NAV = 0 do backoff timer is frozen, and the current value is recorded.
10: the slave nodes listen channel; The channel then becomes idle, and after the DIFS time,
11: while the channel time 6 DIFS do the NAV backoff timer continues to be started from the last
12: backoff; record. Until the value of the timer reaches 0, another frame
13: end while begins to be sent. If there is no collision and the packet is not
14: transmit; successfully transmitted, the slave node waits for SIFS (short
15: while a collision happens do inter-frame space) and then attempts to resend. If a collision
16: send collision signal, collision number +1; occurs, the number of backoff stages is increased until it is
17: if N = mopt then equal to the optimal number of backoff stages, and the access
18: the host node broadcasts access mode switches mode switches to TDMA.
into TDMA;
19: else B. DETERMINATION OF THE SWITCHING KEY VALUE
20: the slave nodes backoff then transmit; OF PROTOCOL
21: end if As the number of competitors is increasing, the probability of
22: if transmission is completed then collision increases. The throughput of the CSMA/CA proto-
23: send successfully; col decreases with the increase of the number of competitors.
24: else In contrast, the throughput of the TDMA protocol increases
25: the slave nodes wait for SIFS time then retrans- with the number of competitors. Therefore, as the number
mit; of competitors is increasing, when the throughput of the
26: end if CSMA/CA protocol is equal to that of the TDMA protocol,
27: end while the number of competitors N is the switching key value of the
28: end while protocol.
29: else The network throughput refers to the ratio of all received
30: the host node brosdcasts into uniform allocation of packet bytes to the transmission time during the simulation,
tiom slots; as shown in (1):
31: if the priority of the node is high then
32: the host node broadcasts into TDMA; received packet bytes
Throughput = (1)
33: else transmission time
34: the host node broadcasts into CSMA/CA; We define N as the key protocol switching value. The
35: end if expected throughput of CSMA/CA is calculated and com-
36: end if pared with the throughput of TDMA. When they are equal,
the number of competitors is set as the switching key value N
of the protocol, as shown in Fig. 3.
Where the throughput of CSMA/CA is equal to that of
of RTS received, and then determines the size of the network TDMA, i.e., where the two curves intersect, the intersection
load. According to that network load, the time slot allocation point indicates the key protocol switching value, N . As can
scheme is determined. When the number of competitors is be seen from Fig. 3, the abscissa of the intersection of TDMA
greater than N and the time slots are continuously distributed, and CSMA/CA is N ,When the number of competitors is less
CSMA/CA is used first. When a collision happens, the total than N , the throughput of CSMA/CA is significantly higher
number of experienced backoff stages is increased by one. than that of TDMA; on the other hand, when the number of
If the number of experienced backoff stages of a slave node competitors is greater than N , the throughput of TDMA is
is equal to the maximum allowed number of backoff stages significantly higher. Therefore, this paper judges the network
and the number of slave nodes is greater than 1, the host load according to the switching key value of the protocol, and
node broadcasts access mode switches to TDMA. When the then switches the access mode.
number of competitors is less than N and the time slots are According to [22], with an increase in network load,
uniformly distributed, the slave nodes with high priority use the number of delayed packets in TDMA is more than that

VOLUME 7, 2019 104545

 Z. Zhang et al.: Load-Based Hybrid MAC Protocol for UWSNs

hand, the host node only sends control messages and receives
data packets, which will not lead to a large amount of energy
consumption.

