ArticlePDF Available

Void-Handling Techniques for Routing Protocols in Underwater Sensor Networks: Survey and Challenges

January 2017
IEEE Communications Surveys & Tutorials PP(99):1-1

January 2017
PP(99):1-1

DOI:10.1109/COMST.2017.2657881

Authors:

Seyed Mohammad Ghoreyshi

University of Southampton

Alireza Shahrabi

Glasgow Caledonian University

Tuleen Boutaleb

Glasgow Caledonian University

From the view of routing protocols in Underwater Sensor Networks (UWSNs), the presence of communication void, where the packet cannot be forwarded further using the greedy mode, is perhaps the most challenging issue. In this paper, we review the state of the art of void-handling techniques proposed by underwater geographic greedy routing protocols. To this, we first review the void problem and its negative impact on the category of the geographic greedy routing protocols, which does not entail any void recovery technique. It is followed by a discussion about the constraints, challenges, and features associated with the design of void-handling techniques in UWSNs. Afterwards, currently available void-handling techniques in UWSNs are classified and investigated. They can be classified into two main categories: location-based and depth-based techniques. The advantages and disadvantages of each technique along with the recent advances are then presented. Finally, we present a qualitative comparison of these techniques and also discuss some possible future directions.

A void area with respect to destinations S 1 and S 2

…

A convex void in a single sink architecture

…

A concave void in a single sink architecture

…

Impact of void problem on DBR protocol [49]

…

+10

The Classification of Void-handling techniques for UWSNs

…

Figures - uploaded by Seyed Mohammad Ghoreyshi

Content may be subject to copyright.

Content uploaded by Seyed Mohammad Ghoreyshi

Content may be subject to copyright.

1553-877X (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2017.2657881, IEEE

Communications Surveys & Tutorials

Void-Handling Techniques for Routing Protocols in

Underwater Sensor Networks: Survey and

Challenges

Seyed Mohammad Ghoreyshi, Alireza Shahrabi, and Tuleen Boutaleb

School of Engineering and Built Environment

Glasgow Caledonian University

Glasgow, UK

{Seyed.MohammadGhoreyshi, A.Shahrabi, T.Boutaleb}@gcu.ac.uk

Abstract—From the view of routing protocols in Underwater

Sensor Networks (UWSNs), the presence of communication void,

where the packet cannot be forwarded further using the greedy

mode, is perhaps the most challenging issue. In this paper, we

review the state of the art of void-handling techniques proposed

by underwater geographic greedy routing protocols. To this, we

ﬁrst review the void problem and its negative impact on the

category of the geographic greedy routing protocols, which does

not entail any void recovery technique. It is followed by a dis-

cussion about the constraints, challenges, and features associated

with the design of void-handling techniques in UWSNs. After-

wards, currently available void-handling techniques in UWSNs

are classiﬁed and investigated. They can be classiﬁed into two

main categories: location-based and depth-based techniques. The

advantages and disadvantages of each technique along with

the recent advances are then presented. Finally, we present a

qualitative comparison of these techniques and also discuss some

possible future directions.

I. INT ROD UC TI ON

Underwater acoustic sensor networks have obtained a con-

siderable attention to support aquatic applications such as

exploration of ocean resource, disaster prevention, intrusion

detection, military applications, and pollution monitoring [1]–

[4]. The sensors are distributed in different depths to collect

information and forward them to a destination, which may be

a sink, a group of sinks or an Autonomous Underwater Vehicle

(AUV) [5]–[8].

Different routing protocols are proposed to improve the

packet delivery with minimum energy and delay cost in

UWSNs, in which greedy routing protocols are the most

prominent approaches due to the simplicity of use in UWSNs

[9]–[11]. With the aid of localization mechanisms, geographic

greedy routing has become a promising scheme in the sensor

and ad hoc networks. Geographic greedy routing (also called

position-based, or location-based) is a routing principle, which

relies on the geographic position information to forward the

data packets closer to the destination in each hop [12]–[14].

In contrast to a table-driven (proactive) routing, which

requires large communication overhead to establish end-to-end

routes [15]–[17], geographic routing does not need to discover

and maintain the full path from the source to the destination.

In most cases, the geographic information of one or two hops

has been held to route the packets. Thus, there is no need for

the routing tables and routing messages to hold and update the

route path. This unique feature makes them scalable to be used

in the large networks with many nodes [9], [18]. Geocasting

is another service which can be supported by the geographic

routing to deliver a packet to an intended geographic region

[19].

Geographic routing follows a greedy-forwarding strategy in

which every node looks for the closest neighbouring node to

the destination. However, greedy forwarding may fail because

of communication voids or local maximum phenomenon [20]–

[22]. In this case, the forwarding node cannot ﬁnd any quali-

ﬁed node with a positive progress towards the destination, so

the packet may be dropped even though there is a valid path

from the sender to the destination.

As depicted in Fig. 1, node eis a void node since it has no

neighbouring node closer to the sinks S1and S2than itself.

Thus, in a greedy-forwarding strategy, the packet is dropped

if node eis selected as the next forwarding node instead of

node d, which has a valid path to a sink. Without resolving

this issue, data packets may drop in the network, wasting the

network resources such as energy and bandwidth. Moreover,

the void problem is more challenging as it is unpredictable as

to when and where a void may occur due to dynamic nature

of operating environment.

A number of factors individually or a combination of them,

cause the void phenomenon, such as the sparse topology,

temporary obstacles, and unreliable nodes or links [20], [23].

Increasing the density of the network is a simple solution;

however, it is not possible all the time and, even so, it cannot

entirely eliminate the void problem [24]. Therefore, in order to

improve the routing efﬁciency, many different techniques and

recovery methods are proposed to handle the void problem in

the wireless and ad hoc networks [21], [25]–[28].

However, due to the different characteristics of UWSNs,

such as the differences in topological and environmental

features, void-handling techniques proposed for terrestrial net-

works are impractical in the underwater environment. This

is attributed to the fact that, ﬁrst, all communications voids

in UWSNs are three-dimensional, which requires different

treatments than two-dimensional holes in the terrestrial net-

works [29]. Second, the mobility of most underwater nodes

makes the void mobile. A mobile void can also result from

Communications Surveys & Tutorials

Fig. 1: A void area with respect to destinations S1and S2

the surrounding environment [30]. This can be the case of

a ship that navigates over an underwater network, it blocks

communications in the nearby area and thus generates a void

that moves along with the ship.

Taking into account the different requirements of the three-

dimensional and mobile voids, designing some efﬁcient void-

handling techniques for the routing protocols speciﬁc to

UWSNs is necessary. The performance of these void-handling

techniques depends on many factors, such as the number of

void nodes, network dynamics, and number of destinations.

Generally, the routing strategies and void-handling tech-

niques can be categorised into two main groups: location-

based and depth-based. In the location-based category, the void

node is determined based on the geographical advancement

of the neighbouring nodes. A node is called a void node,

if it cannot ﬁnd any other node within its neighbourhood

with shorter euclidean distance toward the destination. In the

depth-based category, the void node is determined based on

the depth advancement of the neighbouring nodes towards the

water surface. Depth information indicates the vertical distance

from each node to the water surface. A node is a void node

if it cannot ﬁnd any neighbouring node with the lower depth

than itself. Because of different features in these categories,

different void-handling techniques are required.

In this paper, our goal is to investigate the void-handling

techniques reported in the literature. To achieve this, we ﬁrst

mention the characteristics of UWSNs in Section II. Then, in

Section III, we investigate the void problem in 3D UWSNs and

its ignorance impact on the protocols, which do not support

any void-handling technique. Then, we indicate the challenges

and required features to design and evaluate a void-handling

technique. In Section IV, we propose a classiﬁcation for all

void-handling techniques in UWSNs. In Section V, almost all

currently reported void-handling techniques in the literature

are discussed in detail. The advantages and disadvantages of

each void-handling technique are then shown using different

examples and analysis. These void-handling techniques quali-

tatively are compared in terms of their performance regarding

the void problem in Section VI. In Section VII, we identify

some directions and guidelines for the future research on the

void-handling techniques in UWSNs. Finally, in Section VIII,

we conclude the paper.

II. CH AR ACT ER IS TI CS O F UND ERWATER SE NS OR

NET WORKS

In this section, we present the characteristics of UWSNs.

The underwater sensor networks pose more severe situation to

cope with void regions. Due to the characteristics mentioned

below, the terrestrial void-handling techniques are quite im-

practical and can not be employed directly in the underwater

environment. Thus, it is required to develop the void-handling

techniques suitable for underwater acoustic communications

taking all the characteristics into account.

A. Three-dimensionality and Node Movement

In contrast to the terrestrial networks, UWSNs are three-

dimensional, and sensors move with the water current. The

three-dimensional holes in the routing path can lead to more

packet failures. The effect of hidden terminal problem [31]

in 3D UWSNs is more intense due to the existence of

more neighbouring nodes in different directions. Moreover,

the topology continuously changes with the nodes movement.

The speed of node depends on the water velocity, which varies

at different times [9], [10].

B. High and Variable Propagation Delay

Underwater sensors use the acoustic waves which are pre-

ferred for the long distance communications. The speed of

sound in underwater is about 1500 m/s [32]. Thus, it causes

a large propagation delay, which is about to ﬁve orders of

magnitude higher than that of radio frequency (RF). The

sound velocity varies based on the different parameters such

as temperature, salinity, and depth of water [33]. It is critical

to taken into account the propagation delay in designing the

void-handling techniques in UWSNs.

C. Limited Bandwidth

Due to features of acoustic waves and environmental noise,

the acoustic bandwidth is severely limited in UWSNs. The

available acoustic bandwidth depends on the communication

range and acoustic frequency. As a result of the limited

bandwidth, the data rate for underwater sensors can rarely

exceed 100 kbps [34]. Therefore, the limited bandwidth of

acoustic channels should be taken into consideration when

designing a void-handling technique.

Communications Surveys & Tutorials

Fig. 2: A convex void in a single sink architecture

D. Path Loss

The underwater environment has higher path loss in com-

parison to terrestrial physical layer. The path loss results from

the attenuation and geometric spreading [35]. The attenuation

also results from the absorption of acoustic waves in water

[36]. The path loss effect can be reduced by increasing the

transmission power or decreasing the traversed distance. Thus,

packet forwarding is more likely to be successful if packets

are relayed over multiple short distances instead of traversing

over long distances [31], [35].

E. Noises

There are two kinds of noise which affect the acoustic

communications: man-made noise and ambient noise. Man-

made noise is mainly generated by human activities like using

pumps and shipping. Ambient noise refers to natural events

such as seismic and tides [3], [36]. The main sources of

the noise include turbulence, shipping, waves and thermal

noise [35]. These noises lead to a lossy and noisy underwater

environment, which should be considered carefully in the void-

handling techniques.

F. Energy Consumption

Energy consumption is another primary concern in UWSNs

since it is hard to replace or recharge the sensor batteries.

In UWSNs, the energy consumed by the sensors is much

more than what is consumed by the terrestrial sensors [37],

[38]. Therefore, energy efﬁciency is an essential requirement

of void-handling techniques in UWSNs.

Fig. 3: A concave void in a single sink architecture

III. VOID PROBLEM AND CHALLE NG ES

In this section, we introduce the void problem, the used

terminology, and its challenges.

A. Deﬁnitions and characteristics

An underwater sensor network has a 3D network topology

in which one or more sinks are located on the water surface

equipped with an acoustic modem for underwater communica-

tion and with a radio modem for out of water communication

[5]. Anchored nodes are located at the bottom of the ocean

in the predetermined locations to collect the information and

deliver them to a sink by using the relay nodes which are

located at different levels in between.

In a greedy-forwarding strategy, each forwarding node trans-

fers packets to a node closer than itself to the destination

[9]. Given a sender node miand a destination node S, the

advancement of a neighbouring node mjis deﬁned as

ADV (mj) = D(mj, S)−D(mi, S),(1)

where D(m, S)denotes the Euclidean distance from node m

to destination S. In the location-based routing, destination S

is considered as the closest sink to the node m[8], [39].

Whereas in pressure-based routing, D(m, S)is simpliﬁed to

the Euclidean distance between node mand the water surface,

i.e. the depth of node m. Depending on the class of routing

protocol, one or a few candidate nodes can participate in the

packet forwarding, which are selected from the following set

Cmi={mj∈N(mi) : ADV (mj)>0},(2)

Communications Surveys & Tutorials

where N(mi)includes all the neighbouring nodes within the

transmission range of mi. An empty Cmimeans that node mi

has no qualiﬁed next-hop node to forward the packet in greedy

mode. Such a node is called a void node. In the literature, a

void node is also referred to as local maxima or stuck node,

which are also used interchangeably throughout this paper.

During the greedy mode of packet forwarding in geographic

routing, if a relay node cannot ﬁnd any neighbouring node

with positive advancement toward the sink(s), it should switch

to the recovery mode to bypass the void area; otherwise, the

packet will be dropped [11], [26], [40], [41]. The void nodes

are generally located on the boundary of a void communication

area. In UWSNs, a void communication area is a three-

dimensional region between underwater nodes which is empty

of any nodes inside (like the void areas shown in Figs. 2 and

3). A void area prevents communication between some of the

nodes in the network. The path between the local maxima

node and a non-local maxima node, where greedy routing can

be resumed, is called the recovery path.

The forwarding direction speciﬁes whether a hole is a

communication void or not. In UWSNs, void areas are usually

considered as the holes between the relay nodes and water

surface where the sink(s) is located. For further clariﬁcation,

Fig. 2 shows a case in which there is a void area between node

eand sink aon the surface. If a greedy routing protocol does

not include any void-handling technique, the packet is dropped

by node e, while there are two valid paths from this node eto

the sink (e-d-c-b-aand e-f-g-h-a). Thus, node eis considered

as a void node with respect to the destination awhile the

empty area between them is called a 3D void communication

area.

In a pressure-based model, a void node is deﬁned in another

way. In this category, a node is called a void node if it is

located in a shallower depth than all of its neighbouring nodes

and it is not connected to any sink on the surface [42]. In

this case, a packet cannot make any upward progress toward

the surface. In Fig. 3, node eis a void node, since all of its

neighbouring nodes have higher depth. The trapped nodes are

those that are located down below the void node and involving

them in packet forwarding leads to getting stuck the packet

(e.g. b,c,d). The area in which the trapped nodes are located

called the trap area [43].

Voids emerge in the underwater environment in different

shapes and sizes. For instance, void areas are emerged in

convex and concave shapes in Figs. 2 and 3, respectively.

There might be cases where nodes seem to be connected to

each other in terms of transmission distance, but they cannot

communicate. This is due to the fact that some other factors

such as obstructions and underwater noises can disparage this

assumption [44]. Thus, nodes are connected to each other if the

transferred signal between them can be decoded without any

error. To this, Signal-to-Noise Ratio (SNR) over the traversed

distance should be higher than detection threshold at the

receiver side [35]. Furthermore, holes in water are not neces-

sarily distributed evenly. It can be formed by many factors such

as deployment model, energy depletion and movement pattern

of underwater nodes, etc [45]. Knowing such characteristics

is very useful when designing a void-handling technique.

Fig. 4: Impact of void problem on HH-VBF protocol [47]

Void areas in the terrestrial sensor network are usually ﬁxed

because they consist of a set of static sensor nodes [20].

However, in UWSNs, by the movement of ﬂoating nodes

with the water current, void areas can gradually move to

other regions. Hence, mobility of the void communication area

is another feature which should be taken into consideration

in this environment. Displacement speed of void area is

dependent on the velocity of the underwater nodes which is

not always so high. Nonetheless, high dynamics is not always

a negative factor for the routing protocols, because sometimes

the temporary voids can be vanished with the aid of the newly

arrived nodes [30].

B. Void-ignorance routing protocols

In this section, we brieﬂy discuss the negative impact of void

on some well-known routing protocols which intentionally (for

the sake of simplicity) or ignorantly do not consider it. Some

routing protocols such as Vector-Based Forwarding (VBF),

and Hop-by-Hop Vector-Based Forwarding (HH-VBF) [46],

[47] are location-based greedy routing in which forwarding

nodes are selected within a virtual pipeline faced toward the

destination. However, no solution when facing with a void in

the pipeline is provided.

In VBF, packets are forwarded within a ﬁxed virtual pipeline

between every pair of source (e.g. an anchored node at

the bottom of ocean) and destination. The performance of

VBF is dropped in the sparse networks, where candidate

nodes inside the pipeline can barely be found. In order to

increase the chance of ﬁnding nodes in the pipeline, HH-VBF

requires a different pipeline at each hop originated from every

intermediate (relay) node. However, the void occurrence in the

pipeline toward the destination still remained as a problematic

issue. This problem is shown in Fig. 4. As can be seen, the

packets generated in source Acan successfully be delivered

to the sink; however, packets generated in source Bis stuck

in node vand dropped, though there exists a valid path like

(v-w-x-y-s) to the sink.

As another protocol, we can consider RDBF [48], which

similarly relies on the use of location-based coordinates but

with no void-handling technique. In this protocol, packets

Communications Surveys & Tutorials

Fig. 5: Impact of void problem on DBR protocol [49]

are relayed through the nodes with the nearest geographical

distance to the sink node. RDBF does not limit the forwarding

nodes in a pipe, or other geometric shape; however, in facing

a void area, RDBF does not present a recovery mode to deal

with the packets which are stuck in a local maxima node.

Some routing protocols can also be found in the pressure-

based category, with no particular solution for void problem.

For instance, DBR [49] is the ﬁrst pressure-based routing

protocol which takes the advantage of the depth of each node

to forward packets towards the surface. However, forwarding

node selection is not performed in a way that packets do

not become stuck in a void node, and neither proposes any

recovery method after getting stuck in a local maxima node.

This problem is shown in Fig. 5. When node kreceives a

packet, since it does not have any neighbour with lower depth

or pressure than itself, the packet will be dropped, though

there are some valid paths to one of the sinks (e.g. k-l-m-n-

S1). DBMR and EEDBR [50], [51] are also proposed in this

category which consider the nodes residual energy in their

forwarding set selection; however they still have no solution

for the local maxima node problem.

All in all, the presence of void area in the routing path

can dramatically decrease the performance of the network.

High packet loss and wasting resources are the immediate

consequence of not including an appropriate void-handling

technique in the routing protocol. Specially, in delay sensitive

applications, dropping a packet can lead to missing a critical

event and failure of the network duty. Thus, an efﬁcient

geographic routing protocol should include a void-handling

mode in addition to the greedy-forwarding mode. In the

following section, we discuss the existing challenges to design

an efﬁcient void-handling technique in UWSNs.

C. Void-handling Techniques: Constraints and Challenges

For 3D sensor networks, it is proved that there is no

local and memoryless routing algorithm which can guarantee

the packet delivery [52]. Hence, packet delivery just can be

guaranteed by utilising the updated reachability information,

or using an expensive exhaustive search like ﬂooding. In the

remaining of this section, we investigate some techniques that

might be used to overcome the void issue followed by a

discussion about the constraints and challenges associated with

each technique. The features of the void-handling techniques

are also summarised in Table I.

1) Increasing the nodes density: Although the density of

the network has, to some extent, an impact on the occurrence

of the void area [24], increasing the network density cannot

entirely eradicate this issue as it cannot be predicted when

and where it can happen. In fact, many other factors, such

as deployment pattern of the nodes, node movement, and

unreliability of some links, are involved in the creation of the

void areas. Hence, the unpredictable nature of void occurrence

makes it a complicated task to be located and avoided.

Moreover, underwater nodes are much more expensive than

the almost costless RF sensor nodes. For these reasons, con-

sidering a dense topology to limit the number of void areas is

an unrealistic solution.

2) Flooding techniques: In the terrestrial networks, ﬂood-

ing is one of the simplest ways to deal with the void area.

However, full ﬂooding technique or original ﬂooding is not a

cost efﬁcient method in the underwater environment. The basic

idea behind the full ﬂooding is to give a copy of the packet to

all the nodes in the network to increase the packet delivery

probability [53]. However, when a stuck node ﬂoods the

packets to all its neighbours, it causes the reception of many

duplicated packets by the destination. Furthermore, packets are

not relayed in an optimal path with minimum distance and

energy cost between the stuck node and destination. Three-

dimensionality of the UWSNs also serves to exacerbate the

situation by increasing the number of duplicated packets.

Duplicated packets waste the network resources and deplete

the energy of the nodes (especially nodes close to the sink).

In total, the full ﬂooding wastes the network resources.

Nevertheless, restricted ﬂooding or partial ﬂooding can be cost

efﬁcient to deal with the void areas. In this way, ﬂooding

range and rate are limited to prevent the packets from being

distributed into the whole network. Thus, the void problem can

be resolved with the minimum cost which is the key aspect of

a workable ﬂooding-based void-handling technique [54]–[56].

3) Heuristic techniques: A heuristic technique uses the

experience to ﬁnd a satisfactory solution by employing a

practical method. In order to handle the voids, heuristics tech-

niques also can be used to ﬁnd the recovery paths [54], [57]–

[59]. Heuristic techniques have the ability to be customised

for different network topologies and void areas. This group

of techniques has their own advantages and disadvantages.

Although heuristic techniques cannot guarantee the packet

delivery, they can signiﬁcantly decrease the complexity of a

solution. These techniques do not follow strict rules to achieve

more simplicity and easier applicability in UWSNs. In order

to understand where heuristic techniques are applicable, some

theoretical analyses usually are required. The derived results

from the theoretical analyses can determine the effectiveness

and efﬁciency of each heuristic technique and its various

possible conﬁguration.

4) Transmission power adjustment: Due to special features

of the acoustic signal, the transmission power of each node

can be adjusted in a cross-layer fashion [60], [61]. Thus, with

the hope of ﬁnding non-void nodes (with positive progress)

in the farther distance, a local maxima node can increase its

Communications Surveys & Tutorials

TABLE I: Summary of the void-handling techniques in UWSNs

Void-handling Technique Advantages Disadvantages

Increasing the nodes density Reducing the void occurrence to some extent A dense topology imposes high cost and it

cannot entirely eradicate the void problem

Flooding techniques Involving more nodes to increase the packet

delivery probability

Increasing the number of duplicated packets and

wasting the network resources

Heuristic techniques

Ability to be customised for different network

topologies and decreasing the complexity of a

solution

They cannot guarantee the packet delivery and

not applicable in some conﬁgurations

Transmission power adjustment Finding non-void nodes in the farther distance

by increasing the transmission power

Energy dissipation, and interference in the MAC

layer

Backward forwarding

Initiating an on-demand packet recovery which

is reasonable when the number of void areas is

limited

It may create a loop between the stuck node and

some other nodes which waste the energy

Passive participation

Contributing to the self-healing property of net-

work topology by the passive participation of the

void nodes

An inappropriate approach to handle voids in a

sparse underwater network

Void Avoidance Minimising the possibility of encountering the

void area during the packet forwarding

Exchanging beacons between nodes imposes

communication overhead in the highly dynamic

networks

Learning techniques The packet delivery can be enhanced by learning

from the past results

It is not applicable to the networks with a high

dynamic

Network topology control Reducing the number of void areas by using the

depth adjustment mechanism It consumes high energy for topology adjustment

Backup network Obtaining more information about the void areas

or collecting data directly from sensor nodes High energy consumption and complexity

Hybrid technique Improving the efﬁciency by using a combination

of the void-handling techniques More complexity

forwarding range by increasing the transmission power. It is

even possible to use a very high power to connect the local

maxima node to the destination directly. However, increasing

the transmission power can lead to energy dissipation, and

interference in the MAC layer. Thus, in order to exploit this

technique, increasing the maximum power and transmission

range should be limited at each node. After ﬁnding a node

with positive progress toward the sink, relaying the packet

with a predeﬁned transmission power can be resumed.

5) Backward forwarding: Some void-handling techniques

allow a packet to get stuck and then initiate a recovery

method to guide the packet to a non-void node. As a recovery

technique suggested in [62], if a relaying node with positive

advancement cannot be found, the packet can be forwarded

back to a node with the least negative advancement to deliver

the packet. However, this approach may create a loop between

the stuck node and some other nodes. Void-handling technique

should obviously be loop-free, otherwise it only wastes the

resources.

6) Passive participation: As another solution, on a volun-

tary basis, void nodes can take themselves out of the packet

forwarding to provide the opportunity for other available nodes

[63]. In this way, each node should be able to recognise

whether it is a void node or not. Afterwards, void nodes

simply drop the packets as soon as they could not ﬁnd any

neighbouring node with positive progress toward the sink.

In this way, when a forwarding node does not receive any

acknowledgement from the void node, it selects another node

to relay the packet, as if the void node does not exist in the

network. This technique contributes to the self-healing prop-

erty of network topology. However, in some cases, removing

a void node may lead to the disconnection with the sink node

if the only valid path to the destination is via the void node.

Thus, this technique is an inappropriate approach to handle

voids in the sparse underwater networks.

7) Void Avoidance: Not sending packets to the stuck nodes

is another strategy to cope with the void problem, while other

techniques let it happens and then is handled by the local

Communications Surveys & Tutorials

maxima node. This strategy can minimise the possibility of

encountering the void area during the packet forwarding. There

are different approaches to achieve this objective like VAPR

[43], OVAR [37], and LLSR [64]. For instance, a local maxima

node can inform its neighbouring nodes about its current state

by sending beacons. Thus, neighbouring nodes can set the cost

of void nodes to an inﬁnite value to prevent sending packets

to them. The difference with the passive participation is that

void nodes actively inform other nodes about their current

status. Thus, these techniques can reduce the long delay in

the routing decision making. However, exchanging beacons

between nodes still imposes communication overhead in the

high dynamic networks.

8) Learning techniques: Void-handling techniques can also

be augmented using some learning methods [65]. As nodes

are mobile, very old information may not be very useful.

Furthermore, it should be noted that nodes usually move

together with the same pattern. In this way, underwater nodes

can learn from the past using the results that have been

achieved so far. For instance, if a node has transmitted the

packets to a speciﬁc route, after a while, it can observe that

what percentage of packets have been successfully delivered.

The node can then decide whether to send the packets as before

or to select a new route. The stability of the decision is based

on the dynamics of the underwater environment.

9) Network topology control: Underwater sensor nodes can

be equipped with the depth adjustment mechanism which

enable them to deal with the communication void problem [8],

[66], [67]. All void nodes can then move vertically to establish

a connection with at least one non-void node. If all topological

void can be removed, then there is no need to any further

technique to bypass the void area. However, this technique

consumes high energy for topology adjustment which must be

justiﬁed for long-term and non-time-critical applications.

10) Backup network: Void-handling techniques can use

some backup facilities like AUV to obtain extra information

about the void boundary and local maxima nodes. Nodes can

forward a data packet on the boundary of a void area by

using information provided by a backup network. The in-depth

knowledge about the characteristics of local maxima nodes and

void areas can assist to design more efﬁcient void-handling

techniques than current approaches.

Nevertheless, high energy consumption and complexity

should be considered as the expenditure side of these tech-

niques. As another solution, AUVs also can identify the

networks holes and act as a relay to cover holes between the

nodes. Moreover, AUVs can be used as an alternate network

to directly collect information from all nodes and deliver them

to the sink on the surface. In this way, an AUV travels in a

predeﬁned path to collect information from sensor nodes and

deliver them to the sink after each complete rotation [68], [69].

11) Hybrid technique: If a void-handling technique is not

able to handle all kinds of voids itself, it can beneﬁts from

a hybrid technique by combining void-handling techniques

together to improve the efﬁciency [54], [70]. Sometimes,

just following a single void-handling technique can be very

expensive in terms of the required resources. Thus, sometimes

a combination of void-handling techniques can be used to

reduce the excessive use of the resources.

D. Features of the void-handling techniques for UWSNs

A wide variety of routing protocols have been proposed and

developed under different assumptions over the past few years.

Hence, it is critical to specify a set of criteria that enables us to

properly evaluate them. Without such criteria, it is impossible

to form an objective judgement, a qualitative comparison,

and a comprehensible understanding of all different factors

affecting the efﬁciency and effectiveness of current void-

handling techniques.

1) Long or short-term application: First of all, the type

of network application has a signiﬁcant impact on the model

that is supposed to be selected [71]. It should therefore be

into consideration whether the void-handling technique is de-

signed for a short-term or long-term application. In short-term

applications, there is usually no need to use some complicated

and expensive void-handling methods. However, for long-term

applications, detecting and handling the void areas with a long-

term effective strategy is of high importance.

Using the backup networks or the network topology control

approaches can only be justiﬁed in the long-term applications

because of imposing high communication overhead. On the

other hand, some void-handling techniques like heuristics,

ﬂooding, and passive participation are suitable for the short-

term applications.

2) Guaranteed delivery: One of the most important criteria

in the void-handling techniques is whether it can guarantee the

packet delivery in theory. Some void-handling techniques can

guarantee the packet delivery [37], [43], [64], [72], as long as

a topologically valid path exists between every local maxima

node and the sink. Guaranteed delivery should be proved

using the proposed void-handling mechanism and network

topological properties assuming that other factors, such as

physical links, MAC layer are in their ideal states.

The void-handling techniques using the reachability infor-

mation from the destination, can guarantee the packet delivery

by avoiding the void and trapped nodes. However, other

techniques which are blind to the network topology may fail

to ﬁnd the valid path.

3) Involving nodes: Involving as less as possible nodes

to bypass a void area is another desirable factor. An efﬁ-

cient void-handling technique should ﬁnd its path toward the

destination with the least possible number of relay nodes

(i.e. minimum number of transmissions). Perhaps the ideal

condition is that packets are not transferred to the void nodes,

which is not always possible, or to return to greedy mode as

soon as possible using the minimum number of nodes.

The strategies like ﬂooding, dense networks, and backwards

forwarding increase the number of involving nodes. However,

some strategies like passive participation, transmission power

adjustment, and void avoidance techniques are able to reduce

the number of involving nodes.

4) Communication overhead: As another metric, void-

handling technique should incur minimal overhead in terms

of the number of control packets. Some techniques require to

exchange a large amount of information between nodes which

Communications Surveys & Tutorials

is not an appropriate manner in UWSNs with long propagation

delay and limited energy. Furthermore, some techniques are

not efﬁcient in terms of the packet size. For instance, a void-

handling technique may include all receiver’s IDs in the header

of the packet which can increase the packet size.

The communication overhead is high for the void-handling

techniques such as ﬂooding, network topology control, backup

network, and learning techniques. In contrast, the heuristic and

passive participation techniques may reduce this overhead, if

designed properly.

5) Computational complexity: A void-handling technique

should be simple enough to perform its tasks under the prac-

tical situations. High complexity can increase the deployment

and computational cost of the routing protocols. Thus, an

efﬁcient void-handling technique should ﬁnd an acceptable

solution for the void problem within the minimal time and

cost range.

The transmission power adjustment techniques need a cross-

layer design to cope with the collisions in the MAC layer,

which make them a complex approach to be used in UWSNs.

The learning techniques and backup networks also increase the

complexity. However, the heuristic, the passive participation,

and the void avoidance techniques are simple enough to be

used in UWSNs.

6) Locality and scalability: Similar to the greedy routing

strategy, a void-handling technique should be able to bypass

a void area only by using the local information and in a

distributed manner, so that the scalability of the routing pro-

tocol can always be preserved [14], [73]. For a void-handling

technique, scalability means that how many local maxima

nodes can be handled without any signiﬁcant reduction in

performance. It should be noted that centralised techniques are

only suitable for small size stationary networks but impractical

in the vast underwater environment.

7) States: A void-handling technique is able to obtain a

better performance and scalability if fewer number of states

is required to be held at each void node [20]. The less

information to store, the longer they remain valid. Thus, the

void-handling technique should rely on as less as possible

number of states. According to this feature, void-handling

techniques can be classiﬁed into four categories of stateful,

stateless, partial stateful, and soft-state.

i)Stateful model: Stateful routing protocols which should

discover and hold a routing path from source to the destination,

are suitable only for the static networks [74]–[76]. Underwater

sensors continuously move with the water currents and conse-

quently the discovered routing path becomes invalid over the

time.

ii)Stateless model: In stateless techniques, nodes are almost

blind to the network topology, but the overhead is reduced

signiﬁcantly.

iii)Partial stateful model: In the partial stateful model,

nodes maintain a path not fully but limited to a few number of

hops. In order to discover and hold the neighbour’s statuses in

the partial stateful model, the impact of obstacles and noises

on the connectivity should be evaluated periodically. It should

be noted that there is always a trade-off between routing

efﬁciency and route maintenance cost.

iv)Soft-state model: In soft-state techniques, no routing path

is maintained in each node, but they use some reachability

information (e.g. hop count distance, forwarding direction)

which are useful, but not essential for efﬁciency as they can

be regenerated or updated when needed [77]. Although this

information gives a general view to each node, all routing

decisions should be made locally to hold the scalability of

the protocol. The void nodes then can be avoided in advance,

using the reachability information from the destination. The

majority of soft-state techniques are beacon-based to distribute

the information between underwater nodes. Beaconing should

be performed in regular time periods to update the information

[37], [43].

8) Path optimality: The path from a void node to a non-

stuck node (a node which can resume greedy forwarding)

should not be much worse than the optimal route [78]–[80].

In other words, the best node out of many available non-stuck

nodes should be ideally selected to provide the closest path

to the shortest path. In the geographic distance context, the

shortest path is expressed as the length of straight line between

two nodes.

The void-handling techniques using the reachability infor-

mation from the destination may discover the near-optimal

paths by avoiding the void and trapped nodes. However, other

techniques, which are blind to the network topology may fail

to ﬁnd the optimal paths.

9) Path reliability: An efﬁcient void-handling technique

can decrease the overall transmission cost per packet by

avoiding lossy links and thereby attain higher throughput.

To ﬁnd a high-quality path, the void node should select the

candidate nodes with the highest packet delivery probability.

In order to calculate the packet delivery probability, some

factors like the attenuation, the ambient noise, and the distance

between nodes should be taken into account [81].

10) Opportunistic forwarding: Any new proposed void-

handling technique should take into account the high bit

error rate issue in the underwater environment. Opportunistic

routing is a promising solution to deal with lossy environ-

ments. In this way, packet forwarding is enhanced by taking

advantage of simultaneous packet reception among one node’s

neighbours and their collaboration to forward the packet [82]–

[85]. Reliability and throughput can be increased by using the

opportunistic forwarding in which packet is relayed by nodes

collaboration in each hop.

11) Sender-based or receiver-based: The void-handling

technique should determine whether it is the void node which

decides to whom the packet is transmitted to (sender-based),

or it just broadcasts the packet and then each receiver decides

whether to include itself in the forwarding process (receiver-

based). In sender-based techniques, the forwarding node puts

the ID numbers of all candidate nodes in the packet’s header.

The receiving node accepts the packet if its ID is included in

the packet header.

In receiver-based techniques, when a neighbouring node

receives a packet, it can accept or drop the packet according to

its current status and the void-handling criteria (e.g. whether

or not to be placed in the forwarding area).

Communications Surveys & Tutorials

12) Energy efﬁciency: Moreover, a void-handling method

should be energy efﬁcient to prolong the life of the nodes

and network [86]. Underwater sensors consume more energy

than terrestrial sensors due to acoustic signals used as their

communication medium [87]–[89]. The packet transmission

by the void node wastes the energy [86]. As an efﬁcient

approach, a local maxima node can inform other nodes about

its current status to prevent them from sending packets to it.

This strategy can reserve energy in the local maxima node.

The implicit ACK can also reduce the energy consumption of

the void-handling technique since no extra packet are required

to conﬁrm the delivery. In this model, when a node overhears

that one of its neighbouring node forwards a packet which is

already in its buffer, it can consider it as an ACK [90].

The uniform energy consumption of the nodes is also of

high signiﬁcance [45], [91], [92]. Each local maxima node

should therefore consider the residual energy of the neighbour-

ing nodes in order to preserve the uniform energy consumption

in the network. In the case of having many destinations, nodes

with lower residual energy can be easily avoided during the

packet forwarding.

Increasing the node density or using the full ﬂooding

techniques can exacerbate the energy consumption; however,

the transmission power adjustment, the void avoidance, and

passive participation approaches can reduce the energy con-

sumption.

13) End-to-end delay: It is deﬁned as the average time

taken from the moment of the creation of packets at the

source node until successfully being delivered to the sink node.

This parameter is critical for the delay-sensitive applications.

Thus, an efﬁcient void-handling technique should decrease the

packet delivery time. The delay is dependent on various factors

such as the packet holding time, the void-handling strategy,

number of hops between the source and destination, network

density, and communication overhead [30], [37].

The full ﬂooding, backwards forwarding, and network topol-

ogy control can increase the latency; however, increasing

the node density, void avoidance, and passive participation

techniques may reduce the end-to-end delay. The network

coding can also be used by the void-handling techniques to

increase the network throughput, and to reduce the delay [90].

14) Quality of service: In some applications, packets are

transmitted with different levels of priority. Some data can

be more critical than other normal packets which only carry

information about the ordinary events [93]–[95]. Thus, an

efﬁcient void-handling technique should deal with the critical

packets differently. In order to maintain vital resources for the

critical packets and deliver them on time, ordinary packets can

be delivered using the best effort approach (e.g. using a longer

path or with more delay).