C. DETERMINATION OF THE OPTIMAL NUMBER OF

BACKOFF STAGE OF PROTOCOL
Suppose there are N nodes in the network. If there is one node
seeking to send packets, the probability that at least one of the
remaining N − 1 nodes will also try to send packets is p, That
means, in each attempt to transmit, regardless of the number
of retransmissions, each packet conflicts with a constant and
independent probability p, then p can be written as:
p = 1 − (1 − q)N −1 (2)
FIGURE 3. A visualization of the key protocol switching value.
where, q represents the probability that each remaining node
will send a packet if it is at steady state. Then q can be written
as:
in CSMA/CA. Therefore, when the number of competitors 2
exceeds N , that is, when the network load is large, the delay q= (3)
1 + Wmin + pWmin m−1 k=0 (2p)
k
P
of TDMA is greater than the delay of CSMA/CA. In the
proposed scheme, when the network load is large, the slave where, m is the number of backoff stages and Wmin is the
nodes first use the CSMA/CA. If there is a collision, the slave minimum collision window. By combining (2) and (3) into
nodes randomly backoff, and the number of experienced a nonlinear system of equations, the solutions for p and q can
backoff stages is incremented by one. When the total number be obtained.
of experienced backoff stages of any node in the network is Throughput τ refers to the ratio of all received packet bytes
equal to the maximum number of backoff stages, the access to the transmission time during the simulation , as shown
mode switches to TDMA. This approach not only helps in (1).
keep the delay low, but it also serves to maximize network Let Ptr be the probability that at least one node in the
throughput. network will send a packet in a σ slot, as shown in (4) [24]:
In underwater acoustic communication networks, two
Nq(1 − q)N −1
basic working modes of CSMA/CA, namely, basic working Ptr = (4)
mode and optional RTS-CTS-DATA-ACK working mode, are 1 − (1 − q)N
simulated and compared in [23].The results show that the Ps represents the probability that one time on the channel
increase of handshake frame not only affects the end-to-end is a successful transmission, as shown in (5) [24]:
delay of the network, but the throughput of the network is
Nq(1 − q)N −1
also significantly reduced. Therefore, the CSMA/CA under Ps = (5)
the basic mode of operation is more suitable for underwa- 1 − (1 − q)N
ter acoustic network, which is the version of CSMA/CA L is the payload size of the average packet, and Ts is the
used in this paper. The number and probability of collisions average time required to transmit a packet of size L (including
under CSMA/CA will increase as the number of competitors the frame interval); Tid is the duration of the idle period
increases. As a result, the number of experienced backoff (single slot time); Tc is the average time spent in a collision.
stages increases, and the throughput of the network con- By means of the unique solutions p and q, the throughput
versely decreases. Therefore, in order to attempt to maintain τ can be calculated from (1), (4), (5) as follows (6), as shown
the overall throughput of the network, the maximum number at the bottom of the next page.
of allowed backoff stages cannot be increased indefinitely. When the number of competitors is the same, the through-
mopt is defined as the optimal number of allowed back- put under different numbers of backoff stages is compared.
off stages after a collision happens. The next focus is to Due to the increase of the number of backoff stages, the delay
determine that optimal number of allowed backoff stages, becomes longer and longer, so when the increase of through-
mopt , to maximize throughput. LBTSA combines TDMA and put is very small, the minimum number of backoff stages is
CSMA/CA. TDMA does not require control messages and selected as the optimal number of backoff stages, mopt .
transmits packets directly. What used is the basic working When the number of competitors is greater than N , the pro-
mode is CSMA/CA without RTS/CTS. Therefore, LBTSA tocol first selects CSMA/CA. As collisions occur due to
does not need to transmit a large number of control messages, CSMA/CA, with each collision the number of experienced
and just shakes hands at the beginning, which can not only backoff stages is increased by one, and when the number of
save energy, but also reduce end-to-end delay. And most of experienced backoff stages exceeds mopt , the access mode
the control messages are sent by the host node. On the other automatically switches to TDMA.

104546 VOLUME 7, 2019

Z. Zhang et al.: Load-Based Hybrid MAC Protocol for UWSNs

TABLE 1. Channel parameter settings.

spherically and the attenuation increases. The transmission

FIGURE 4. TDMA time slot allocation method.
loss in the spherical model is shown in equation (7) [26].
D. A LOAD-BASED TIME SLOT ALLOCATION SCHEME TL = 10 lg r + αr × 10−3 + A (7)
Since the propagation delay of UWSNs is high, the time slot
allocation is also an important factor affecting the perfor- where, TL represents transmission loss in dB. α represents the
mance of the network, and the location allocation of the time absorption coefficient of the medium, the unit is dB/m. A is a
slot in the super frame can be performed in various ways, representation of the impact of abnormal transmission, with a
such as continuous allocation, uniform allocation, random range of values between 5 − 10dB. γ represents the transmis-
allocation, and the like. This paper will discuss the two time sion distance, and the unit is m. The absorption coefficient of
slot allocation methods shown in Fig. 4. the medium α is a function of the carrier frequency f , the unit
In Fig. 4, the TDMA slot is also called the contention-free of f is kHz, Therefore, according to the empirical equation,
slot, and the CSMA/CA slot is called the contention slot. it can be shown that:
A uniform allocation of time slots is shown in Fig. 4 (a), 
0.1f 2 40f 2