15) Activeness: Depending on application requirements, a

void-handling technique can be divided into four categories of

reactive, proactive, preventative, and hybrid.

i)Reactive model: In a reactive model, the void-handling

technique is triggered when a packet is stuck in a local maxima

node. The path discovery for each stuck packet is performed

on demand. After bypassing the void and ﬁnding a node with

positive progress toward the destination, the greedy strategy is

resumed.

ii)Proactive model: In a proactive strategy, stuck nodes in

the network are discovered in a preprocessing phase and the

path to bypass the void is stored in each stuck node. When a

packet is stuck at a void node, it follows a predeﬁned path to

bypass the void area.

iii)Preventative model: In the preventative model, en-

countering a void area is prevented with the aid of some

precautionary measures such as excluding the void and trapped

nodes from the packet forwarding. These techniques try to

send the packets to the non-void nodes before encountering

a void area. Nevertheless, in some of these techniques, local

maxima nodes cannot be avoided thoroughly. Moreover, these

techniques need some extra information about the network

topology which should be obtained and updated periodically.

iv)Hybrid model: In the hybrid model, void-handling

techniques consist of at least two void-handling techniques

together to obtain more reliability. For instance, void-handling

technique can try to avoid the holes with a preventative

technique as far as it is possible, and also applies a reactive

or proactive technique to deal with the packets may be stuck

in a local maxima node.

IV. CLA SS IFI CATION O F VOI D HAN DL IN G TEC HN IQ UE S

Several routing protocols have been proposed for UWSNs

over the past few years. There are many ways to classify

geographic routing strategies and void-handling techniques in

UWSNs. Generally, they can be categorised into two groups:

location-based and depth-based. The main difference between

these protocols is related to the location service which is

responsible for determining the position of the nodes [5], [9].

In the location-based category, all nodes are aware of their

3D location information by the aid of some localisation ser-

vices [96]–[99]. However, it should be mentioned that Global

Positioning System (GPS) cannot be used in underwater envi-

ronment as a localisation system because of quick attenuation

of its waves in water [5].

During the data forwarding of location-based routing pro-

tocols, each node can decide about relaying the packet based

on its position, the position of the destination, and routing

criteria. In some of the location-based protocols, each node

should have a table to hold the positions of the neighbouring

nodes [30], [100]. The main difference between these protocols

is that each protocol tries to apply different ﬁtness factor for

selecting the next forwarding nodes.

In this category, a node is called a local maxima node, if

it cannot ﬁnd any other node with positive progress toward

the destination in terms of geographical distance (euclidean

distance). After encountering a void area, local maxima node

may initiate a local search, within multi-hop neighbouring

nodes vicinity, to ﬁnd a node which is geographically closer

to the destination than itself, or just drop the received packet.

Depth-based routing is another class of geographic routing

protocols which is simpliﬁed to use only depth information to

route the packets [49], [101], [102]. The depth of each node

in water can be estimated through a pressure gauge which is

embedded on it [43]. The ﬁnal destination is located on the

Communications Surveys & Tutorials

TABLE II: Comparison between the location-based and pressure-based routing protocol classes

Location-based Pressure-based

All nodes are aware of their 3D location information All nodes are aware of their depth using a pressure gauge

Dependent on the localisation services No need to any localisation services

The destination location should be known to all nodes No need to the destination location knowledge

In some protocols under this class, each node holds the positions of the

neighbouring nodes

In some protocols under this class, each node holds the depths of the

neighbouring nodes

Every packet carries the position of the destination No need to carry the destination location in each packet

Routing decision is based on the geographical locations Routing protocols is simpliﬁed to use only the depth information

The ﬁnal destination can be placed everywhere The ﬁnal destination is assumed to be on the surface

A void node is determined based on the geographical positions of the neigh-

bouring nodes A void node is determined based on the depth of the neighbouring nodes

A recovery node is geographically closer to the destination than void node A recovery node has a lower depth than the void node

TABLE III: Comparison between the unicast, anycast, and geocast models

Unicast Anycast Geocast

Single sink architecture Multi-sink architecture Nodes in a particular geographical area as destination

Suitable for small networks Suitable for large networks Suitable for large networks

Packet delivery is successful if packet is received

by the single sink

Packet delivery is successful if packet is received

by any sink

Packet delivery is successful if packet is received

by all the nodes within the geocast region

Void is determined with respect to the single sink Void is determined with respect to all the sinks Void is determined with respect to the packet

entry area to the geocast region

Limited number of paths from a source to des-

tination More available paths from a source to sinks Available paths depend on the covering area

around the geocast region

surface and, therefore, packets are forwarded to lower depth

at each hop.

A node is a local maxima node if all of its neighbouring

nodes make negative progress towards the surface. In this

case, the void-handling problem is simpliﬁed to a searching

process for a node with a lower depth value than that of

forwarding node (but not necessarily a node with the closest

geographical distance to the sink node) [42]. After ﬁnding such

a node, depth-based greedy routing can be reactivated [42].

The features of location-based and depth-based techniques are

also summarised in Table II.

Moreover, geographic routing protocols in UWSNs also can

be classiﬁed based on the number and position of destinations

into three groups: unicast, anycast, and geocast. In the unicast

model, all forwarding nodes should deliver the packets to a

speciﬁc sink on the surface (single sink architecture). Accord-

ing to the characteristics of unicast model, void occurs with

respect to the position of the single sink on the surface.

On the other hand, in the anycast model (multi-sink archi-

tecture), there are a number of destinations (sinks or buoys) on

the surface which can be utilised during the packet collecting

phase [42]. In this way, all packets can be delivered via anycast

routing to any sink or buoy on the surface. Due to the existence

of different paths, a node can be considered as a local maxima

Communications Surveys & Tutorials

Fig. 6: The Classiﬁcation of Void-handling techniques for UWSNs

node with respect to one of the sinks, but at the same time,

it is a non-void node from another sink viewpoint. Thus, a

local maxima node simply can select another sink as its ﬁnal

destination to solve the void problem.

In the geocast model, a group of nodes in a particular

geographical area and in a speciﬁc time interval will be

selected as the destinations [19]. In this case, a packet should

be received by all the nodes inside the target region to have

a successful delivery. In surveillance applications, when an

anonymous sensor or any underwater vehicle is sensed, it may

be required that a data packet is generated and forwarded to

a group of sensor nodes in a speciﬁc geographical region [7],

[103]. Geocasting also can be used to initiate a query asking

for needed information from the subset of underwater sensors

[103]. In this group of protocols, the void within the geocast

area should be addressed and resolved [7], [104]. The void is

determined with respect to the packet entry area to the geocast

region and often is resolved by considering a covering area

around the geocast region [7], [103]. The features of unicast,

anycast, and geocast models are also summarised in Table III.

Depending on the availability of location service and also

the destination status, a classiﬁcation of the state-of-the-art

void-handling techniques in UWSNs is presented in Fig. 6.

Under the class of location-based protocols, different routing

protocol with void-handling techniques have been proposed in

the literature for different architectures such as unicast (single

sink), and geocast. For the pressure-based protocols, anycast

(multi-sink) architecture also has been used. However, to the

best of our knowledge, no geocast protocol has been proposed

so far for pressure-based class of protocols, as it is really

challenging to identify a group of nodes as the temporary

destination without relying on a robust localisation system.

The next section is presented to review the void-handling

techniques in details.

V. VOID-HA ND LI NG TE CH NI QU ES

In this section, we review the basic principles of UWSNs

void-handling techniques reported in the literature. To be

more speciﬁc, we focus mainly on the void-handling tech-

nique rather than the routing strategy. Thus, for each routing

protocol, the routing strategy is brieﬂy explained and then

the void-handling technique is comprehensively analysed. In

this analysis, our main focus is to study the void-handling

techniques without considering other unrelated information

such as network characteristics. According to our taxonomy

presented in Fig. 6, existing void-handling techniques are

classiﬁed into two main categories of location-based and

depth-based. We present all void-handling techniques under

each category along with a qualitative discussion. The features

of all void-handling techniques presented in this paper are also

summarised in Table IV. Our discussion relies on the features

presented in Section III-D.

A. Location-based void-handling techniques

In this section, we classify existing location-based void-

handling techniques into two subcategories of unicast, and

geocast and individually discuss each of them. First, we

discuss the void-handling techniques proposed under unicast

category (single-sink architecture) like VBVA [29], AHH-

VBF [30], DFR [54], and FBR [57]. Then, it is followed

by discussing the void-handling techniques proposed under

geocast category like RMTG [103], and Mobicast [7]).

1) Vector-Based Void Avoidance (VBVA): VBVA [29] is a

reactive, stateless, and receiver-based technique which pro-

posed to mitigate the negative impact of void communications

on the vector-based routing protocols such as VBF [46], and

HH-VBF [47]. In VBVA, each node knows the location of the

sink, the source node (via the packet header), and itself. VBVA

exploits two approaches, vector-shift and back-pressure for

dealing with the convex voids and concave voids, respectively.

In the packet forwarding section, VBVA exactly follows a

Communications Surveys & Tutorials

Fig. 7: Vector-shift mechanism in VBVA [29]

vector-based approach (VBF) to forward the packets toward

the sink.

When the routing procedure faces a convex void like Fig. 7,

VBVA tries to route the packets along the boundary of the void

with the aid of the vector-shift mechanism. To do this, local

maxima node (node S) broadcasts a recovery packet called

vector-shift to all its neighbours (nodes aand d). The vector-

shift packet enables the nodes outside the current pipeline

to participate in the packet forwarding by creating the new

vectors emitted from them toward the sink. This procedure

can be repeated by other nodes till packet is delivered to the

destination.

When VBVA cannot ﬁnd any neighbouring node by shifting

method because of placing in a concave void, it initiates the

back-pressure mechanism to route the packet backward to ﬁnd

some suitable nodes to do vector-shift (like the procedure

shown in Fig. 8). To do this, local maxima node broadcasts

a control packet, called back-pressure to let the other nodes

with negative progress perform the vector-shift, in the hope

that any available path toward the destination can be found. In

the case of not ﬁnding any path, the back-pressure procedure

is continued in the receiving nodes until the vector-shift

mechanism can successfully be accomplished. For instance, in

Fig. 8, the packet will be forwarded back from node cto the

node Swhere the vector-shift mechanism can be successfully

applied.

VBVA initiates the void-handling mechanism on demand

while no extra information (e.g. neighbouring information,

void characteristics) is required to be stored in each node.

This feature increases the scalability and robustness of VBVA

for highly dynamic and large network applications. However,

the recovery procedure of VBVA is too complicated to be

performed in the real underwater environment. VBVA lets

the packets to be trapped in a concave hole and then tries

to recover them using a time-consuming procedure which

increases the end-to-end delay. As a further matter, it is

Fig. 8: Back-pressure mechanism in VBVA [29]

obvious that vector-based protocols suffer from duplicated

paths, and vector-shift mechanism can exacerbate the problem

as can be seen in Fig. 8. By receiving the packets in two

different sides of the void node, packets are subsequently

delivered along both boundaries of the void area, resulting

in more energy expenditure.

2) Adaptive Hop-by-Hop Vector-Based Forwarding (AHH-

VBF): AHH-VBF [30] routing protocol applies a preventative

technique to cope with the void problem. In AHH-VBF, every

node knows the locations of the sink, the sender node (via the

packet header), one-hop neighbours, and itself. This routing

protocol is on the basis of HH-VBF, in which direction of

the forwarding pipeline is changed hop by hop. In terms of

dealing with the void, AHH-VBF is equipped with an adaptive

approach which not only changes the direction of the pipeline

but also the radius of the pipeline based on the neighbouring

nodes distribution. For instance, when the region ahead is

sparse, the radius of virtual pipeline (between the forwarding

node and sink) is increased to cover a broader and larger area.

In this way, more candidate nodes may be found to relay the

packet and the transmission reliability is enhanced.

Unlike HH-VBF which uses a constant power level, AHH-

VBF is able to adaptively adjust the power level according

to the density of neighbouring nodes. Thus, the transmission

power level can be increased to cover longer distance in sparse

networks, or decreased to save more energy in dense networks.

In AHH-VBF, each node has a local knowledge about its

neighbours which is utilised to select the proper power level. In

order to update information about neighbours, each node sends

control packets in periodic times which are determined accord-

ing to the speed of network topology changes. For instance,

when the network topology changes very fast, neighbouring

information should be exchanged within a shorter period of

time.

AHH-VBF does not always guarantee the delivery of

packets to the destination. Although adaptively changing of

the forwarding area in a hop-by-hop manner and adjusting

Communications Surveys & Tutorials

Fig. 9: Packet forwarding in DFR [54]

transmission power can handle small void areas within the

pipeline, it is not ﬂexible enough to change the forwarding

direction when confronting with large holes. It is always

possible that the pipeline at its maximum size excludes any

node inside (like the example in Fig. 4). Thus, an efﬁcient

void-handling technique should be able to shift the packet to

another area except the predeﬁned area by the protocol. AHH-

VBF, however, lacks such an ability.

Furthermore, the decision on the size of the forwarding

region only depends on the distribution of candidate nodes

which is not an appropriate approach when nodes follow an

irregular distribution. For instance, majority of neighbours

may be gathered in a location close to the forwarding node,

but because of sensing a neighbour in a far distance, power

level should be set at its maximum level which is clearly

a waste of energy. As an another problem, if the pipeline

radius is set too large to bypass a void, forwarding nodes on

the opposite corners of the pipeline will not be able to hear

each other and probably forward the same packet concurrently.

As a consequence, many duplicated paths are established

between the sender and receiver which causes more energy

consumption and collision.

3) Directional Flooding-based Routing (DFR): DFR [54] is

a location-based, stateless, and receiver-based routing protocol

in which every node knows the locations of the sink, one-

hop neighbours, and itself. DFR takes beneﬁt of a controlled

ﬂooding approach to achieve more reliability in confronting

with the various link qualities in UWSNs. For this purpose,

each node adjusts the ﬂooding zone based on the link quality

of the area ahead. The intended progress area is considered

toward the only existing sink on the surface. For instance,

Fig. 10: Request to detour phase in DFR [54]

when facing a poor link toward the sink, ﬂooding zone will

be set in a way that more nodes can participate in the packet

forwarding. On the other hand, if the network is strongly

connected, packets can be relayed with the collaboration of

few nodes.

However, the void problem is still unresolved where no node

can be found in the ﬂooding zone. Accordingly, two types of

void problems can appear during the packet forwarding. The

ﬁrst type of void is when a ﬂooding zone without any node

is established which causes the packet delivery failure. This

phenomenon happens when the ﬂooding zone is continuously

decreased due to the good link quality among neighbours while

no node can be located in the zone. Thus, DFR exploits a

preventative void-handling technique to adjust the ﬂooding

zone to cover at least one node to relay the packet.

As shown in Fig. 9, in order to determine the ﬂooding zone

at each forwarding node, DFR considers two angles including

a Reference-Angle value which is selected by the forwarding

node (node p), and a Current-Angle which is determined by the

geographic location of the receiving node respect to the source

and destination (angle between fs and fd). The nodes with the

higher Current-Angle values than Reference-Angle value are

placed in the forwarding zone and qualiﬁed to participate in

the packet forwarding. Reference-Angle value at least should

be smaller than one of the candidates’ Current-Angle values, to

meet the void-handling requirements. Thus, it is only sufﬁcient

that Reference-Angle value is set smaller than the maximum

angle value (amongst the candidates’ Current-Angle values) to

cover at least one node in the ﬂooding zone, which helps to

prevent voids.

The second model of the void is that none of the forwarding

Communications Surveys & Tutorials

Fig. 11: Forwarding node selection in FBR [57]

node’s neighbours is closer to the destination than itself. So

the ﬂooding zone cannot be established in any way. However,

it is possible that a topological detour path can be found via a

neighbour with negative progress. To solve this problem, DFR

ceases the greedy forwarding phase and tries to bypass the

void by ﬁnding a detour path. Thus, DFR initiates a reactive

void-handling technique which is presented as follows.

As shown in Fig. 10, if forwarding node fcannot ﬁnd any

node with positive progress, it seeks to ﬁnd a neighbour which

has the smallest Current-Angle value among neighbours. After

detecting the eligible node j, forwarding node unicasts a RTD

(request to detour) packet including the original data for this

intended node. Upon receiving the RTD packet, receiving node

sends an acknowledgement in response to the forwarding node,

while it updates the variables in the data packet and forwards

it toward the sink. If forwarding node does not receive an

acknowledgement from the target node, it continues unicasting

to other candidates until ﬁnding a detour path.

DFR exploits a hybrid void-handling technique which is

simple enough to be implemented and also scalable to be

used for large networks. However, in the face of poor link

quality and void communications, ﬂooding zone becomes so

big which makes it prone to have the hidden terminal problem

[37], [42]. Thus, multiple nodes can forward the same packet

due to the hidden terminal problem and waste the network

resources. Furthermore, when a closer node to the sink cannot

be found, the recovery technique is not reliable because of

using a unicast approach. This approach does not comply

with the opportunistic characteristics of the routing technique

and also increases the end-to-end delay and communication

overhead. The recovery model is not loop-free and detour path

is not optimal, because DFR has no knowledge about the void

boundaries and shapes.

4) Focused Beam Routing (FBR): FBR [57] exploits a

preventative void-handling technique to avoid void commu-

nications areas. In FBR, each node knows the location of the

sink node, the sender node (via the packet header), and itself.

The main objective of FBR, as a cross-layer protocol, is to

minimise energy consumption by controlling the forwarding

nodes transmission power. The main assumption is that each

node can adjust its transmission power with a choice between

a ﬁnite number of power levels as well as having the ability

to perform beam forming.

In order to conserve energy, each forwarding node initiates

an exploration for available candidates by multicasting a

RTS (Request to Send) packet, at the lowest power level.

If no neighbour responds to this request, forwarding node

is able to increase the transmission power stepwise, until at

last a suitable candidate can be located. Those nodes are

eligible candidates which lie within a cone emanating from the

forwarding node towards the ﬁnal destination. All the receiving

nodes are able to determine that they are within the transmit-

ter’s cone or not, based on the provided location information

in the packet (source location, destination location) and their

own locations.

Considering the case in Fig. 11, node Aintends to send

a packet to node B. At the ﬁrst packet transmission with

lowest power level, node Acannot ﬁnd any candidate node.

However, by increasing the power level, it can detect two

candidate nodes in its cone (node Dand C). Eventually, node

Dis selected to relay the packet because it is closer to the

destination among candidate nodes. If no relay node can be

located toward the destination at the maximum transmission

power, the main cone will be shifted to the left or right for

bypassing the void ahead. By using this strategy, data packet

observes a minimum deviation from the straight line between

the source and destination (minimum amount of zigzagging),

and it also gives another chance to the packet for being

forwarded through other available paths.

The advantage of transmissions in the short geographic

distances is twofold. First, it can increase the reliability due

to access higher bandwidth. Second, it decreases the energy

consumption because of using low power levels. If needed,

transmission power can be increased to propagate the signal

beyond of the current transmission range.