which means two time slots are cross-distributed and scat- α = 1.094 × + + 2.75 × 10 −4 2
f
1+f2 4100 + f 2
tered uniformly within each frame. Fig. 4 (b) illustrates a con-
(8)
tinuous allocation of time slots, which refers to the centralized
arrangement of TDMA time slots, so that the CSMA/CA
B. CHANNEL PARAMETER SETTING
slots in each frame can be combined into a large continuous
contention interval [25] . The research on MAC protocol of UWSNs is inseparable
According to [25], the uniform allocation of time slots can from experimental verification, however, due to the high
ensure low delay and delay jitter, while the continuous allo- cost of underwater experiments, simulation experiments have
cation mode can achieve higher throughput for CSMA/CA. become an important means to verify the performance of
Therefore, when the load is large, the continuous allocation is UWSNs MAC protocol. At present, mainstream line sen-
selected to obtain higher throughput; when the load is small, sor network simulation software includes NS2, OPNET,
the slave nodes with high priority use TDMA. Due to the large and OMNET++, etc., of which NS2 and OMNET++ are
delay caused by TDMA, uniform allocation is used to ensure open source software. The OMNET++ is used for simu-
lower delay and delay jitter. When the load is small, the time lation in this paper. The results of the simulation and the
slots are uniformly allocated; when the load is large, the time results of some formulas are calculated and represented
slots are continuously allocated. by the software Origin. In order to simulate and verify
the performance of proposed protocol, the simulation is as
IV. LBTSA MAC PROTOCOL SIMULATION ENVIRONMENT follows.
SETTINGS 16 nodes are randomly deployed, and the nodes ran-
A. PROPAGATION LOSS MODEL OF UNDERWATER domly generate data packets according to a Poisson
ACOUSTIC CHANNEL distribution model. The channel parameters are shown
In deep water environments, signals propagate vertically, in Table 1. The experimental simulation time is set to
so when the distance increases, the signal is transmitted 200 frames.

Ps Ptr L
τ =
Ps Ptr Ts + Ptr (1 − Ps ) Tc + (1 − Ptr ) Tid
NLq(1 − q)N −1
= (6)
Nq(1 − q)N −1 · Ts + 1 − (1 − q)N − Nq(1 − q)N −1 · Tc + (1 − p)N · Tid
 

VOLUME 7, 2019 104547

 Z. Zhang et al.: Load-Based Hybrid MAC Protocol for UWSNs

TABLE 2. Parameter setting for frame structure.

C. FRAME STRUCTURE PARAMETER SETTINGS

It is assumed that all packets have the same size and a packet
can be sent within a single time slot. The time interval obeys
Exponential Distribution. The length of the control packet is
less than the length of the data packet, and each packet length FIGURE 5. Relationship between the number of competitors and
throughput.
is 500 bits. The simulation parameter settings are shown
in Table 2.
In CSMA/CA, when the channel is idle, the slave nodes the switching key value of the protocol to determine the
will wait for some time. The length of waiting time is defined network load, and then selects the corresponding time slot
by DIFS. After a period of time, whose length is defined allocation mode based on the network load. Finally, the opti-
by short inter-frame space (SIFS), the node will start the mal access mode is selected according to the priority of the
transmission of control packets, such as RTS, CTS, and ACK. slave nodes and the optimal number of backoff stages, and
Once transmission begins, the channel is busy. the access mode is switched.
In the underwater acoustic channel environment described
D. MAC PROTOCOLS USED IN THE SIMULATION in section A, the parameter values are set according to Table 1
Five MAC protocols are compared in the simulation, which and Table 2 , and the key protocol switching value, N , can be
are described as follows: obtained by simulation, as shown in Fig. 5.
a) TDMA: slot-based contention-free MAC protocol, each Normalize throughput. Fig. 5 shows that the abscissa of
node is assigned an independent time slot, and all nodes can the intersection of TDMA and CSMA/CA is N , i.e., N = 10.
only transmit data in their own slot. When the number of competitors is less than 10, the through-
b) CSMA/CA: slot-based contention MAC protocol, each put of CSMA/CA is higher than that of TDMA; when the
node sends a RTS to reserve a channel. If the channel is idle, number of competitors is greater than 10, the throughput
a CTS is sent, and the sender sends the data if it receives the of TDMA is higher than that of CSMA/CA. Therefore, this
CTS. If the receiver correctly receives the frame, an ACK is paper judges the network load according to the switching key
sent to the sender after a period of time. If the sender receives value N of the protocol, and then switches the access mode.
the ACK, the transmitted data is acknowledged to have been When the number of competitors is less than or equal to 10,
received without error; otherwise, it randomly backs off for a the load is smaller; when the number of competitors is greater
period of time, and then re-attempts communication. than 10, the load is larger. When the load is small, a lower
c) HCR [27] protocol: based on the time slot design, HCR delay and lower delay jitter can be guaranteed; when the load
combines TDMA, S-Aloha and random access protocol, and is large, the network throughput can be maximized.
operates protocol switching according to the number of com- In the underwater acoustic channel environment described
petitors. In HCR, the length of the scheduling phase and the in section B, the parameter values refer to Table 1 and Table 2,
random competition phase are variable. This random access the throughput of different competitors can be obtained by
protocol is a relatively simple and random scheme. When a simulation. Fig. 6 shows the throughput in the case of differ-
node is to transmit a data packet, it will transmit the data ent number of competitors, with each curve corresponding to
packet with a constant probability in a time slot. a different value for the number of backoff stages, i.e., m.
d) H-MAC (hybrid MAC) [28] protocol: the first hybrid As can be seen from Fig. 6, for a fixed number of competi-
MAC protocol for underwater wireless sensor networks. tors, the throughput increases as the number of allowed back-
Based on slot design, this protocol combines the TDMA off stages increases. However, when the number of allowed
protocol with a random access protocol, but the length of the backoff stages is greater than or equal to the optimal number
protocol is fixed in both the planned transmission phase and of backoff stages, mopt , the growth of throughput is very
the random competition phase. small, and with an increase in the number of allowed backoff
e) LBTSA: a load-based time slot allocation hybrid wire- stages, the delay becomes longer and longer. Therefore, when
less sensor network hybrid MAC protocol merges TDMA and the number of competitors is greater than N , the optimal
CSMA/CA, and can dynamically adapt to changes in network number of backoff stages mopt is the maximum number of
load. The protocol compares the number of competitors with backoff stages, and when the number of backoff stages is

104548 VOLUME 7, 2019

Z. Zhang et al.: Load-Based Hybrid MAC Protocol for UWSNs

FIGURE 6. Relationship between the number of competitors, throughput FIGURE 8. Relationship between the number of competitors, total
and the number of backoff stages. throughput and time.

competitors is 4, the throughput of LBTSA is 11% higher

than that of HCR protocol,and 48% higher than that of the
H-MAC protocol. When the number of competitors is 10,
the throughput of LBTSA is 38% higher than that of HCR
protocol, and 70% higher than that of H-MAC protocol. At
a competitor count of 16, the throughput of LBTSA is 21%
than that of the HCR protocol, and 115% higher than that of
the H-MAC protocol. The throughput of the LBTSA protocol
is larger than that of H-MAC and HCR for all times and
it demonstrates good adaptability to network load changes.
No matter how the network load changes, the host node
can choose the optimal access mode and the best time slot
allocation. When the load is light, uniform time slot allocation
FIGURE 7. Relationship between the number of competitors and ensures lower delay and delay jitter; when the load is heavy,
throughput.
continuous time slot allocation allows for higher throughput.
Fig. 8 shows the total throughput under different number of
equal to mopt , the access mode is switched to TDMA. How- competitors over a period of time. During the simulation time,
ever, when the number of allowed backoff stages is greater the number of competitors changes via a time-dependent
than or equal to 4, the marginal increase in throughput is Poisson distribution. As can be seen from the figure, the total
relatively small, and with the increase of the number of throughput of LBTSA is 10% higher than that of TDMA
backoff stages, the throughput becomes larger and larger. and 39% higher than that of CSMA/CA. An increase in
Therefore, when the number of competitors is greater than 10, the number of competitors leads to more network conflicts,
the optimal number of allowed backoff stages, mopt , is 4 under so as it has been seen already, the total network throughput
the conditions simulated. When the number of competitors under CSMA/CA decreases. However, when the number of
is greater than 10 and the slot is allocated continuously, competitors is large, TDMA can guarantee contention-free
CSMA/CA is selected first, however, if the number of backoff transmission, so the total throughput increases. Therefore,
stages is more than 4, the access mode switches to TDMA. LBSTA merges the advantages of TDMA and CSMA/CA
We compare LBTSA with TDMA, CSMA/CA, HCR, and adopts slot allocation, which is more suitable for the
and H-MAC and evaluate protocol performance from three changing load of the network environment. In Fig. 7, although
aspects, i.e., throughput, total throughput, and end-to-end the throughput of LBTSA is lower than that of CSMA/CA
delay. The total throughput represents the throughput overlay when there are fewer competitors, the throughput of LBTSA
for the specified simulation time, and the throughput repre- is lower than that of TDMA when there are more competitors.
sents the instantaneous throughput. The HCR and H-MAC However, from Fig. 8, the total throughput of LBTSA is
protocols use the same key switching value for simulation and higher than that of TDMA and CSMA/CA over a period of
comparison. time when the number of competitors is changing. Under
conditions where network load and the number of competitors
V. SIMULATION RESULTS are dynamically changing as a function of time, a fixed, single
Fig. 7 shows the normalized throughput as a function of the MAC protocol cannot adapt in order to maintain total network
number of competitors. From Fig. 7, when the number of the throughput. LBTSA, however overcomes this limitation.