However, variable transmission ranges can interfere to the

other nodes’ activities while requiring complicated MAC

protocols to handle it. In FBR, sending and receiving the

control packets in each hop to establish a connection is very

time-consuming. The long-delay problem is exacerbated in the

sparse networks, because nodes are usually far away from each

other and only high power transmissions can connect them

together, while sending the control packets already have to be

done for all the lower power levels. In the case of cone shifting,

although FBR can efﬁciently bypass the convex voids with

slim shape, it has difﬁculties with other kinds of voids (e.g.

concave, fat shape), due to the lack of a proper mechanism.

For instance, when a packet is trapped in a concave void, cone

should be rotated 180 degrees to forward the packet back,

which seems impractical in the existing approach.

5) Routing and Multicast Tree based Geocasting (RMTG):

RMTG [103] is a 2-D geocast technique in UWSNs with

the hole detection ability to distribute data in a speciﬁed

geographical area covering a group of sensors (geocast re-

gion). In RMTG, each node knows the location of destination

area (via the packet header), neighbouring nodes, and itself.

Unlike many of the geocast techniques which use the ﬂooding

approach in the target region, RMTG utilises a covering area

Communications Surveys & Tutorials

Fig. 12: Route maintenance in RMTG [103]

around the geocast region for data dissemination.

With the aid of greedy forwarding, the packet is relayed

toward the target region. During the process of ﬁnding a

route to reach the geocast region, if a packet is stuck in a

local maxima node, the packet will be forwarded back to the

previous hop node. For instance, in Fig. 12, node bcannot ﬁnd

any neighbouring node with positive progress toward the target

region, so the packet is forwarded back to node a. By receiving

the error packet, the previous hop node divides the region into

four quadrants and selects the best next node from the quadrant

which is nearer to the geocast region (like selecting node c

in Fig. 12). This procedure will continue until the packet is

delivered to the geocast region.

Upon receipt of the packet in the target region, the ﬁrst

receiving node acts as a root node and creates a multicast

shortest path tree to disseminate the packet within this area.

Sometimes the constructed tree in the geocast region is not

able to cover all destination nodes as the leaf nodes due to

the presence of a void area (e.g. nodes e,f,gin Fig. 13).

Nevertheless, RMTG is able to detect hole inside the geocast

region where a packet cannot be forwarded any longer and

the boundary of the target region is not still reached. In this

situation, a virtual area of radius ris established around the

boundaries of the target region to involve more forwarding

nodes to handle the void problem.

In order to enter the geocast region from the other faces,

the root node initiates boundary traversing around the geocast

region, passing through the virtual area by selecting the

clockwise or counter-clockwise direction. During boundary

traversing, if the packet is received by a node within the

target region for the ﬁrst time, the packet will be delivered

to this node and subsequently its neighbouring nodes. Again,

boundary traversing is resumed until all the selected faces are

traversed and the packet is delivered to all remaining nodes in

the target area. A sample of this procedure is shown in Fig.

13.

In RMTG, the assumptions about two-dimensionality of the

underwater environment and using GPS in UWSNs are clearly

improper. The route discovery and route maintenance are also

Fig. 13: Boundary Routing in RMTG [103]

inappropriate with respect to the nodes movement and rapid

changes in the network topology. Moreover, a larger virtual

area around the target region may involve extra relaying nodes

which causes more energy consumption and also a smaller area

resulting in the lower chance of packet delivery.

6) Mobicast routing protocol: Mobicast [7] is a mobile

geocasting approach which aims to collect data from a 3D

underwater area in the presence of various water currents and

void areas. In Mobicast, each node knows its current speed, the

location of itself in different time stamps, and geacast region

(via the control packet by AUV). In this approach, there is

an AUV as a mobile sink which traverses a predetermined

route (usually a circle path) to collect data from sensor nodes

in different geographic regions called 3-D zone of references

(3-D ZOR). Nodes usually stay in sleep mode to save energy.

As shown in Fig. 14, when AUV collects data packets from

sensors within 3-D ZORt, it should notify the sensors within

3-D ZORt+1 to enter the active mode to be ready for the arrival

of AUV. However, message delivery to the next group of nodes

(within 3-D ZORt+1) that is supposed to be investigated at

the next time slot is a challenging task due to the presence

of topology holes. If a topological hole blocks the routing

path between AUV and next target region, sensor nodes cannot

wake up on time to send their packets to AUV.

To overcome the hole problem, Mobicast routing protocol

creates a covering area for surrounding the hole to ﬁnd

alternative paths to deliver the packet to all nodes within

the 3-D ZORt+1. As can be seen in Fig. 14, a 3-D zone

of forwarding (3-D ZOFt+1) around the region of interest is

considered which is larger than or equal to the size of 3-D

ZORt+1. If a feasible routing path cannot be found in the

region of interest (3-D ZORt+1) due to the void problem, it

is possible that an alternative path can be discovered in 3-D

ZOFt+1 (e.g. the discovered path via 3-D ZOFt+1 in Fig. 14).

The size of covering area (3-D ZOFt+1 ) depends on the net-

work density and the velocity of ocean current. For instance,

if there is no topological hole and water current, the size of

3-D ZOFt+1 is exactly equal to the size of 3-D ZORt+1. On

the contrary, when there is a hole along with the ocean current

Communications Surveys & Tutorials

Fig. 14: Void handling in Mobicast protocol [7]

Fig. 15: Problem in void-handling in Mobicast protocol

and low network density, the covering area should be enlarged

to cover more sensor nodes for route discovery. Nevertheless,

estimating the accurate size of the covering area according to

the available information of the current area (3-D ZORt) is

unreliable. Hence, with the aid of the real-time information,

Mobicast exploits an adjustment scheme to determine a proper

size for the covering area.

Another important issue is that how many sensor nodes

within the covering area should be used to deliver the packet.

The AUV takes into account the impact of water current on the

successful delivery rate within different parts of the covering

area and only wakes up the sensor nodes with high successful

delivery rate. Inspired by the apple slice concept, the AUV

divides the covering area into several identical parts (segments)

and only selects those parts which are able to deliver the packet

successfully.

Some constraints can conﬁne the performance of Mobicast.

First of all, collecting information in this manner is suitable

for short-term applications; however, in the long-term applica-

tions, AUVs and ﬁxed underwater sensors should collaborate

to monitor the underwater environment. Secondly, Mobicast

cannot always ﬁnd a topological valid path from AUV to the

nodes within the 3-D ZORt+1, if any. This is due to the fact

that the alternative path may be passed from another area rather

than ”hold and forward zone” which is an overlapping area

between ZORtand ZORt+1 . This problem is shown in the

Fig. 15, in which the topological valid paths of A-B-C-D and

E-F-G-H cannot be discovered using the Mobicast protocol.

As the last point, the performance of Mobicast highly depends

on some parameters which can be set by the user such as user-

deﬁned path and response time, the radius of hold and forward

zone, and expected range for data collection of AUV.

B. Depth-based void-handling techniques

In this category, greedy routing can be accomplished by

utilising the depth information (or along with some extra in-

formation) without knowing the full geographical coordinates.

Hence, there is no need to support the routing protocol with a

costly distributed localisation mechanism in order to provide

each node’s coordinates [42], [49], [105].

In depth-based protocols, a node is considered as a local

maxima node, if it cannot ﬁnd any node with lower depth

in its neighbourhood. Hence, the void-handling problem can

be simpliﬁed to a route discovery method to ﬁnd a node

whose depth is lower than that of the current node to resume

the greedy approach. Some routing protocols (e.g. DCR [66],

and GR+DTC [8]) are location-based in the routing phase;

however, their void-handling techniques are mostly based on

the depth values. Therefore, we classiﬁed them in the pressure-

based category in our analysis.

Again, we ﬁrst discuss the depth-based void-handling tech-

niques with unicast objective (single-sink architecture) like

LLSR [64], IVAR [72], and OVAR [37]. Afterwards, we will

discuss the void-handling techniques with anycast objective

like DCR [66], GR+DTC [8], HydroCast [42], VAPR [43],

and WDFAD-DBR [63].

1) Location-free Link State Routing (LLSR): LLSR [64]

uses a greedy hop-by-hop routing by relying on the parameters

such as hop count, path quality, and pressure. Hop count shows

the proximity of node to the sink which enables LLSR to

bypass void area in a preventative way. In this protocol, path

quality indicates redundancy of routes which is measured by

counting the number of neighbouring nodes with lower hop

count values. This location-free approach is placed under the

pressure-based and beacon-based categories.

In beacon dissemination phase, each node periodically

broadcasts a beacon including hop count, path quality, and

pressure. The beaconing starts from the sink node instead of

a source node. The hop count value of sink is equal to zero

and other nodes gradually obtain their hop distance and path

quality toward the sink by receiving the beacons. In the routing

phase, each node selects a one-hop neighbour with lowest hop

count value as its next hop node. In the case of a tie, the node

with greatest path quality is selected to relay the packet. If the

tie persists after considering the path quality, the neighbouring

node with lowest pressure is selected as the next hop. In LLSR,

Communications Surveys & Tutorials

selecting a node with the lowest pressure contributes to higher

progress toward the surface where the sink is located, and it

can decrease the routing distance between the source node and

sink.

By moving the nodes over time, some established paths are

not valid any more and should be updated. So nodes are able

to update their tables and broadcast a beacon according to the

network topology changes. When a node recognises that it is

a void node, it should send a beacon with a hop count value

equal to inﬁnity for its neighbours. Upon receiving a beacon

from a void node, receivers are aware of the lost connection

and change their routing path, if necessary. When a void node

ﬁnds a new path toward the sink, it changes its hop count

value according to the newly discovered path.

The strength of this approach is that it is a loop-free strategy

with a mechanism to address the network topology changes

(e.g. broken links). By exploiting the reachability information

from the sink, the void nodes can be detected and bypassed,

and by using the pressure information, a positive progress

toward the sink can be obtained. One major drawback of

this approach is that LLSR does not take into account the

opportunistic data forwarding in UWSNs. In this way, only

one node is selected to relay the packet in each step which

increases the chance of packet loss. Also when the path quality

(redundancy of routes) is prioritised over the pressure metric

(closeness to the surface) in the forwarding node selection, the

packet advancement (toward the sink) may be sacriﬁced.

Furthermore, this method is not perfectly optimal in terms

of energy consumption because beacons are always being

sent even for the isolated nodes and the void nodes with

no connections to the sink. Although LSSR can bypass all

kinds of voids by using reachability information, the void-

handling efﬁciency depends on different parameters such as

becoming period, node movement speed, the network topology

dimensions. If a node becomes a void node in the upper layers

of water, it may affect many nodes status in deeper layers.

2) Void Avoidance Routing Protocols (IVAR, OVAR): Inher-

ently Void Avoidance Routing (IVAR) [72] proposes a soft-

state routing protocol which inherently excludes all the routes

leading to a void area and therefore does not need to switch

to recovery mode. IVAR can transfer a packet around the

boundary of a hole and deliver it to the destination only by

using depth and hop count information in each node.

This protocol initiates a beaconing process from the desti-

nation node instead of a source node. In this way, sensor nodes

can obtain reachability information via periodic beaconing by

the sink and relay nodes. Each beacon includes the hop count

information, which shows the proximity of nodes to the sink.

As can be seen in Fig. 16, when a node broadcasts a data

packet, all the receiving nodes with smaller hop count are

potentially a candidate node to forward the packet. To avoid

multiple forwarding of a packet by more than a node, each

node considers its depth as the second metric to set a relaying

timer. Each receiving node individually sets a relaying timer

based on its depth advancement toward the water surface. The

relaying timer of the node with the lowest depth is expired

ﬁrst allowing the node to forward the packet. Other candidate

nodes in the vicinity of the forwarding node should discard

Fig. 16: Void-handling technique in IVAR [72] and OVAR [37]

the packet after hearing this packet transmission.

IVAR is a receiver-based forwarding model in which no

neighbouring node information is required to be held by

a forwarding node. Each receiver node is able to locally

decide whether to participate in packet forwarding only by

comparing its hop count value with the sender node. As can

be concluded from Fig. 16, when the trapped and void nodes

receive a data packet, they simply drop the packet having a

hop count value equal or higher than that of the sender node.

However, IVAR is unable to suppress all duplicated paths and

transmissions resulting from broadcast nature of this method.

The candidate nodes may be located in different direction

of the forwarding node which leads to the hidden terminal

problem and subsequently energy dissipation.

Opportunistic Void Avoidance Routing (OVAR) [37] is

proposed to overcome the drawbacks of IVAR in dealing

with the duplicated packets and hidden node problem in the

forwarding set. OVAR takes the advantage of beaconing proce-

dure, similar to IVAR, to handle the void communication issue.

The only difference is that, in OVAR beaconing procedure,

the one-hop neighbouring information is held to construct an

adjacency graph at each forwarding node. OVAR is a sender-

based approach, in which the forwarding node selects the

candidate nodes and put their IDs in the packet header. In

order to suppress the duplicated packets, the candidate nodes

are selected in the vicinity of each other to exclude any hidden

node in the forwarding set. With a view to managing the

energy, the number of collaborative nodes can be adjusted

according to the density of the network.

In IVAR and OVAR, the near-optimal path is selected in

almost all cases and void areas are smoothly and efﬁciently

bypassed while its path selection is almost insensitive to

node mobility. In dense scenarios, these protocols can deliver

packets using a shorter distance compared to other unicast

routing protocols. In sparse networks with many possible void

areas, these protocols are also able to ﬁnd the best, or close

to, path, if any, with minimum communication overhead [37].

However, there are some limitations which can conﬁne their

performance. Selecting an appropriate value for beaconing

Communications Surveys & Tutorials

Fig. 17: Depth adjustment in DCR [66]

intervals also has great impact on the network performance

in terms of communication efﬁciency and information va-

lidity. Beaconing intervals with lower values impose high

communication overhead to the network, but nodes have more

accurate information about the network topology. Furthermore,

higher values of beaconing intervals lead to unreliability and

inaccuracy of information hold at each node. Despite the fact

that the scalability of soft-state routing protocols is much better

than that of stateful protocols, it is still less scalable compared

to the stateless protocols.

Although OVAR has eliminated drawbacks of IVAR in the

packet forwarding, OVAR is slightly more complex than IVAR

due to its mechanisms to remove the effects of hidden nodes

and also to hold energy-reliability trade-off.

3) Depth-Controlled Routing (DCR): DCR [66] is the ﬁrst

geographic routing protocol which exploits a network topology

control scheme (as described in Section III-C9), to deal with

the void problem in UWSNs. In DCR, each node knows its

pressure and the location of all sinks, the neighbouring nodes,

and itself. Network topology control improves the network

connectivity and diminishes the impact of the void problem by

utilising the vertical movement capability of the nodes. Due

to the fact that underwater nodes can move vertically, void

nodes and disconnected nodes are able to change their depth

to be connected to other nodes with an available path to a

sonobuoy. In this protocol, the AUVs and on-board hydraulic

pressure gauge are used to provide 3D location information

for the underwater nodes.

In the routing stage, each node forwards the packet to the

nearest sonobuoy via a greedy approach. However, some nodes

fail to locate a next-hop node to reach any destination on the

surface which makes them eligible for depth adjustment. First,

all void nodes are identiﬁed by the DCR protocol using a

centralised algorithm to determine the set of void nodes and

calculate new depth values for them where greedy routing

becomes possible. In depth-control stage, the Depth-First

Search algorithm is initiated by all surface sonobuoys as root

to identify all connected and disconnected nodes. Afterwards,

all disconnected nodes are sorted from the shallowest to the

deepest nodes and depth adjustment will be performed with

this prioritisation.

Following this approach, each void node considers a set

of candidate nodes with an available path to a sonobuoy and

inside the cylinder shape, centred in the void node with a

speciﬁed radius (e.g., Fig. 17). The new depth of the void

node is examined with respect to all candidate neighbours and

eventually void node moves to a new depth where it can be

connected to a candidate node with minimum displacement.

Void nodes are informed about their new depths via AUVs.

When the depth adjustment phase begins, the network opera-

tion will be stopped until the topology is reformed. During the

packet forwarding phase, if a node realises that it cannot locate

a next-hop node, it broadcasts a message to all neighbours to

exclude it from the routing path.

By using the topology-control approach, void nodes are

reduced or even eliminated without relying on any recovery

technique. Nevertheless, there are some limitations when using

DCR: i) the high cost of localisation by AUVs, ii) ignoring

the movement of nodes with the water current, iii) the impos-

sibility of ﬁnding connected nodes in the deeper holes, and

iv) the high delay caused by the topology control procedure.

4) Greedy Routing with Distributed Topology Control

(GR+DTC): In order to improve the robustness of DCR in

dealing with the void, GR+DTC [8] proposes a distributed al-

gorithm for topology control which can react to any change in

the network topology caused by nodes mobility. In GR+DTC,

each node knows its pressure and the location of all sinks,

the neighbouring nodes, and itself. Each node locally is able

to determine if it is in a void area and accordingly selects a

new depth value, if necessary. The network model (multi-sink

architecture) and routing model (greedy forwarding strategy)

is exactly the same as the one presented in DCR.

During the topology control, each void node initiates node

adjustment on a priority basis on its distance to the nearest

sink node. In this way, the nodes with the shorter distance to

the surface have shorter waiting time to decide about depth

adjustment. The key aspect of this prioritisation is that all

nodes with smaller distance to the surface already performed

their depth adjustment and other nodes can rely on them as

next-hop nodes, if there exists any path between them and any

sink. Whenever a void node adjusts its depth, it periodically

broadcasts an adjusting information message to detect whether

it is now connected to a non-void node. At the same time, non-

void nodes can respond by a reply message in order that void

node can stop its depth adjustment and set the responder as the

next hop node. Upon ﬁnding a next hop node, the void node

updates its status and broadcasts a beacon for all neighbours

to inform its connectivity to a sink node.

An important aspect of GR+DTC is that void nodes do not

have to wait for receiving the optimal location information

from the monitoring centre and they can react immediately

to any change in the topology. Furthermore, it is notable

that beacon dissemination is only performed by the connected

nodes (which have any available path to any sink) and not as

a periodic mechanism for all nodes. Thus, unnecessary energy

Communications Surveys & Tutorials

Fig. 18: Recovery mode in HydroCast [42]

consumption will be controlled by excluding the void nodes

from the periodic beaconing.

A serious weakness with this routing protocol, however, is

that it cannot deal with all kinds of void areas in UWSNs

(similar to the drawback of DCR) and also the greedy routing

is not equipped with an efﬁcient void-handling technique to

resolve this problem. Furthermore, node replacement is not

performed in an optimal way.

5) HydroCast: HydroCast [42] protocol is consisted of

two parts: greedy pressure-based routing algorithm and a

local lower-depth-ﬁrst recovery method. In HydroCast, each

node knows the pressure of itself and neighbouring nodes,

and two hop neighbouring distances. In the routing part,

HydroCast tries to select a subset of neighbouring nodes with

maximum greedy progress towards the destination taking also

into account the hidden terminal problem.