VOLUME 7, 2019 104549

 Z. Zhang et al.: Load-Based Hybrid MAC Protocol for UWSNs

TABLE 3. Comparison of MAC protocols (a network with a small number of competitors).

TABLE 4. Comparison of MAC protocols (a network with a large number of competitors).

delay, overheads, and complexity of all protocols are provided

for qualitative comparison in Table 3 and Table 4 for the cases
where the number of competitors is less than 10 and where it
is greater than or equal to 10.
Although the performance of H-MAC is better when
the load is light, H-MAC exhibits higher delays and lower
throughput in dense-and-heavy-load network conditions.
HCR combines the advantages of TDMA and random con-
tention protocols and has enough adaptability to changing
network loads. But the transmission delay and the throughput
of HCR are between these protocols. Furthermore, as with
other fixed protocols, HCR cannot dynamically adjust the
communication protocol with a changing number of com-
petitors, and the random access protocol is likely to ulti-
FIGURE 9. Relationship between the number of competitors and delay.
mately conflict. Since H-MAC and HCR join random access,
the overhead of control messages increases. With TDMA
Fig. 9 shows the average end-to-end delay as a function of added to LBTSA, the transmission conflict between data
different number of competitors within a system. From the packets is reduced to the greatest possible extent, the through-
figure, it is seen that at a number of competitors of 4, the end- put is improved, and the overhead due to control information
to-end delay in LBTSA is 50% lower than that of the HCR is reduced, and because of the addition of the load-based
protocol, and the end-to-end delay of LBTSA is 75% lower slot allocation scheme, delay is reduced. Therefore, LBTSA
than that of the H-MAC protocol. When the number of com- performs well in terms of throughput and delay, at the expense
petitors is 10, the end-to-end delay of LBTSA is 84% lower of added complexity due to the introduction of CSMA/CA.
than that of the HCR protocol, and the end-to-end delay of Although the complexity and overhead of LTBSA are higher
LBTSA is 80% lower than that of the H-MAC protocol. When than that of TDMA, LTBSA has lower end-to-end delay than
the number of the competitor is 16, the end-to-end delay of that of TDMA and can better adapt to dynamic network loads.
LBTSA is 60% lower than that of the HCR protocol, and LBTSA adds CSMA/CA without RTS/CTS, therefore, it need
the end-to-end delay of LBTSA is 87% lower than that of not transmit too much control messages. LBTSA combines
the H-MAC protocol. When the number of competitors is TDMA and CSMA/CA. Compared with CSMA/CA, both
small, the end-to-end delays under LBTSA and HCR are the complexity and the overhead of LBTSA are low. Since
also small. With an increase in the number of competitors, LBTSA does not require a lot of overhead, it just shakes hands
the end-to-end delays under CSMA/CA and H-MAC increase at the beginning, which can not only save energy, but also
rapidly. The HCR delay curve shows an upward trend, and the reduce end-to-end delay.
increase in delay of LBTSA is much smaller than that of HCR
and H-MAC. The reason is that LBTSA adds a load-based
slot allocation scheme. With the increase in the number of VI. CONCLUSION
competitors, the uniform distribution of time slots ensures a LBTSA, a load-based time slot allocation hybrid MAC proto-
lower delay and lower delay jitter. col for use in UWSNs, is proposed in this paper. The protocol
The throughput and end-to-end delay of different protocols takes advantages of the features of TDMA and CSMA/CA,
are compared in the simulation. The throughput, end-to-end and dynamically adapts according to network load.

104550 VOLUME 7, 2019

Z. Zhang et al.: Load-Based Hybrid MAC Protocol for UWSNs

The crucial part of the protocol is how to fuse TDMA and [10] S. Jiang, ‘‘On reliable data transfer in underwater acoustic networks:
CSMA/CA, that is, how to determine the switching threshold A survey from networking perspective,’’ IEEE Commun. Surveys Tuts.,
vol. 20, no. 2, pp. 1036–1055, 2nd Quart., 2018.
as a function of the number of competitors. The point at which [11] J. Jiang, G. Han, L. Shu, S. Chan, and K. Wang, ‘‘A trust model based on
the throughputs of the CSMA/CA and TDMA protocols cloud theory in underwater acoustic sensor networks,’’ IEEE Trans. Ind.
are equal is used as the indication that it is time to switch Informat., vol. 13, no. 1, pp. 342–350, Feb. 2017.
[12] I. F. Akyildiz, D. Pompili, and T. Melodia, ‘‘Challenges for efficient
the access mode. The proposed protocol determines the key communication in underwater acoustic sensor networks,’’ ACM SIGBED
switching value, N , through the pre-calculated throughput. Rev., vol. 1, no. 2, pp. 3–8, Jul. 2004.
Since CSMA/CA results in collisions, as a function between [13] J. Partan, J. Kurose, and B. N. Levine, ‘‘A survey of practical issues in
underwater networks,’’ ACM SIGMOBILE Mobile Comput. Commun. Rev.,
the throughput and the number of backoff stages, the opti- Los Angeles, CA, USA, vol. 11, no. 4, pp. 23–33, Oct. 2006.
mal number of backoff stages, mopt , is found to maximize [14] D. Pompili and I. F. Akyildiz, ‘‘Overview of networking protocols for
the throughput. Considering that time slot allocation is an underwater wireless communications,’’ IEEE Commun. Mag., vol. 47,
no. 1, pp. 97–102, Jan. 2009.
important factor affecting system performance, a load-based [15] J.-I. Namgung, S.-Y. Shin, N.-Y. Yun, and S.-H. Park, ‘‘Adaptive GTS
time slot allocation scheme is adopted to improve throughput. allocation scheme based on IEEE 802.15.4 for underwater acoustic sen-
From the simulation results under OMNET++, it can be sor networks,’’ in Proc. IEEE/IFIP 8th Int. Conf. Embedded Ubiquitous
Comput., Hong Kong, China, Dec. 2010, pp. 297–301.
seen that not only is the total throughput of LBTSA higher [16] D. Pompili, T. Melodia, and I. F. Akyildiz, ‘‘A CDMA-based medium
than that of TDMA and CSMA/CA, but the throughput of access control for underwater acoustic sensor networks,’’ IEEE Trans.
LBTSA is also higher than that of H-MAC and HCR under the Wireless Commun., vol. 8, no. 4, pp. 1899–1909, Apr. 2009.
[17] D. Pompili, T. Melodia, and I. F. Akyildiz, ‘‘A distributed CDMA medium
condition of dynamically changing number of competitors. access control for underwater acoustic sensor networks,’’ IEEE Trans.
Moreover, the end-to-end delay increment of LBTSA is much Wireless Commun., vol. 8, no. 4, pp. 1899–1909, Jan. 2007.
smaller than the two protocols, so it has good adaptability to [18] E. Moon, S. Lee, M. A. Yaqub, and D. Kim, ‘‘p-BORE: Prioritized beacon
repetition and contention window selection based MAC protocol in under-
the dynamic network load. The simulation results show that water wireless sensor networks,’’ in Proc. ICUFN, Jul. 2017, pp. 269–271.
LBTSA is a promising MAC protocol, based on contention [19] M. Deng, H. Chen, and L. Xie, ‘‘DCO-MAC: A hybrid MAC protocol for
and contention-free fusion, and its performance is better than data collection in underwater acoustic sensor networks,’’ Sensors, vol. 18,
no. 7, p. 2300, Jul. 2018.
existing single and hybrid MAC protocols. In large UWSNs, [20] G. Han, S. Shen, H. Song, T. Yang, and W. Zhang, ‘‘A stratification-based
although TDMA can transmit more data packets, the greater data collection scheme in underwater acoustic sensor networks,’’ IEEE
number of collisions and experienced backoff times under Trans. Veh. Technol., vol. 67, no. 11, pp. 10671–10682, Nov. 2018.
[21] Z. Li, Z. Guo, F. Hong, and L. Hong, ‘‘E2 DTS: An energy efficiency