In the void-handling mode, local-maxima nodes and recov-

ery paths will be discovered in advance to bypass the void

areas during the packet forwarding. The main idea in this

technique is to identify stuck nodes by making use of the depth

properties of deployed nodes. In this scenario, each node is

able to determine if it is a void node or not, only by searching

for a neighbouring node with lower depth than itself. If it

cannot ﬁnd any neighbouring node with lower depth, the node

is counted as a void node.

In a proactive way, local maxima nodes try to discover a

recovery path to a node with lower pressure by using a ﬂooding

approach. The discovered node (with lower depth) may itself

be another void node which has a new recovery path to the

sink. In Fig. 18, for instance, LM1is a void node which has a

recovery path to another void node LM2. As can be seen, node

LM2is also connected to the node Swith another recovery

path.

The discovered path from a stuck node to a non-void node

with lower depth is stored in each local maxima node for

the future use. After applying this method, each void node

knows an alternative path to a non-void node or directly to a

sonobuoy on the surface, if any. After reaching a packet in a

void node, it exploits an opportunistic data forwarding over

recovery path to deliver packets to a non-void node or a new

void node which has connectivity to a destination.

In order to minimise the ﬂooding energy cost, HydroCast

uses 2D surface ﬂooding instead of 3D ﬂooding. This is due to

the fact that 3D ﬂooding can involve a large number of sensors

in the network; however, 2D ﬂooding is very manageable in

terms of the number of involved nodes. In this way, only the

nodes which are not dominated by the surface neighbouring

nodes are able to participate in the route discovery.

In HydroCast, the packet forwarding on the recovery path

is performed in an energy efﬁcient manner by suppressing

the duplicated transmissions. Void-handling technique used in

HydroCast is a loop-free technique which also guarantees the

packet delivery.

HydroCast only can deal with the void areas near to the

water surface. However, the void areas can appear in deeper

regions of the water which are not addressed in this protocol.

Furthermore, when there is a large bubble shape void area near

to the water surface, more nodes may be involved in the path

discovery, which wastes the energy. HydroCast also imposes

high communication overhead to obtain two-hop neighbouring

nodes information. Also, route discovery and maintenance

in the void-handling mode incurs high overhead. Each node

requires additional resources such as memory storage to record

the discovered path, especially when the recovery path is very

long. In terms of path optimality, 2D surface ﬂooding cannot

ensure the expected quality of discovered paths in all cases.

6) Void Aware Pressure Routing (VAPR): VAPR [43] is a

preventive and soft-state technique which keeps packets away

from the voids during the packet forwarding. In VAPR, each

node knows the pressure of itself and neighbouring nodes,

hop count information, and two hops neighbouring distances.

VAPR beneﬁts from enhanced beaconing and opportunistic di-

rectional data forwarding in order to handle the void problem.

In the beaconing phase, each beacon includes a sequence

number, hop count and depth information which are used to

determine the next hop direction (up or down) to reach the

closest sink on the surface. The reachability information is

propagated across the whole network by using the periodic

beaconing initiated by the sink nodes. The sequence number

is used to update nodes based on the most recent beacons.

Upon receiving a beacon from a neighbour, the receiving

node updates its data forwarding direction (DF dir) and hop

distance based to the closest available sink. If a beacon with

lower hop count is received from a shallower depth, data

forwarding direction should be set as up and otherwise down.

The receiving node also extracts the sender’s data forwarding

direction from the received beacon and sets it as the next-hop

data forwarding direction (NDF dir). Accordingly, each node

knows the direction of packet forwarding for only two hops.

Any direction change is a sign of the void area existence.

As can be seen in Fig. 19, when node breceives a packet

from node c, its DF dir and NDF dir are equal to Up-Up.

Thus, it only sends the packet to node abecause it is in

shallower depth than node band also node a’s DF dir (Up)

matches with that of NDF dir (Up). As can be seen, node x

is a trapped node which is ﬁltered out by node bbecause its

DF dir is set to down.

In this protocol, opportunistic data forwarding is only per-

formed based on the directional trails and not on the basis of

hop count values. VAPR can ﬁlter out the trapped nodes by

checking the next-hop data forwarding direction. If void ap-

Communications Surveys & Tutorials

Fig. 19: Directional data forwarding in VAPR [43]

Fig. 20: Problem in directional data forwarding in VAPR

pears in the routing path (e.g. observing any direction change),

the data forwarding direction of two hops is used to determine

the correct routing path. In facing a void area, the forwarding

node only considers the neighbouring nodes with a change

in the routing direction (up-down or down-up) as candidate

nodes. Generally, the forwarding direction is exactly equal to

the opposite direction of the beacon reception direction. After

selecting the candidate nodes, VAPR uses HydroCast approach

to select a forwarding set without hidden terminal problem

which can maximise expected packet advancement in upward

or downward direction (according to the selected direction).

VAPR can provide nodes with a partial view of the network

topology to diminish the impact of nodes blindness to the net-

work topology. By propagating surface reachability informa-

tion, there is no need to use any void recovery technique which

imposes an extra cost of route discovery and maintenance. The

strength of this technique is to guarantee the packet delivery by

using an opportunistic directional forwarding. VAPR also has

loop-free property in a static and dynamic 3D environment.

However, VAPR tries to bypass void areas by holding

information of up to two-hop neighbouring nodes which

impose high overhead to the system. Moreover, the beaconing

procedure in VAPR (for multi-sink architecture) is not properly

utilised in a way that beacons carry some useful information

in addition to the hop count. For this reason, each node in

VAPR is forced to periodically measure the distance to every

neighbour and broadcast the measured information to all other

one-hop neighbours.

As another problem, packet can only be forwarded up or

down depending on the selected direction which cannot utilise

subsets of nodes in the horizontal direction (including nodes

with lower depth and higher depth together in the forwarding

set). Because of this, in facing a convex void (similar to Fig.

20), packets will traverse longer distance because they cannot

be forwarded in a horizontal direction to bypass the void and

also the number of available candidates will be reduced which

can increase the packet failure probability. Thus, to bypass a

convex void region, data packets are routed along the longer

route leading to an increase in the energy consumption, if this

route is highly used.

7) Weighting Depth and Forwarding Area Division DBR

(WDFAD-DBR): WDFAD-DBR [63] is a pressure-based rout-

ing protocol in which void nodes can take themselves out of

the packet forwarding set to provide the opportunity for other

available candidate nodes. In this protocol, each node knows

its depth, the depth difference of two hops, and neighbouring

distances. In WDFAD-DBR, the forwarding area is divided

into a primary forwarding area (Reuleaux triangle) and two

auxiliary forwarding areas. The primary forwarding area is

constant all the time, but auxiliary forwarding areas may

adaptively expand based on the node density and channel

quality. This feature is useful to suppress the duplicate packets

in a dense topology or increase the chance of packet delivery

in a sparse network.

WDFAD-DBR also tries to estimate the relative position of

neighbouring nodes with the aid of nodes movement pattern

and speed, which then assists the underwater nodes to ﬁnd out

whether a neighbouring node is still located in the transmission

range or not. The position prediction mechanism also aids

to prolong the network lifetime by increasing the updating

request interval. To address the void issue, WDFAD-DBR

follows a preventive void-handling technique, which can avoid

void nodes in advance by considering two hops information

and also suppressing the packets in the void nodes. By

considering the depth difference of two hops, the chance of

encountering a void node is reduced.

As shown in Fig. 21, node Sis a sending node and node A

and node Bare two forwarding candidate nodes because they

are located above the sending node. In DBR strategy, node

Awill ﬁrst forward the packet having the lowest depth. The

packet transmitted by node Awill suppress the transmission

of other neighbours like node B. However, after forwarding

the packet by node A, there exist no nodes above node Ato

continue the packet forwarding leading to void communication

issue. This case indicates the weakness of greedy strategies to

fall in a local optimal solution. In WDFAD-DBR, however,

node Bis selected to forward the packet rather than node A.

This is due to the fact that the depth difference between node

Communications Surveys & Tutorials

Fig. 21: Void handling in WDFAD-DBR

Eand Bis also considered in addition to the depth difference

between node Sand B. Moreover, node Adrops the receiving

packet after realising that there is no node above itself. This

strategy efﬁciently decreases the probability of packet loss due

to the void problem.

However, the issue with WDFAD-DBR is that it cannot

identify the trapped nodes in advance. Trapped nodes may

lead a forwarding packet to a void node. Thus a forwarding

packet does not have any chance to bypass a convex void area.

Moreover, in WDFAD-DBR, the packet is dropped if a void

or trapped node senses an event and intends to send a packet

toward the sink node. This is due to the fact that the forwarding

direction in WDFAD-DBR is considered only upward which

makes it impossible to consider a void or trapped node as

a source node. As another problem, it needs to periodically

update the neighbouring information resulting in generating

more control packets and corresponding Acknowledgement

Packets (ACKs).

VI. QUA LI TATIV E COMPARI SO NS

In this section, we compare all existing void-handling

techniques in terms of quality to evaluate their effectiveness

to deal with the void problem. Our comparison is founded

on some criteria presented in Section III-D including ac-

tiveness, opportunistic forwarding, states, guaranteed delivery,

path optimality, end-to-end delay, communication overhead,

scalability, and energy efﬁciency.

Table IV shows how different void-handling techniques are

located under different categories while also highlighting their

signiﬁcant features. The majority of these techniques have

already been evaluated with the aid of network simulators or

using a real testbed. Obviously, our analysis is also consistent

with those reported in the literature. To the best of our

knowledge, this is the ﬁrst research study which compares all

void-handling techniques from different categories altogether.

Before comparison, it should be mentioned that all void-

handling techniques have their own advantages and disad-

vantages. Since the metrics are not fully independent from

each other, sometimes improving one metric (e.g. guaranteed

delivery) may adversely affect another metric (e.g. lower

complexity and cost). Thus, in order to obtain the maximum

efﬁciency, some key points such as the environmental char-

acteristics, intended application, and unique characteristics of

the routing protocol should be considered when designing a

new void-handling technique or selecting an existing one [20],

[106].

Furthermore, note that we only evaluate the void-handling

technique proposed for each protocol independently regardless

of their routing algorithms. As another important point, some

void-handling techniques are very similar when qualitatively

compared; however, they may perform differently when all

details are applied in quantitative studies.

A. Activeness

This feature indicates whether a void-handling technique is

able to handle void communications reactively (on demand) or

proactively (with a previous plan). Also, some approaches have

no recovery method which requires them to use a preventative

technique.

As can be seen in Table IV, most of the underwater void-

handling techniques use a preventative approach. This is due

to the fact that in UWSNs, reactive and proactive techniques

often impose high communication overhead to the network [5],

[107]. For instance, VBVA and RMTG are reactive techniques

which are only activated when a packet is stuck at a void

node. However, they should tolerate high cost to be able to

recover packets from the void area. Nonetheless, some of

the preventative techniques still suffer from the packet loss

because they cannot efﬁciently bypass a void area. In order to

increase robustness, DFR exploits a hybrid technique including

a preventative (ﬂooding zone adjustment) and reactive (ﬁnding

detour path) approach. HydroCast is the only technique which

uses a proactive technique to hold the recovery path from

the void nodes to the non-void nodes. However, permanently

keeping a recovery path at each node is very costly [108].

In general, preventative techniques can be less costly in

terms of resource consumption when the topology changes

are slow and a large number of control packets is not required

to ﬁnd the detour paths.

B. Opportunistic Forwarding

This metric speciﬁes whether a void-handling technique

exploits a subset of neighbouring nodes to relay a packet in

order to increase the transmission reliability [109], [110]. The

opportunistic data forwarding is usually neglected in some

void-handling techniques.

For instance, DFR uses opportunistic data forwarding in

the routing algorithm with a non-opportunistic void-handling

technique. DFR void-handling technique will be satisﬁed just

with placing one node in the ﬂooding zone or only by ﬁnding

one of the best neighbouring nodes in the search for the detour

path. This feature is in contradiction with the opportunistic

nature. DCR, GR+DTC, RMTG and LLSR are also not oppor-

tunistic neither in routing nor in the void-handling technique.

For instance, DCR and GR+DTC will be satisﬁed if only one

Communications Surveys & Tutorials

TABLE IV: Characteristics of void-handling techniques.

Protocol Void-handling

Technique Activeness Opportunistic

Forwarding States Guaranteed

Delivery

Path

Opti-

mality

Delay Overhead Scalability

Energy

Efﬁ-

ciency

VBVA [29] Vector-shift &

back-pressure Reactive Yes Stateless No No High High Medium Low

AHH-VBF

[30]

Transmission

power adjustment

& Pipeline’s

radius adjustment

Preventative Yes Stateless No No Low Low High High

DFR [54]

Flooding zone

adjustment &

ﬁnding detour

path

Hybrid No Stateless No No Medium Low High High

FBR [57]

Transmission

power adjustment

& cone shifting

Preventative Yes Stateless No No High High High High

RMTG

[103]

Covering area

around target

region &

Backward-

forwarding

Reactive No Stateful No No High High Low Low

Mobicast

[7]

Covering area

around target

region

Preventative Yes Stateless No No Low Medium High High

LLSR [64] Beaconing initi-

ated by sink Preventative No Soft-state Yes Near-

optimal Low Medium Medium Medium

IVAR [72] Beaconing initi-

ated by sink Preventative Yes Soft-state Yes Near-

optimal Low Medium Medium Medium

OVAR [37] Beaconing initi-

ated by sink Preventative Yes Soft-state Yes Near-

optimal Low Medium Medium High

DCR [66] Network

topology control Preventative No Stateless No No High High Low Low

GR+DTC

[8]

Network

topology control Preventative No Stateless No No High Medium High Medium

Hydrocast

[42]

Local lower

depth-ﬁrst

recovery

Proactive Yes Partial-path

state No No Medium High Low Medium

VAPR [43] Beaconing initi-

ated by sinks Preventative Yes Soft-state Yes Near-

optimal Low Medium Medium High

WDFAD-

DBR [63]

Passive participa-

tion Preventative Yes Stateless No No Low Medium High High

Communications Surveys & Tutorials

node covers the void area which can make it a bottleneck

node in the routing path. Other techniques listed in the table

are opportunistic; however, some of them suffer from the

hidden terminal problem [37], [42]. Only OVAR, HydroCast

and VAPR remove the hidden nodes from their forwarding

set and therefore have no duplicated packet transmissions.

WDFAD-DBR also suppress all duplicated packets by using

a Reuleaux triangle as the primary forwarding area, but it

may intentionally let to have duplicated packets in the sparse

density by including the auxiliary forwarding areas.

In some void-handling techniques like AHH-VBF, DFR,

OVAR, and WDFAD-DBR, the number of candidate nodes

is adjusted based on the density of area ahead which makes

them more suitable to be used in UWSNs in terms of energy

and reliability tradeoff.

C. States

It is desirable to hold fewer states at each node in order to

increase the scalability and performance of the void-handling

technique [20], [111]. The majority of void-handling tech-

niques in UWSNs are stateless. This means that each node

only holds the states of one hop or up to two hops [5].

RMTG is the only stateful approach, since a path is es-

tablished between the source and geocast region and also

root node holds the states of all nodes in the geocast region.

HydroCast is considered as a partial-path state approach which

needs to maintain states with a partial path from a void node to

a non-void node or ﬁnal destination. LLSR, IVAR, OVAR and

VAPR are considered as soft-state models, because they rely

on the beaconing information which is useful for the routing

efﬁciency, but not essential, as it can be updated or replaced

if needed. They do not maintain a path toward the sink, but

they use reachability information to decide about the routing

path [43], [72].

According to the speed of the nodes and topology changes,

the appropriate approach should be selected. The stateful, soft-

state, and stateless approaches are suitable for stationary, semi-

dynamic, and high-dynamic topologies, respectively.

D. Guaranteed Delivery

We measure this feature when a void node is able to deliver

a packet to the destination if there exists a valid topological

path or simply valid path between them. It is proved that no

local and memoryless technique exists to guarantee the packet

delivery in 3D networks [52].

Nevertheless, LLSR, IVAR, OVAR and VAPR by using the

reachability information, are able to bypass all kinds of voids.

Although VBVA also has a solution for both kinds of voids

including convex and concave, this technique is not loop-free.

Back pressure in VBVA may send the packet to a non-void

node and resume vector-based forwarding, but packet again

can be stuck in another void node which has no valid path

toward the sink except the previous traversed non-void node.

On the other hand, the main goal of DCR and GR+DTC is to

change the network topology in a way that all void nodes

are moved to a non-void position. When there is no void

node, in theory, the packet can be delivered by each node

toward the destination [21], [112]. However, these techniques

cannot eliminate all void nodes just by changing their depth,

so they cannot deliver the packet in some cases. FBR and

HydroCast only can guarantee the packet delivery if they are

not faced with any concave void. Mobicast is successful in

packet delivery if there is a valid path started from a node

inside the ”hold and forward zone”, otherwise it fails. DFR

and RMTG are not loop-free techniques and AHH-VBF is not

able to change its pipeline direction if it is required, so they

may fail in the operation. WDFAD-DBR also ﬁnds no path

toward the surface if the only valid topological path is via a

void node.

It can be observed that only those techniques which have

used the hop count distance information for the packet for-

warding, are successful in bypassing all the void areas.

E. Path Optimality

Path optimality shows whether the traversed distance by

a packet is close to the optimal path which is expressed

as the length of straight line between the source node and

destination or not [113]. In general, none of the existing void-

handling techniques can ensure that an optimal path is always

discovered.

Nevertheless, LLSR, IVAR, OVAR and VAPR are able

to ﬁnd a near-optimal path when the updated reachability

information is properly supplied. Note that packets are not

forwarded on an optimal path when VAPR confronts a wide

convex void area, as shown in Fig. 20. Although other tech-

niques are unable to ﬁnd the optimal path, some of them may

have other advantages such as simplicity of implementation,

reliability of the proposed path, and also resource management

efﬁciency.

As can be observed, in a 3D environment, ﬁnding a near-

optimal path depends on the nodes viewpoint of the network

topology [43], [114]. Thus, the soft-state techniques may route

the packets in a shorter path.

F. End-to-end Delay

This metric shows the time required to deliver a packet from

a source node to the sink [115]. The void-handling technique

used by VBVA imposes high delay because it lets the packets

be stuck in a void node and then tries to recover them using

a time-consuming procedure. FBR also has a high end-to-end

delay because it sends and receives the control packets in each

hop to establish a connection. Although DFR tries to decrease

the delay by using a preventative approach at the ﬁrst step,

the recovery technique may be required, which causes more

delay. RMTG has a high delay because of using a reactive

void-handling technique and also need to the route discovery

between the source node and geocast region. Mobicast has less

delay by using a preventative approach. Sof-state techniques

such as LLSR, IVAR, OVAR, and VAPR have low latency by

forwarding packets in a near-optimal path. DCR and GR-DTC

are not suitable for the delay-sensitive applications due to the

high delay caused by the topology control procedure. Although

Hydrocast uses the predeﬁned recovery paths to decrease the

delay, it lets the packets be stuck in a void node which is still

Communications Surveys & Tutorials

time-consuming. AHH-VBF and WDFAD-DBR have lower

delay by using a preventative approach and without any need

to perform a recovery phase.