CSMA/CA will occupy more time slots. A fixed number of distributed time synchronization algorithm for underwater acoustic mobile
time slots per frame leads to the reduction of TDMA time sensor networks,’’ Ad Hoc Netw., vol. 11, no. 4, pp. 1372–1380, Jun. 2013.
slots and negatively affects network performance. By using [22] R. K. Wang, M. Jing, and L. F. Kang, ‘‘Simulation analysis of
TDMA/CSMA hybrid protocol based on MATLAB,’’ Fire Command Con-
the LBTSA protocol, we can achieve higher throughput and trol, vol. 42, no. 2, pp. 163–167, 2017.
lower end-to-end delay, so the proposed LBSTA is suitable [23] X. L. Yu, ‘‘Research on MAC layer protocol in underwater acoustic
for small underwater wireless sensor networks with high load. communication network,’’ M.S. thesis, Harbin Eng. Univ., Harbin, China,
2009.
[24] G. Bianchi, ‘‘Performance analysis of the IEEE 802.11 distributed coordi-
REFERENCES nation function,’’ IEEE J. Sel. Areas Commun., vol. 18, no. 3, pp. 535–547,
[1] M. C. Domingo, ‘‘An overview of the Internet of underwater things,’’ Mar. 2000.
J. Netw. Comput. Appl., vol. 35, no. 6, pp. 1879–1890, Nov. 2012. [25] J. Q. Cui, J. F. Yang, and J. Xu, ‘‘CSMA/CA performance analysis in hybrid
[2] S. Jiang, ‘‘State-of-the-art medium access control (MAC) protocols for CSMA/CA and TDMA networks,’’ Comput. Eng. Appl., vol. 3, no. 20,
underwater acoustic networks: A survey based on a MAC reference pp. 83–88, 2017.
model,’’ IEEE Commun. Surveys Tuts., vol. 20, no. 1, pp. 96–131, [26] Y. K. Chen, ‘‘Time delay analysis and media access protocol design of
1st Quart., 2018. under-water acoustic communication network,’’ Ph.D. dissertation, South
[3] L. Liu, S. Zhou, and J.-H. Cui, ‘‘Prospects and problems of wireless com- China Univ. Technol., Guangzhou, China, 2018.
munication for underwater sensor networks,’’ Wireless Commun. Mobile [27] L. Hong, F. Hong, Z. Guo, and Y. Guo, ‘‘HCR: A hybrid MAC protocol
Comput., Underwater Sensor Netw., Archit. Protocols, vol. 8, no. 8, for underwater sensor networks using channel reservation,’’ Int. J. Comput.
pp. 977–994, 2008. Appl., vol. 33, no. 2, pp. 154–159, Jul. 2011.
[4] E. P. Wigner, ‘‘Theory of traveling-wave optical laser,’’ Phys. Rev., [28] K. B. Kredo and P. Mohapatra, ‘‘A hybrid medium access control protocol
vol. 134, pp. A635–A646, Dec. 1965. for underwater wireless networks,’’ in Proc. WUWNET, New York, NY,
[5] H. Luo, K. Wu, R. Ruby, Y. Liang, Z. Guo, and L. M. Ni, ‘‘Software- USA, Sep. 2007, pp. 33–40.
defined architectures and technologies for underwater wireless sensor
networks: A survey,’’ IEEE Commun. Surveys Tuts., vol. 20, no. 4,
pp. 2855–2888, 4th Quart., 2018.
[6] D. N. Sandeep and V. Kumar, ‘‘Review on clustering, coverage and
connectivity in underwater wireless sensor networks: A communication
techniques perspective,’’ IEEE Access, vol. 5, pp. 11176–11199, 2017.
[7] J. Heidemann, M. Stojanovic, and M. Zorzi, ‘‘Underwater sensor networks: ZIWEI ZHANG received the B.S. degree from
Applications, advances and challenges,’’ Philos. Trans. Roy. Soc. London the Qingdao Institute of Technology, Qingdao,
A, Math. Phys. Sci., vol. 370, no. 1958, pp. 158–175, Jan. 2012. China, in 2012. She is currently pursuing the
[8] G. Han, X. Long, C. Zhu, M. Guizani, and W. Zhang, ‘‘A high-availability M.S. degree with the Department of Information
data collection scheme based on multi-AUVs for underwater sensor net- Science and Technology, Qingdao University of
works,’’ IEEE Trans. Mobile Comput., to be published. Science and Technology, Qingdao. Her research
[9] G. Han, X. Long, C. Zhu, M. Guizani, Y. Bi, and W. Zhang, ‘‘An AUV interests include underwater wireless sensor net-
location prediction-based data collection scheme for underwater wire- works, acoustic communication, and underwater
less sensor networks,’’ IEEE Trans. Veh. Technol., vol. 68, no. 6, acoustic sensor networks.
pp. 6037–6049, Jun. 2019.