It can be concluded that the reactive techniques impose a

higher delay, by letting the packet being stuck in a void node at

the ﬁrst stage [116]. Among the preventative techniques, those

are performed with a lower delay, which collect the required

routeing information before the forwarding phase.

G. Communication Overhead

This metric shows the amount of communication overhead

is imposed when handling a void. Generally, the void-handling

techniques in the underwater environment have higher over-

head in comparison to the terrestrial void-handling techniques

because of the dynamic nature of UWSNs [1], [117].

VBVA has high overhead due to a large number of control

packets generated in a chaotic manner. DFR and AHH-VBF

can adjust their forwarding zone hop-by-hop which enables

them to control the protocol overhead to some extent. Thus,

these approaches are considered as the low overhead tech-

niques in UWSNs. FBR can be considered as a high overhead

technique, due to sending and receiving Request To Send

(RTS) and Clear to Send (CTS) in each forwarding node

and also for each power level setting. DCR needs Depth-

ﬁrst Search to identify all void nodes, and also an AUV

should gather information from relay nodes and then inform

them about their new positions which make this technique a

high overhead approach. However, GR+DTC decreases this

overhead by using a distributed approach in which each node

individually starts their depth adjustment initiated from the

shallowest depth. In geocasting, RMTG uses a large number of

control packets, while Mobicast relies on its estimation about

the covering area around the geocast region and does not need

to exchange control packets. LLSR, IVAR, OVAR and VAPR

are not considered as high overhead techniques. Although

beaconing technique inherently imposes communication over-

head, it can be applied over long intervals due to the fact that

nodes move slowly with water current [37], [43]. Moreover,

the communication overhead in the beaconing can be justiﬁed

against localisation overhead in the location-based techniques

[96], [97]. WDFAD-DBR incurs low overhead during the

routing phase but the updating phase for obtaining one-hop

information still imposes overhead to the protocol. Finally,

HydroCast is categorised as a high overhead technique because

of its proactive approach and also recovery path discovery (2D

ﬂooding) and maintenance. Generally, the stateful, reactive,

and heuristic techniques using multiple control packets, are

prone to having the higher overhead [118].

H. Scalability

This feature shows that the performance of each void-

handling technique is not affected by increasing the number

of void nodes [20]. DCR is not scalable to the number of

void nodes, since it follows a centralised approach; however,

GR+DTC obtains more scalability by performing a localised

approach for depth adjustment. RMTG is not scalable because

the routing path between the source and geocast root should

be held which is costly in a 3D dynamic environment. The

opportunistic data forwarding in HydroCast is scalable, but

its 2D surface ﬂooding is not scalable because every void

node should search and ﬁnd a recovery path. Note that

recovery paths are actually not used most of the time. Soft-

state protocols such as LLSR, IVAR, OVAR, and VAPR are

scalable, but not as well as the stateless routing protocols.

Although VBVA uses a stateless approach, its scalability is not

high because of using a reactive approach. Other void-handling

techniques can be considered as high scalable because they use

a stateless and preventative approach. In whole, the majority

of stateless and distributed void-handling techniques are more

scalable than stateful, soft-state, and centralised approaches

[53], [106].

I. Energy Efﬁciency

This metric reﬂects how void-handling technique is energy-

efﬁcient by considering all inﬂuential factors such as number

of transmissions, communication overhead, involving nodes,

hidden terminal problem and so forth [87], [119], [120].

VBVA is not energy efﬁcient due to a large number of

generated control packets. DFR, FBR, and AHH-VBF exploit

a forwarding zone which can prevent packets to be ﬂooded in

the unnecessary areas of the network. FBR and AHH-VBF

also have the ability to adjust their transmission ranges to

further control the energy consumption. Vertical movement of

nodes is an energy-consuming task which only can be justiﬁed

when used less frequently [8]. DCR is not considered as an

energy efﬁcient technique for exchanging a large number of

control packets with monitoring centre; however, GR+DTC

can diminish this energy dissipation by reducing the number

of control packets. RMTG does not consider the energy issue

in its geocasting, while Mobicast tries to precisely estimate

the covering area to wake up only the required nodes at the

right times.

Beaconing-based techniques are able to compensate beacon-

ing energy consumption by traversing the optimal path in the

routing (reducing the number of transmissions), and also this

cost can be justiﬁed by considering the localisation energy cost

in the location-based techniques. VAPR and OVAR are more

energy efﬁcient among them because of taking advantages of

an opportunistic data forwarding which efﬁciently addresses

the hidden terminal problem. WDFAD-DBR also resolves the

duplicated packets in a dense network which turns it into an

energy-efﬁcient routing protocol. Flooding techniques usually

consume high energy; nonetheless, HydroCast controls energy

consumption by utilising a 2D ﬂooding instead of 3D ﬂooding

and also uses opportunistic data forwarding on the recovery

path to suppress duplicated packets. Overall, the void-handling

techniques with lower communication overhead, power level

adjustment, and resizeable forwarding area based on the net-

work density, have achieved a higher energy efﬁciency.

VII. OPE N RES EA RC H CHA LL EN GES

From all discussions in the preceding sections, it seems that

void-handling techniques have received great attention when

designing efﬁcient routing protocols for UWSNs. However,

Communications Surveys & Tutorials

some technical, practical, and environmental issues are still

remained for further investigation. Based on the literature

surveyed above, we present the following potential directions

that can be considered in the void-handling techniques.

First, hybrid void-handling techniques should be devised

based on the new deployment environment. It would be inter-

esting to know what kinds of characteristics can be inherited

from the previous techniques and which of them should be

devised based on the new deployment environment. Thus, with

previous knowledge of void-handling techniques presented in

the literature, it becomes increasingly clear why a hybrid

approach in UWSNs is vital.

Second, the majority of current void-handling techniques

are designed for the shallow waters with a limited number of

void areas. However, it is interesting to study whether these

techniques are still effective in a deep multi-holes environment

under varying pressure, temperature, and salinity. In some

cases, very speciﬁc solutions are required to deal with multi-

holes in UWSNs [28].

Third, the trapped nodes issue is deserved to receive higher

attention in void-handling techniques due to its direct impact

on the performance of void-handling techniques. As an efﬁ-

cient solution, a void-handling technique should proactively

discover the trapped nodes in a preprocessing phase and avoid

them during the packet forwarding.

Fourth, to our best knowledge, the impact of nodes move-

ment on the void area have not been investigated thoroughly in

the literature. The void area is continuously reshaped or move

with the water current [30]. The void-handling techniques also

suffer from lack of a realistic model for node mobility. Most

of the existing protocols assume that nodes are mobile at a low

rate or they are stationary. Therefore, investigating the impact

of node movement on the void-handling techniques seems to

be a challenging issue.

Fifth, designing the void-handling techniques with a cross-

layer view can enhance the performance of the routing proto-

cols during packet delivery. The existing void-handling tech-

niques have only focused on the network layer. However, with

a cross-layer design, the number of collisions can be managed

more efﬁciently over the MAC layer, while the results of some

tasks, such as beaconing, can be shared between layers. [57],

[121].

Sixth, dealing with a void area within a geocast region is still

a challenging issue. The existing model involves many relay

nodes to cover the geocast region with a larger area. However,

it is necessary to design the new void-handling techniques to

further decrease the number of involving nodes.

Finally, some existing void-handling techniques have been

proposed under unrealistic assumptions about the underwater

environment (e.g. availability of precise full coordinates in-

formation, noise-free environment, etc). Thus, conducting a

realistic study of the proposed void-handling techniques using

a real testbed can easily enlighten their weakness as well as

their strengths.

VIII. CON CL US IO N

In this paper, we investigated the state of the art of void-

handling techniques in UWSNs. First of all, we discussed the

different features of void communications in the terrestrial and

underwater environments and mentioned the unique challenges

of designing void-handling techniques in UWSNs. Afterwards,

the main features for designing efﬁcient void-handling tech-

niques have been introduced. To facilitate comparison of

different techniques, we classiﬁed the current void-handling

techniques into two main categories of location-based and

depth-based techniques. For each category, all existing void-

handling techniques have individually been explored in detail.

Then, a comprehensive comparison of the currently available

techniques has been proposed. It is shown that each void-

handling technique is designed for a speciﬁc environment

which has its own strengths as well as its constraints. Finally,

some open research challenges are mentioned to deal with the

void problem.

REF ER EN CE S

[1] J. Heidemann, W. Ye, J. Wills, A. Syed, and Y. Li, “Research challenges

and applications for underwater sensor networking,” in Wireless Com-

munications and Networking Conference, Las Vegas, USA, pp. 228–

235, IEEE, 2006.

[2] J. Partan, J. Kurose, and B. N. Levine, “A survey of practical issues

in underwater networks,” ACM SIGMOBILE Mobile Computing and

Communications Review, vol. 11, no. 4, pp. 23–33, 2007.

[3] I. F. Akyildiz, D. Pompili, and T. Melodia, “Challenges for efﬁcient

communication in underwater acoustic sensor networks,” ACM Sigbed

Review, vol. 1, no. 2, pp. 3–8, 2004.

[4] M. R. Arshad, “Recent advancement in sensor technology for under-

water applications,” Indian Journal of Marine Sciences, vol. 38, no. 3,

pp. 267–273, 2009.

[5] M. Ayaz, I. Baig, A. Abdullah, and I. Faye, “A survey on routing tech-

niques in underwater wireless sensor networks,” Journal of Network

and Computer Applications, vol. 34, no. 6, pp. 1908–1927, 2011.

[6] C. Giantsis and A. A. Economides, “Comparison of routing protocols

for underwater sensor networks: a survey,” International Journal of

Communication Networks and Distributed Systems, vol. 7, no. 3-4,

pp. 192–228, 2011.

[7] Y.-S. Chen and Y.-W. Lin, “Mobicast routing protocol for underwater

sensor networks,” IEEE Sensors Journal, vol. 13, no. 2, pp. 737–749,

2013.

[8] R. W. Coutinho, A. Boukerche, L. F. Vieira, and A. A. Loureiro, “A

novel void node recovery paradigm for long-term underwater sensor

networks,” Ad Hoc Networks, vol. 34, pp. 144–156, 2015.

[9] M. T. Kheirabadi and M. M. Mohamad, “Greedy routing in under-

water acoustic sensor networks: a survey,” International Journal of

Distributed Sensor Networks, vol. 2013, Article ID 701834, 21 pages,

2013. doi:10.1155/2013/701834.

[10] N. Li, J.-F. Mart´

ınez, J. M. Meneses Chaus, and M. Eckert, “A survey

on underwater acoustic sensor network routing protocols,” Sensors,

vol. 16, no. 3, p. 414, 2016, doi:10.3390/s16030414.

[11] G. Han, J. Jiang, N. Bao, L. Wan, and M. Guizani, “Routing proto-

cols for underwater wireless sensor networks,” IEEE Communications

Magazine, vol. 53, no. 11, pp. 72–78, 2015.

[12] I. Stojmenovic, “Position-based routing in ad hoc networks,” IEEE

communications magazine, vol. 40, no. 7, pp. 128–134, 2002.

[13] M. Mauve, J. Widmer, and H. Hartenstein, “A survey on position-based

routing in mobile ad hoc networks,” IEEE network, vol. 15, no. 6,

pp. 30–39, 2001.

[14] X. Hong, K. Xu, and M. Gerla, “Scalable routing protocols for mobile

ad hoc networks,” IEEE network, vol. 16, no. 4, pp. 11–21, 2002.

[15] S.-J. Lee, M. Gerla, and C.-K. Toh, “A simulation study of table-driven

and on-demand routing protocols for mobile ad hoc networks,” IEEE

network, vol. 13, no. 4, pp. 48–54, 1999.

[16] S. Mohseni, R. Hassan, A. Patel, and R. Razali, “Comparative review

study of reactive and proactive routing protocols in manets,” in 4th

International Conference on Digital Ecosystems and Technologies,

Dubai, UAE, pp. 304–309, IEEE, 2010.

[17] K. U. R. Khan, R. U. Zaman, and A. V. Reddy, “Performance com-

parison of on-demand and table driven ad hoc routing protocols using

nctuns,” in Tenth International Conference on Computer Modeling and

Simulation, Washington, DC, USA, pp. 336–341, IEEE, 2008.

Communications Surveys & Tutorials

[18] S. Souiki, M. Feham, M. Feham, and N. Labraoui, “Geographic

routing protocols for underwater wireless sensor networks: a survey,”

International Journal of Wireless and Mobile Networks, vol. 6, no. 1,

pp. 69–87, 2014.

[19] C. Maihofer, “A survey of geocast routing protocols,” IEEE Commu-

nications Surveys and Tutorials, vol. 6, no. 2, pp. 32–42, 2004.

[20] D. Chen and P. K. Varshney, “A survey of void handling techniques

for geographic routing in wireless networks,” IEEE Communications

Surveys and Tutorials, vol. 9, no. 1, pp. 50–67, 2007.

[21] W.-J. Liu and K.-T. Feng, “Greedy routing with anti-void traversal for

wireless sensor networks,” IEEE Transactions on Mobile Computing,

vol. 8, no. 7, pp. 910–922, 2009.

[22] S. Mitra and A. Roy, “Communication void free routing protocol in

wireless sensor network,” Wireless Personal Communications, vol. 82,

no. 4, pp. 2567–2581, 2015.

[23] S. Basagni, C. Petrioli, R. Petroccia, and D. Spaccini, “Carp: a channel-

aware routing protocol for underwater acoustic wireless networks,” Ad

Hoc Networks, vol. 34, pp. 92–104, 2015.

[24] M. Younis and K. Akkaya, “Strategies and techniques for node place-

ment in wireless sensor networks: A survey,” Ad Hoc Networks, vol. 6,

no. 4, pp. 621–655, 2008.

[25] M. Aissani, A. Mellouk, N. Badache, and M. Boumaza, “A novel

approach for void avoidance in wireless sensor networks,” International

Journal of Communication Systems, vol. 23, no. 8, pp. 945–962, 2010.

[26] X. Fan and F. Du, “An efﬁcient bypassing void routing algorithm for

wireless sensor network,” Journal of Sensors, vol. 2015, Article ID

686809, 9 pages, 2015. doi:10.1155/2015/686809.

[27] Q. Fang, J. Gao, and L. J. Guibas, “Locating and bypassing holes in

sensor networks,” Mobile Networks and Applications, vol. 11, no. 2,

pp. 187–200, 2006.

[28] C.-Y. Chang, C.-T. Chang, Y.-C. Chen, and S.-C. Lee, “Active route-

guiding protocols for resisting obstacles in wireless sensor networks,”

IEEE Transactions on Vehicular Technology, vol. 59, no. 9, pp. 4425–

4442, 2010.

[29] P. Xie, Z. Zhou, Z. Peng, J.-H. Cui, and Z. Shi, “Void avoidance in

three-dimensional mobile underwater sensor networks,” in 4th Interna-

tional Conference in Wireless Algorithms, Systems, and Applications,

Boston, MA, USA, pp. 305–314, Springer, 2009.

[30] H. Yu, N. Yao, and J. Liu, “An adaptive routing protocol in underwater

sparse acoustic sensor networks,” Ad Hoc Networks, vol. 34, pp. 121–

143, 2015.

[31] S. M. Ghoreyshi, A. Shahrabi, and T. Boutaleb, “An opportunistic void

avoidance routing protocol for underwater sensor networks,” in 30th

International Conference on Advanced Information Networking and

Applications, Crans-Montana, Switzerland, pp. 316–323, IEEE, 2016.

[32] G. E. Burrowes and J. Y. Khan, “Investigation of a short-range un-

derwater acoustic communication channel for mac protocol design,” in

4th International Conference on Signal Processing and Communication

Systems, Gold Coast, Australia, pp. 1–8, IEEE, 2010.

[33] K. Chen, M. Ma, E. Cheng, F. Yuan, and W. Su, “A survey on mac

protocols for underwater wireless sensor networks,” IEEE Communi-

cations Surveys and Tutorials, vol. 16, no. 3, pp. 1433–1447, 2014.

[34] P. Casari and M. Zorzi, “Protocol design issues in underwater acoustic

networks,” Computer Communications, vol. 34, no. 17, pp. 2013–2025,

2011.

[35] M. Stojanovic, “On the relationship between capacity and distance in

an underwater acoustic communication channel,” ACM SIGMOBILE

Mobile Computing and Communications Review, vol. 11, no. 4, pp. 34–

43, 2007.

[36] J. Preisig, “Acoustic propagation considerations for underwater acoustic

communications network development,” ACM SIGMOBILE Mobile

Computing and Communications Review, vol. 11, no. 4, pp. 2–10, 2007.

[37] S. M. Ghoreyshi, A. Shahrabi, and T. Boutaleb, “A novel cooperative

opportunistic routing scheme for underwater sensor networks,” Sensors,

vol. 16, no. 3, p. 297, 2016.

[38] V. Rodoplu and M. K. Park, “An energy-efﬁcient mac protocol for

underwater wireless acoustic networks,” in Proceedings of MTS/IEEE

OCEANS, Washington, DC, USA, pp. 1198–1203, IEEE, 2005.

[39] T. Melodia, D. Pompili, and I. F. Akyildiz, “Optimal local topology

knowledge for energy efﬁcient geographical routing in sensor net-

works,” in Twenty-third Annual Joint Conference of the IEEE Computer

and Communications Societies, INFOCOM, Hong Kong, PR China,

pp. 1705–1716, IEEE, 2004.

[40] S. Chen, G. Fan, and J.-H. Cui, “Avoid’void’in geographic routing for

data aggregation in sensor networks,” International Journal of Ad Hoc

and Ubiquitous Computing, vol. 1, no. 4, pp. 169–178, 2006.

[41] S. Parvin, M. A. Sarram, G. Mirjalily, and F. Adibnia, “A survey on

void handling techniques for geographic routing in vanet network.,”

International Journal of Grid and Distributed Computing, vol. 8, no. 2,

pp. 101–114, 2015.

[42] U. Lee, P. Wang, Y. Noh, F. Vieira, M. Gerla, and J.-H. Cui, “Pressure

routing for underwater sensor networks,” in 29th conference on In-

formation communications (INFOCOM), San Diego, California, USA,

pp. 1676–1684, 2010.

[43] Y. Noh, U. Lee, P. Wang, B. S. C. Choi, and M. Gerla, “Vapr:

void-aware pressure routing for underwater sensor networks,” IEEE

Transactions on Mobile Computing, vol. 12, no. 5, pp. 895–908, 2013.

[44] S. T. Nguyen, E. Cayirci, L. Yan, and C. Rong, “A shadow zone

aware routing protocol for acoustic underwater sensor networks,” IEEE

Communications Letters, vol. 13, no. 5, pp. 366–368, 2009.

[45] J. Jiang, G. Han, H. Guo, L. Shu, and J. J. Rodrigues, “Geographic

multipath routing based on geospatial division in duty-cycled under-

water wireless sensor networks,” Journal of Network and Computer

Applications, vol. 59, pp. 4–13, 2016.

[46] P. Xie, J.-H. Cui, and L. Lao, “Vbf: vector-based forwarding protocol

for underwater sensor networks,” in Proceedings of IFIP Networking,

Coimbra, Portugal, pp. 1216–1221, Springer, 2006.

[47] N. Nicolaou, A. See, P. Xie, J.-H. Cui, and D. Maggiorini, “Improving

the robustness of location-based routing for underwater sensor net-

works,” in Proceedings of OCEANS, pp. 1–6, IEEE, 2007.

[48] Z. Li, N. Yao, and Q. Gao, “Relative distance based forwarding

protocol for underwater wireless networks,” International Journal of

Distributed Sensor Networks, vol. 2014, Article ID 173089, 11 pages,

2014. doi:10.1155/2014/173089.

[49] H. Yan, Z. J. Shi, and J.-H. Cui, “Dbr: depth-based routing for under-

water sensor networks,” in Proceedings of the 7th international IFIP-

TC6 networking conference on adhoc and sensor networks, wireless

networks, next generation internet. Singapore, pp. 72–86, 2008.

[50] L. Guangzhong and L. Zhibin, “Depth-based multi-hop routing protocol

for underwater sensor network,” in 2nd International Conference on

Industrial Mechatronics and Automation (ICIMA), vol. 2, pp. 268–270,

IEEE, 2010.

[51] A. Wahid and D. Kim, “An energy efﬁcient localization-free routing

protocol for underwater wireless sensor networks,” International Jour-

nal of Distributed Sensor Networks, vol. 2012, Article ID 307246, 11

pages, 2012. doi:10.1155/2012/307246.

[52] S. Durocher, D. Kirkpatrick, and L. Narayanan, “On routing with

guaranteed delivery in three-dimensional ad hoc wireless networks,”

Wireless Networks, vol. 16, no. 1, pp. 227–235, 2010.

[53] J. N. Al-Karaki and A. E. Kamal, “Routing techniques in wireless

sensor networks: a survey,” IEEE Wireless communications, vol. 11,

no. 6, pp. 6–28, 2004.

[54] D. Shin, D. Hwang, and D. Kim, “Dfr: an efﬁcient directional ﬂooding-

based routing protocol in underwater sensor networks,” Wireless Com-

munications and Mobile Computing, vol. 12, no. 17, pp. 1517–1527,

2012.

[55] R. Otnes, P. A. van Walree, H. Buen, and H. Song, “Underwater

acoustic network simulation with lookup tables from physical-layer

replay,” IEEE Journal of Oceanic Engineering, vol. 40, no. 4, pp. 822–

840, 2015.

[56] E. Isuﬁ, H. Dol, and G. Leus, “Advanced ﬂooding-based routing proto-

cols for underwater sensor networks,” EURASIP Journal on Advances

in Signal Processing, vol. 2016, no. 1, pp. 1–12, 2016.

[57] J. M. Jornet, M. Stojanovic, and M. Zorzi, “Focused beam routing

protocol for underwater acoustic networks,” in Proceedings of the third

ACM international workshop on Underwater Networks, San Francisco,

California, USA, pp. 75–82, ACM, 2008.

[58] R. Otnes, A. Asterjadhi, P. Casari, M. Goetz, T. Husøy, I. Nissen,

K. Rimstad, P. van Walree, and M. Zorzi, “Underwater acoustic

networking techniques,” springerbriefs in electrical and computer en-

gineering, 2012, doi:10.1007/978-3-642-25224-2-1.”

[59] T. He, J. A. Stankovic, C. Lu, and T. Abdelzaher, “Speed: A stateless

protocol for real-time communication in sensor networks,” in 23rd In-

ternational Conference on Distributed Computing Systems, Providence,

RI, United States, pp. 46–55, IEEE, 2003.

[60] J. M. Jornet and M. Stojanovic, “Distributed power control for under-

water acoustic networks,” in Oceans, Quebec, Canada, pp. 1–7, IEEE,

2008.

[61] M. Al-Bzoor, Y. Zhu, J. Liu, R. Ammar, J.-H. Cui, and S. Rajasekaran,

“An adaptive power controlled routing protocol for underwater sensor

network,” International Journal of Sensor Networks, vol. 18, no. 3-4,

pp. 238–249, 2015.

Communications Surveys & Tutorials

[62] H. Takagi and L. Kleinrock, “Optimal transmission ranges for randomly

distributed packet radio terminals,” IEEE Transactions on Communi-

cations, vol. 32, no. 3, pp. 246–257, 1984.

[63] H. Yu, N. Yao, T. Wang, G. Li, Z. Gao, and G. Tan, “Wdfad-dbr:

Weighting depth and forwarding area division dbr routing protocol for

uasns,” Ad Hoc Networks, vol. 37, pp. 256–282, 2016.

[64] M. Barbeau, S. Blouin, G. Cervera, J. Garcia-Alfaro, and E. Kranakis,

“Location-free link state routing for underwater acoustic sensor net-

works,” in 28th Canadian Conference on Electrical and Computer

Engineering (CCECE), Halifax, Canada, pp. 1544–1549, IEEE, 2015.

[65] T. Hu and Y. Fei, “Qelar: a machine-learning-based adaptive routing

protocol for energy-efﬁcient and lifetime-extended underwater sensor

networks,” IEEE Transactions on Mobile Computing, vol. 9, no. 6,

pp. 796–809, 2010.

[66] R. W. Coutinho, L. F. Vieira, A. Loureiro, et al., “Dcr: Depth-controlled

routing protocol for underwater sensor networks,” in IEEE Symposium

on Computers and Communications, Split, Croatia, pp. 453–458, IEEE,

2013.

[67] R. W. Coutinho, A. Boukerche, L. F. Vieira, and A. A. Loureiro,

“Gedar: geographic and opportunistic routing protocol with depth

adjustment for mobile underwater sensor networks,” in International

Conference on Communications, Sydney, Australia, pp. 251–256, IEEE,

2014.

[68] M. Akbar, N. Javaid, A. H. Khan, M. Imran, M. Shoaib, and A. Vasi-

lakos, “Efﬁcient data gathering in 3d linear underwater wireless sensor

networks using sink mobility,” Sensors, vol. 16, no. 3, p. 404, 2016.

[69] N. Ilyas, M. Akbar, R. Ullah, M. Khalid, A. Arif, A. Hafeez, U. Qasim,

Z. A. Khan, and N. Javaid, “Sedg: Scalable and efﬁcient data gathering

routing protocol for underwater wsns,” Procedia Computer Science,

vol. 52, pp. 584–591, 2015.

[70] L.-J. Chen, C.-H. Yu, T. Sun, Y.-C. Chen, and H.-h. Chu, “A hybrid

routing approach for opportunistic networks,” in Proceedings of the

SIGCOMM workshop on Challenged networks, Pisa, Italy, pp. 213–

220, ACM, 2006.

[71] E. Felemban, F. K. Shaikh, U. M. Qureshi, A. A. Sheikh, and S. B.

Qaisar, “Underwater sensor network applications: A comprehensive

survey,” International Journal of Distributed Sensor Networks, vol.

2015, Article ID 896832, 14 pages, 2015. doi:10.1155/2015/896832.

[72] S. M. Ghoreyshi, A. Shahrabi, and T. Boutaleb, “An inherently void

avoidance routing protocol for underwater sensor networks,” in Twelfth

International Symposium on Wireless Communication Systems, Brus-

sels, Belgium, pp. 1–6, IEEE, 2016.

[73] S.-C. M. Woo and S. Singh, “Scalable routing protocol for ad hoc

networks,” Wireless Networks, vol. 7, no. 5, pp. 513–529, 2001.

[74] M. Goetz and I. Nissen, “Guwmanetmulticast routing in underwater

acoustic networks,” in Military Communications and Information Sys-

tems Conference, Gdansk, Poland, pp. 1–8, IEEE, 2012.

[75] G. Toso, R. Masiero, P. Casari, O. Kebkal, M. Komar, and M. Zorzi,

“Field experiments for dynamic source routing: S2c evologics modems

run the sun protocol using the desert underwater libraries,” in Proceed-

ings of MTS/IEEE OCEANS, Bergen, Norway, pp. 10–13, IEEE, 2012.

[76] C. Tapparello, P. Casari, G. Toso, I. Calabrese, R. Otnes, P. van

Walree, M. Goetz, I. Nissen, and M. Zorzi, “Performance evaluation of

forwarding protocols for the racun network,” in Proceedings of the 8th

ACM International Conference on Underwater Networks and Systems,

Kaohsiung, Taiwan, p. 36, ACM, 2013.

[77] J. Lui, V. Misra, and D. Rubenstein, “On the robustness of soft state

protocols,” in Proceedings of the 12th IEEE International Conference

on Network Protocols, Berlin, Germany, pp. 50–60, IEEE, 2004.

[78] D. Zhang and E. Dong, “A virtual coordinate-based bypassing void

routing for wireless sensor networks,” IEEE Sensors Journal, vol. 15,

no. 7, pp. 3853–3862, 2015.

[79] V. G. Menon and P. J. Prathap, “Opportunistic routing with virtual co-

ordinates to handle communication voids in mobile ad hoc networks,”

in Advances in Signal Processing and Intelligent Recognition Systems,

Trivandrum, India, pp. 323–334, Springer, 2016.

[80] H. Badis, A. Munaretto, K. Al Agha, and G. Pujolle, “Optimal path

selection in a link state qos routing protocol,” in 59th Vehicular

Technology Conference (VTC), Milan, Italy, pp. 2570–2574, IEEE,

2004.

[81] S. Azad, K. T. Hasan, D. Nandi, and A.-S. K. Pathan, “A high-

throughput routing metric for multi-hop underwater acoustic networks,”

Computers & Electrical Engineering, vol. 44, pp. 24–33, 2015.

[82] H. Liu, B. Zhang, H. T. Mouftah, X. Shen, and J. Ma, “Opportunistic

routing for wireless ad hoc and sensor networks: Present and future

directions,” IEEE Communications Magazine, vol. 47, no. 12, pp. 103–

109, 2009.

[83] L. F. M. Vieira, “Performance and trade-offs of opportunistic routing

in underwater networks,” in Wireless Communications and Networking

Conference (WCNC), Paris, France, pp. 2911–2915, IEEE, 2012.

[84] T. Patel and P. Kamboj, “Opportunistic routing in wireless sensor

networks: A review,” in International Advance Computing Conference

(IACC), Bangalore, India, pp. 983–987, IEEE, 2015.

[85] R. W. Coutinho, A. Boukerche, L. F. Vieira, and A. A. Loureiro,

“Design guidelines for opportunistic routing in underwater networks,”

IEEE Communications Magazine, vol. 54, no. 2, pp. 40–48, 2016.

[86] R. W. Coutinho, A. Boukerche, L. F. Vieira, and A. A. Loureiro, “On

the design of green protocols for underwater sensor networks,” IEEE

Communications Magazine, vol. 54, no. 10, pp. 67–73, 2016.

[87] M. C. Domingo and R. Prior, “Energy analysis of routing protocols

for underwater wireless sensor networks,” Computer communications,

vol. 31, no. 6, pp. 1227–1238, 2008.

[88] F. Al Salti, N. Alzeidi, and B. R. Arafeh, “Emggr: an energy-

efﬁcient multipath grid-based geographic routing protocol for under-

water wireless sensor networks,” Wireless Networks, pp. 1–14, 2016,

doi:10.1007/s11276-016-1224-0.

[89] T. Wu and N. Sun, “A reliable and evenly energy consumed routing pro-

tocol for underwater acoustic sensor networks,” in 20th International

Workshop on Computer Aided Modelling and Design of Communica-

tion Links and Networks (CAMAD), Toronto, Canada, pp. 299–302,

IEEE, 2015.

[90] D. E. Lucani, M. M´

edard, and M. Stojanovic, “Network coding

schemes for underwater networks: the beneﬁts of implicit acknowl-

edgement,” in Proceedings of the second workshop on Underwater

networks, Montreal, Canada, pp. 25–32, ACM, 2007.

[91] C. Zidi, F. Bouabdallah, and R. Boutaba, “Routing design avoiding

energy holes in underwater acoustic sensor networks,” Wireless Com-

munications and Mobile Computing, vol. 16, pp. 2035–2051, 2016.

[92] N. Javaid, M. Shah, A. Ahmad, M. Imran, M. I. Khan, and A. V.

Vasilakos, “An enhanced energy balanced data transmission protocol

for underwater acoustic sensor networks,” Sensors, vol. 16, no. 4,

p. 487, 2016.

[93] Z. Guo, G. Colombo, B. Wang, J.-H. Cui, D. Maggiorini, and G. P.

Rossi, “Adaptive routing in underwater delay/disruption tolerant sensor

networks,” in Fifth Annual Conference on Wireless on Demand Network

Systems and Services, Garmisch-Partenkirchen, Germany, pp. 31–39,

IEEE, 2008.

[94] C. R. Lin and J.-S. Liu, “Qos routing in ad hoc wireless networks,”

IEEE Journal on Selected Areas in Communications, vol. 17, no. 8,

pp. 1426–1438, 1999.

[95] C. Zhu and M. S. Corson, “Qos routing for mobile ad hoc networks,” in

Twenty-First Annual Joint Conference of the IEEE Computer and Com-

munications Societies, INFOCOM, New York, USA, vol. 2, pp. 958–

967, IEEE, 2002.

[96] G. Han, J. Jiang, L. Shu, Y. Xu, and F. Wang, “Localization algorithms

of underwater wireless sensor networks: A survey,” Sensors, vol. 12,

no. 2, pp. 2026–2061, 2012.

[97] V. Chandrasekhar, W. K. Seah, Y. S. Choo, and H. V. Ee, “Localization

in underwater sensor networks: survey and challenges,” in Proceedings

of the 1st ACM international workshop on Underwater networks, Los

Angeles, USA, pp. 33–40, ACM, 2006.

[98] Z. Zhou, J.-H. Cui, and S. Zhou, “Localization for large-scale underwa-

ter sensor networks,” in Ad hoc and sensor networks, wireless networks,

next generation internet, Atlanta, USA, pp. 108–119, Springer, 2007.

[99] M. Erol-Kantarci, H. T. Mouftah, and S. Oktug, “A survey of ar-

chitectures and localization techniques for underwater acoustic sensor

networks,” IEEE Communications Surveys & Tutorials, vol. 13, no. 3,

pp. 487–502, 2011.

[100] X. Wu, G. Chen, and J. Chen, “Energy-efﬁcient and topology-aware

routing for underwater sensor networks,” in Proceedings of 19th

International Conference on Computer Communications and Networks

(ICCCN), Zurich, Switzerland, pp. 1–6, IEEE, 2010.

[101] A. Khasawneh, M. S. B. A. Latiff, H. Chizari, M. Tariq, and A. Bama-

traf, “Pressure based routing protocol for underwater wireless sensor

networks: A survey,” KSII Transactions on Internet and Information

Systems (TIIS), vol. 9, no. 2, pp. 504–527, 2015.

[102] R. Mohammadi, R. Javidan, and A. Jalili, “Fuzzy depth based routing

protocol for underwater acoustic wireless sensor networks,” Journal

of Telecommunication, Electronic and Computer Engineering (JTEC),

vol. 7, no. 1, pp. 81–86, 2015.

[103] S. K. Dhurandher, M. S. Obaidat, and M. Gupta, “A novel geocast

technique with hole detection in underwater sensor networks,” in

ACS International Conference on Computer Systems and Applications

(AICCSA), Hammamet, Tunisia, pp. 1–8, IEEE, 2010.

Communications Surveys & Tutorials

[104] S. K. Dhurandher, M. S. Obaidat, and M. Gupta, “An efﬁcient tech-

nique for geocast region holes in underwater sensor networks and its

performance evaluation,” Simulation Modelling Practice and Theory,

vol. 19, no. 9, pp. 2102–2116, 2011.

[105] B. Diao, Y. Xu, Z. An, F. Wang, and C. Li, “Improving both energy

and time efﬁciency of depth-based routing for underwater sensor

networks,” International Journal of Distributed Sensor Networks, vol.

2015, Article ID 781932, 9 pages, 2015. doi:10.1155/2015/781932.

[106] K. Akkaya and M. Younis, “A survey on routing protocols for wireless

sensor networks,” Ad hoc networks, vol. 3, no. 3, pp. 325–349, 2005.

[107] D. Pompili, T. Melodia, and I. F. Akyildiz, “Routing algorithms for

delay-insensitive and delay-sensitive applications in underwater sensor

networks,” in Proceedings of the 12th annual international conference

on Mobile computing and networking, Los Angeles, USA, pp. 298–309,

ACM, 2006.

[108] I. Stojmenovic, M. Russell, and B. Vukojevic, “Depth ﬁrst search and

location based localized routing and qos routing in wireless networks,”

in international conference on Parallel processing, pp. 173–180, IEEE,

2000.

[109] A. Darehshoorzadeh and A. Boukerche, “Underwater sensor networks:

a new challenge for opportunistic routing protocols,” IEEE Communi-

cations Magazine, vol. 53, no. 11, pp. 98–107, 2015.

[110] N. Nowsheen, G. Karmakar, and J. Kamruzzaman, “Pradd: A path

reliability-aware data delivery protocol for underwater acoustic sensor

networks,” Journal of Network and Computer Applications, vol. 75,

pp. 385–397, 2016.

[111] D. Pompili, T. Melodia, and I. F. Akyildiz, “Distributed routing algo-

rithms for underwater acoustic sensor networks,” IEEE Transactions

on Wireless Communications, vol. 9, no. 9, pp. 2934–2944, 2010.

[112] X. Wang, G. Xing, Y. Zhang, C. Lu, R. Pless, and C. Gill, “Integrated

coverage and connectivity conﬁguration in wireless sensor networks,”

in Proceedings of the 1st international conference on Embedded

networked sensor systems, Los Angeles, USA, pp. 28–39, ACM, 2003.

[113] L. Moraru, P. Leone, S. Nikoletseas, and J. D. Rolim, “Near optimal

geographic routing with obstacle avoidance in wireless sensor networks

by fast-converging trust-based algorithms,” in Proceedings of the 3rd

ACM workshop on QoS and security for wireless and mobile networks,

Chania, Greece, pp. 31–38, ACM, 2007.