VOLUME 7, 2019 104551

 Z. Zhang et al.: Load-Based Hybrid MAC Protocol for UWSNs

WEI SHI was born in Shandong, China, in 1986. JINGJING WANG (M’15) received the B.S.
He received the B.S. degree from Ludong Uni- degree in industrial automation from Shandong
versity, Yantai, China, in 2009, and the master’s University, Jinan, China, in 1997, the M.Sc. degree
degree in signal and information processing and in control theory and control engineering from the
the Ph.D. degree in computer application technol- Qingdao University of Science and Technology,
ogy from the Ocean University of China, Qingdao, Qingdao, China, in 2002, and the Ph.D. degree in
China, in 2011 and 2014, respectively. Since 2014, computer application technology from the Ocean
he has been a Lecturer with the School of Informa- University of China, Qingdao, in 2012. From
tion Science and Technology, Qingdao University 2014 to 2015, she was a Visiting Professor with
of Science and Technology. His research interests The University of British Columbia. She is cur-
include 5G, UWB, 60-GHz wireless communication, and underwater wire- rently a Professor with the School of Information Science and Technol-
less sensor networks. ogy, Qingdao University of Science and Technology. Her research interests
include underwater wireless sensor networks, acoustic communications,
QIUNA NIU received the B.S. degree from the ultra-wideband radio systems, and MIMO wireless communications. She
Changchun University of Science and Technology, serves as the Vice Chair for the ACM Qingdao Chapter.
Changchun, China, in 1997, and the M.S. degree
from Shandong University, Jinan, China, in 2004.
From 2017 to 2018, she was a Visiting Scholar
with the University of Arkansas. She is currently
an Associate Professor with the School of Informa-
tion Science and Technology, Qingdao University
of Science and Technology. Her current research
interests include channel encoding technology in
wireless communication, indoor wireless localization, and target tracking. HANJIANG LUO (M’09) received the Ph.D.
degree from the Department of Computer Science,
YING GUO received the B.S. and M.S. degrees Ocean University of China. He was a Visiting
from the Qingdao University of Science and Tech- Scholar with HKUST, in 2014. He is currently
nology, Qingdao, in 2004 and 2007, respectively, an Associate Professor with the School of Com-
and the Ph.D. degree from the Department of Com- puter Science and Engineering, Shandong Univer-
puter Science and Technology, Ocean University sity of Science and Technology. He has published
of China, Qingdao, in 2010. From 2018 to 2019, research papers in some prestigious journals and
she was a Visiting Professor with Arizona State conferences, such as ACM Computing Surveys, the
University. She is currently an Associate Profes- IEEE COMMUNICATIONS SURVEYS AND TUTORIALS, the
sor with the Qingdao University of Science and IEEE TPDS, IEEE TVT, and the IEEE ICDCS. His research interests include
Technology. Her research interests include wire- wireless sensor networks, underwater wireless sensor networks, the Internet
less sensor networks, underwater acoustic networks, and the Internet of of Things, data science, and artificial intelligence.
Things.

104552 VOLUME 7, 2019