[114] M. Soyturk and T. Altilar, “The challenges and the approaches for the

geographic routing protocols in wireless sensor networks,” Proceedings

of the IEEE International Conference on Technologies for Homeland

Security and Safety (TEHOSS), Istanbul, Turkey, pp. 1–7, 2006.

[115] A.-d. Zhan, T.-y. Xu, G.-h. Chen, B.-l. Ye, and S.-l. Lu, “A survey

on real-time routing protocols for wireless sensor networks,” Chinese

Journal of Computer Science, vol. 3, no. 11, pp. 234–238, 2008.

[116] R. Ding and L. Yang, “A reactive geographic routing protocol for

wireless sensor networks,” in Sixth International Conference on Intel-

ligent Sensors, Sensor Networks and Information Processing (ISSNIP),

Brisbane, Australia, pp. 31–36, IEEE, 2010.

[117] I. F. Akyildiz, D. Pompili, and T. Melodia, “Underwater acoustic sensor

networks: research challenges,” Ad hoc networks, vol. 3, no. 3, pp. 257–

279, 2005.

[118] G. Acs and L. Butty´

an, “A taxonomy of routing protocols for wireless

sensor networks,” BUTE Telecommunication department, pp. 32–40,

2007.

[119] R. C. Shah and J. M. Rabaey, “Energy aware routing for low energy

ad hoc sensor networks,” in Wireless Communications and Networking

Conference, Orlando, USA, pp. 350–355, IEEE, 2002.

[120] C. Schurgers and M. B. Srivastava, “Energy efﬁcient routing in wireless

sensor networks,” in Military communications conference, Washington

USA, pp. 357–361, IEEE, 2001.

[121] I. F. Akyildiz, D. Pompili, and T. Melodia, “State-of-the-art in protocol

research for underwater acoustic sensor networks,” in Proceedings of

the 1st ACM international workshop on Underwater networks, Los

Angeles, USA, pp. 7–16, ACM, 2006.

Seyed Mohammad Ghoreyshi received the B.E.,

and M.E. degrees from Mazandaran University of

Science and Technology, and University of Tehran,

in 2009, and 2013, respectively. He is currently

a PhD student at Glasgow Caledonian University

in the UK. His current research interests include

underwater sensor networks, designing the routing

protocols and void-handling techniques.

Dr. Alireza Shahrabi received the B.Sc. and M.Sc.

degree in Computer Engineering from Sharif Univer-

sity of Technology, Tehran, Iran, in 1991 and 1994,

respectively, and the PhD degree in Computing

Science from University of Glasgow, Glasgow, UK

in 2003. Since September 2001 he has been with

the School of Engineering and Built Environment,

Glasgow Caledonian University where he is now a

Reader. His current research interests are network

protocols, wireless, mobile and sensor networks and

also performance modelling and evaluation of paral-

lel and distributed systems. He has published extensively in leading journals

and well-established conferences. He has been on the editorial board of

some international journals and also served as the organising and program

committees of many international conferences and workshops. Dr Shahrabi is

a member of the IEEE Computer Society.

Dr. Tuleen Boutaleb received the BEng degree

in Electronics Engineering and the PhD degree in

cellular mobile networks for telemetry from Glas-

gow Caledonian University, UK, in 1995 and 2006,

respectively. From 1995 to 1998 she was a researcher

on a European project working on the communica-

tions infrastructure. From 1998 to 2001 she was a

research assistant working on a European project.

Since 2001, she has been a telecommunications

engineering lecturer within the School of Engineer-

ing and Built Environment, Glasgow Caledonian

University, UK. Her research interests include cellular communication net-

works performance enhancement, messaging in VANET, WSN and satellite

communications for remote monitoring.

Cluster-Based Multi-attribute Routing Protocol for Underwater Acoustic Sensor Networks

Article

Full-text available

Mar 2024
WIRELESS PERS COMMUN

Underwater Acoustic Sensor Networks play a significant role in various underwater applications. There are several challenges in underwater communications like high bit-error-rate, low bandwidth, high energy consumption, void-node during routing, etc. Handling void-node during routing is a major challenge in underwater routing. There are well-known void-handling protocols like Energy-efficient Void-Aware Geographic Routing protocol, HydroCast, etc. However, these routing protocols require all neighboring nodes must be a part of the cluster which increases the overhead on clustering, or void-node has a part of the routing. This paper proposes an underwater routing protocol referred to as Cluster-based Multi-Attribute Routing (CMAR) to overcome these issues. It is a sender-based, opportunistic underwater routing protocol. CMAR uses the Technique for Order of Preference by Similarity to Ideal Solution to evaluate the suitability of the neighboring nodes and the basis for clustering process initialization. Through MATLAB simulations, the performance of the CMAR is compared with HydroCast in terms of the number of nodes selected in the forwarding set, number of clusters formed, number of times void-node becomes part of routing and transmission reliability.

Dynamic Packet Routing Based on Acoustic Signal Curve Propagation in the AUV-Assisted IoUT

Article

Full-text available

Jan 2023

Autonomous underwater vehicles (AUVs) can function as sensor nodes in Internet of Underwater Things (IoUT), contributing to ocean exploration and monitoring by collecting and transmitting data to the base station. Most of the routing algorithms applied to IoUT require the participation of stationary nodes and seldom consider the fluctuations of network topology, which cannot be directly applied to the IoUT composed of AUVs. The focus of this research is to examine the problem of packet routing in a dynamic AUV-assisted IoUT, with the ultimate goal of ensuring the effective transmission of underwater information. We analyze the transmission pattern of underwater acoustic signals and the consequent communication disruption between AUVs, which helps establish the age of information (AoI) and bit error rate (BER) of data through modeling. A routing algorithm that utilizes the branch and bound (BB) technique has been suggested, alongside the introduction of the value of information (VoI) to enable the joint optimization of the AoI and bit error rate (BER). We describes two nearly optimal heuristic algorithms for networks with a high number of AUVs. The AFA-ACOBB strategy is designed based on the above algorithms and the influence of AUV motion on link reliability is considered. Moreover, we have developed a power regulation mechanism that can effectively minimize the occurrence of network packet loss and energy waste. The simulation results demonstrate that the proposed scheme outperforms certain classically-related schemes in terms of AoI and BER, while simultaneously maintaining superior packet loss rate (PLR) and energy consumption.

A comprehensive survey of localization schemes and routing protocols with fault tolerant mechanism in UWSN- Recent progress and future prospects

Article

Full-text available

Feb 2024
MULTIMED TOOLS APPL

Kamal Kumar Gola

The underwater sensor network (UWSN) has become the most demandable area because of its numerous advancements in the real world. Water covers over 70% of the earth's surface, and there is more important information below the water's surface. Sensor networks detect information in the water and transmit it to the base station via the sink. UWSN is made up of autonomous and independent sensor nodes. These are placed spatially in the water to gather and transmit pertinent data. However, the communication may suffer due to several challenges like 3D topology, less bandwidth, routing and power constraints. This paper reviews underwater sensor networks based on routing, localization, and fault tolerance. It also reviews and compares the recently proposed algorithms to fulfil drawbacks in existing methods. Each factor’s challenges are briefly discussed and compared in this paper. Future directions for underwater sensor networks are analyzed and discussed, along with their emerging applications and challenges.

A Novel Hybridized Cluster-Based Geographical Opportunistic Routing Protocol for Effective Data Routing in Underwater Wireless Sensor Networks

Article

Full-text available

Nov 2023

Underwater wireless sensor nodes comprise hundreds to thousands of battery-operated sensor nodes with limited bandwidth. These networks are employed to transmit the data with enhanced quality of service (QoS). However, efficient data routing is the most challenging obstacle in many underwater applications. To solve the issues in underwater sensor nodes, the hybridized cluster-based geographical opportunistic routing protocol with distance vector establishment has been proposed to transmit the data efficiently. Primarily, the proposed methodology finds out the shortest path with minimal hop count whereas the void node can be updated with infinite hop count. Thereafter, the sleep/wake scheduling and waiting mechanism and periodic beaconing algorithm are incorporated into the proposed model to attain a higher packet delivery ratio with minimal energy consumption. This proper scheduling and optimal cluster routing enhance the continuous data transmission in underwater applications. The simulation result reveals that the proposed method achieves better energy efficiency and higher network lifetime when compared with the existing clustering methods.

ARO-RTP: Performance analysis of an energy efficient opportunistic routing for underwater IoT networks

Article

Full-text available

Oct 2023

The term "Underwater Internet of Things" (UIoT) refers to a system of intelligent, networked devices that may function in a variety of water conditions. During communication, the transmission quality is diminished by high levels of interference and collisions, which in turn causes a slow End-to-End (E2E), low latency, and a low Packet Delivery Ratio (PDR). And when the basis node does not choose an immediate forwarder node, an issue known as the "void hole" occurs and thus leads to the wastage of more power. Despite playing a crucial role in enabling underwater communication, IoUT frameworks still face difficulties because of their unstable radio signals, low resources, unreliable transmission medium, low bandwidth, limited range, low transmission rate, inborn noise, node mobility, slow propagation speed, and limited battery capacity. Therefore, it is greatly desirable to have secure communication with a prolonged network lifetime. In this research work, an opportunistic routing-based reliable transmission protocol (OR-RTP) is proposed to send the packs toward the superficial sink to lessen the high energy consumption (EC). The source node recognizes the furtherance relay set in the proposed protocol based on the forwarder's local information. The proposed protocol uses a meta-heuristic-based relay selection scheme (Artificial rabbits’ optimization-ARO) to pick the best relay by balancing the forwarder's EC and packet delivery ratio (PDR). During packet transmission, most of the energy is ruined because of collisions among Under Water Acoustic Sensor Networks (UW-ASNs') sensor nodes. This work developed an RTP for every forwarder node that sends data to the surface basin to reduce the problem of collision. The OR-RTP protocol has been extensively simulated, and the results are compared to those of other routing protocols in terms of EC, PDR, throughput, and network lifetime. The proposed model achieved 85% of PDR on 400 nodes and 92% of PDR for 700 nodes, whereas the existing model Grey Wolf Optimization achieved 51% of PDR on 400 nodes and 73% of PDR for 700 nodes.

EERP: An energy-efficient and reliable routing protocol for underwater acoustic wireless networks

Conference Paper

Jun 2024

WLQRP: A Weighting Link-Quality-Based Routing Protocol for Underwater Sensor Networks

Article

Full-text available

Dec 2023

Due to challenges posed by long propagation distances, high mobility, limited bandwidth, multipath propagation, and Doppler effects in underwater acoustic sensor networks (UASNs), the design of routing protocols faces numerous complexities. To enhance reliability in sparse node regions and reduce network energy consumption, this paper proposes a dependable and energy-efficient routing protocol termed WLQRP for weighting link-quality-based routing protocol. WLQRP leverages link quality information between nodes for data forwarding. When computing the link quality between nodes, it considers not only the historical transmission success ratio between nodes and their immediate relay but also incorporates the link quality between the relay and the relay’s next-hop nodes. This approach effectively reduces the likelihood of encountering void regions. Moreover, the protocol judiciously selects the next-hop node to improve transmission success rates, thereby minimizing data packet retransmissions and reducing energy consumption. We conduct MATLAB simulations to evaluate the WLQRP protocol, comparing it against the VBF (vector-based forwarding) and the HH-VBF (hop-by-hop vector-based forwarding) underwater routing protocol. Experimental results demonstrate that WLQRP enhances network performance in terms of data transmission rate, energy consumption, and end-to-end delay. These findings validate the efficacy of the proposed protocol.

Enhanced Link Quality-based Routing Protocol for underwater networks

Conference Paper

Jul 2023

Spatial Anti-Void Querying in IoT-enabled Wireless Sensor Networks

Conference Paper

Oct 2023

An Adaptive Routing Protocol for Underwater Acoustic Sensor Networks With Ocean Current

Article

Nov 2023

Long-term underwater acoustic sensor networks (UASNs) are widely deployed for marine scientific and industrial applications. The widespread ocean currents lead to variations in the communication range of location-fixed nodes equipped with non-omnidirectional transducers, which will reduce the successful packet relaying probability. In order to reduce the impact of ocean current on packet relaying, we propose an adaptive routing protocol for underwater acoustic sensor networks with ocean current (OCAR). By adaptively selecting candidate nodes based on communication range, OCAR improves the successful packet relaying probability in poor communication conditions caused by ocean current. Firstly, the impact of ocean current on communication range is demonstrated by constructing a communication range model and a sea trial. Secondly, successful packet relaying probability is improved by selecting the communicable nodes under ocean current as the candidate nodes. Next, the transmitter-based adaptive routing scheme reduces the number of nodes responding to the relay packet and adaptively reselects the destination node per hop. Therefore, energy consumption is reduced and the packet delivery ratio (PDR) is improved. Finally, routing recovery and candidate set dynamic update are achieved to improve successful packet relaying probability. Simulation results show that, compared with Flood, OCAR can achieve a maximum 98.29% reduction in energy consumption and 37.05% reduction in average end-to-end delay while maintaining close to 100% PDR. In addition, compared with VBF, OCAR can achieve close to 100% PDR when the PDR of VBF is 0, maximum energy consumption reduction of 84.41% and average end-to-end delay reduction of 96.34%.

On the design of green protocols for underwater sensor networks

Article

Full-text available

Oct 2016

Underwater sensor networks have enabled a new era in scientific and industrial underwater monitoring and exploration applications. However, these networks are energy-constrained and, more problematically, energy-hungry, as a consequence of the use of underwater acoustic links. In this work, we thoroughly review potential techniques for greening underwater sensor networks. In a top-down approach, we discuss the principal design and challenges of the appealing highlighted techniques. We also exemplify their use by surveying recent proposals in underwater sensor networks. Finally, we describe potential future research directions for energy conservation in underwater networks.

Location-free link state routing for underwater acoustic sensor networks

Conference Paper

Full-text available

May 2015

Relative Distance Based Forwarding Protocol for Underwater Wireless Networks

Article

Full-text available

Feb 2014
INT J DISTRIB SENS N

Underwater wireless sensor network (UWSN) features many unique characteristics, including slow propagation speed, high end-to-end delay, low available bandwidth, variable link quality, and energy constraint. All these problems pose a big challenge to devise a transmission efficient, energy-saving, and low delay routing protocol for UWSNs. In this paper we devise a relative distance based forwarding (RDBF) routing protocol, which aims to provide transmission efficient, energy-saving, and low delay routing. We utilize a fitness factor to measure and judge the degree of appropriateness for a node to forward the packets. Under the limitations of the fitness factor, RDBF can confine the scope of the candidate forwarders and find the beneficial relays to forward packets. In this way, only a small fraction of nodes are involved in forwarding process, which can distinctly reduce the energy consumption. Moreover, using only the selected beneficial nodes as forwarders can both enhance the transmission efficiency and reduce the end-to-end delay. This is because the distances of these nodes to the sink are the shortest and the hop counts of routing paths consisted by these nodes are minimum. We use the ns-2 based simulator to conduct our experiment; the results show that RDBF performs better in terms of packet delivery ratio, end-to-end delay, and energy efficiency.

An inherently void avoidance routing protocol for Underwater Sensor Networks

Conference Paper

Full-text available

Aug 2015

An Opportunistic Void Avoidance Routing Protocol for Underwater Sensor Networks

Conference Paper

Full-text available

Mar 2016

EMGGR: an energy-efficient multipath grid-based geographic routing protocol for underwater wireless sensor networks

Article

Full-text available

May 2017
WIREL NETW

Routing in underwater wireless sensor networks (UWSN) is an important and a challenging activity due to the nature of acoustic channels and to the harsh environment. This paper extends our previous work [Al-Salti et al. in Proceedings of cyber-enabled distributed computing and knowledge discovery (CyberC), Shanghai, pp 331–336, 2014] that proposed a novel multipath grid-based geographical routing (MGGR) protocol for UWSNs. The extended work, EMGGR, viewed the network as logical 3D grids. Routing is performed in a grid-by-grid manner via gateways that use disjoint paths to relay data packets to the sink node. The algorithm consists of three main components: (1) a gateway election algorithm; responsible for electing gateways based on their locations and remaining energy level (2) a mechanism for updating neighboring gateways’ information; allowing sensor nodes to memorize gateways in local and neighboring cells, and (3) a packet forwarding mechanism; in charge of constructing disjoint paths from source cells to destination cells, forwarding packets to the destination and dealing with holes (i.e. cells with no gateways) in the network. The performance of EMGGR has been assessed using Aqua-Sim, which is an NS2 based simulator for UWSNs. Results show that EMGGR is an energy efficient protocol in all simulation setups used in the study. Moreover, EMGGR can also maintain good delivery ratio and end-to-end delay.

On the Relationship between Transmission Power and Capacity of an Underwater Acoustic Communication Channel

Article

Full-text available

Jan 2008

The underwater acoustic channel is characterized by a path loss that depends not only on the transmission distance, but also on the signal frequency. As a consequence, transmission bandwidth depends on the transmission distance, a feature that distinguishes an underwater acoustic system from a terrestrial radio system. The exact relationship between power, transmission band, distance and capacity for the Gaussian noise scenario is a complicated one. This work provides a closed-form approximate model for 1) power consumption, 2) band-edge frequency and 3) bandwidth as functions of distance and capacity required for a data link. This approximate model is obtained by numerical evaluation of analytical results which takes into account physical models of acoustic propagation loss and ambient noise. The closed-form approximations may become useful tools in the design and analysis of underwater acoustic networks.

Advanced flooding-based routing protocols for underwater sensor networks

Article

Full-text available

May 2016

Flooding-based protocols are a reliable solution to deliver packets in underwater sensor networks. However, these protocols potentially involve all the nodes in the forwarding process. Thus, the performance and energy efficiency are not optimal. In this work, we propose some advances of a flooding-based protocol with the goal to improve the performance and the energy efficiency. The first idea considers the node position information in order to reduce the number of relays that may apply flooding. Second, a network coding-based protocol is proposed in order to make a better use of the duplicates. With network coding, each node in the network recombines a certain number of packets into one or more output packets. This may give good results in flooding-based protocols considering the high amount of packets that are flooded in the network. Finally, a fusion of both ideas is considered in order to exploit the benefits of both of them.

Pressure routing for underwater sensor networks

Article

Jan 2010

A reliable and evenly energy consumed routing protocol for underwater acoustic sensor networks

Conference Paper

Sep 2015

Void-Handling Techniques for Routing Protocols in Underwater Sensor Networks: Survey and Challenges

Abstract and Figures

Recommended publications

A Survey on Multi-path Routing Protocols in Wireless Multimedia Sensor Networks

An Underwater Routing Protocol with Void Detection and Bypassing Capability

A Stateless Opportunistic Routing Protocol for Underwater Sensor Networks

A Novel Cooperative Opportunistic Routing Scheme for Underwater Sensor Networks

An inherently void avoidance routing protocol for Underwater Sensor Networks