research article routing protocols in underwater acoustic...

Research ArticleRouting Protocols in Underwater Acoustic Sensor Networks:A Quantitative Comparison

Guangjie Han,1 Na Bao,1 Li Liu,1 Daqiang Zhang,2 and Lei Shu3

1Department of Information & Communication Systems, Hohai University, Changzhou 213022, China2School of Software Engineering, Tongji University, Shanghai 200092, China3Guangdong Petrochemical Equipment Fault Diagnosis Key Laboratory, Guangdong University of Petrochemical Technology,Guangdong 525000, China

Correspondence should be addressed to Guangjie Han; [email protected]

Received 14 May 2015; Accepted 12 August 2015

Academic Editor: Jianping He

Copyright © 2015 Guangjie Han et al. This is an open access article distributed under the Creative Commons Attribution License,which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Underwater Acoustic Sensor Networks (UASNs) have drawn great attention for their potential value in ocean monitoring andoffshore exploration. In order to make the underwater application possible, the unique characteristics of underwater acousticchannels and continuous node movement inspired the emergence of routing protocols for underwater environment. In this paper,we introduce and compare four prominent routing protocols proposed for UASNs, namely, H2-DAB, GEDAR, E-PULRP, andPER. Performances of the routing protocols are evaluated in terms of the average number of control packets, end-to-end delay,data delivery ratio, and total energy consumption. The impact of water currents on the routing algorithms is also analyzed in oursimulation. Experimental results demonstrate that E-PULRP provides high data delivery ratio at the cost of end-to-end delay. H2-DAB has better real-time performance for minimal delay transmission. GEDAR efficiently addresses the problem of void regionwithout introducing extra energy. PER requires the most control packets in the process of routing establishment. Our work aimsto provide useful insights to select appropriate routing protocols to fulfil different application requirements in UASNs.

1. Introduction

Nearly 71% of the Earth’s surface is covered bywater.The deepocean is a vast and mostly unexplored habitat on our planet.Recently, there has been a growing interest in exploring andmonitoring ocean environments for scientific exploration,commercial exploitation, or defense and security purposes.Because of the inhospitable environment (e.g., unpredictableunderwater activities and high water pressure), unmannedexploration is a promising solution for discovering the aque-ous environment. Over the past few years, UnderwaterAcoustic Sensor Networks (UASNs) have shown great valuein ocean exploration activities and attracted the interestof many researchers [1–4]. In an UASN, sensor nodes aredeployed in underwater environment, covering the entiremonitored space to cooperatively fulfill monitoring tasks.In order to make the underwater application possible, it isessential to propose efficient routing protocols that achievereliable communications.

Althoughmany routing protocols for Terrestrial WirelessSensor Networks (TWSNs) [5, 6] have been researched upto now, the design of routing protocol in UASNs is still achallenging problem due to the following reasons: (1) limitedcommunication bandwidth, (2) high propagation delay, (3)high bit error ratio, and (4) batteries being energy constrainedandnot able to be recharged (solar energy cannot be exploitedin underwater environment) [7]. Also, in a TWSN, nodes arealways considered to be static so that the location informationcan be easily acquired through localization algorithms orGPS [8]. Oppositely, due to the water currents or marineorganisms, floating node mobility is an inevitable issue inUASNs [9, 10]. In this regard, traditional routing protocolsdesigned for TWSNs cannot be directly applied to UASNs.

UWSNs rely on underwater acoustic communicationsbecause high-frequency radio signals used in TWSNs canbe rapidly absorbed by water [11]. The propagation delayis five orders of magnitude higher than in radio frequency

Hindawi Publishing CorporationInternational Journal of Distributed Sensor NetworksVolume 2015, Article ID 858593, 11 pageshttp://dx.doi.org/10.1155/2015/858593

2 International Journal of Distributed Sensor Networks

terrestrial channels [2]. Moreover, data forwarding is acomplex procedure in underwater environment because ofthe time-dependent temperature, pressure, salinity, lightness,and biology activity [9]. Due to the fact that continuous nodemovement makes it hard to manage the location informationof sensor nodes, it is a challenging task to find and maintainthe routes for dynamic underwater environments. Sinceenergy harvesting is almost unavailable in the ocean [12],optimizing the amount of energy used for data transmissionis also an important issue in routing protocol design. Based onthe analysis above, more efficient routing protocols in UASNsshould be developed to ensure reliable data delivery whileoptimizing energy consumption.

The main contribution of this paper is that we introduceand compare the performance of four routing protocolswhich are suitable for large-scale UASNs. The schemes arehop-by-hop dynamic addressing based (H2-DAB) [13], geo-graphic and opportunistic routing with depth adjustment-based topology control for communication recovery(GEDAR) [14], energy optimized path unaware layered rout-ing protocol (E-PULRP) [15], and power-efficient routing(PER) [16] protocol. To the best of our knowledge, this is thefirst work that studies performance evaluations of H2-DAB,GEDAR, E-PULRP, and PER. We use MATLAB as the sim-ulation tool and compare the performance of the protocols interms of the average number of control packets, end-to-enddelay, data delivery ratio, and total energy consumption.To embody the dynamic characteristic of UASNs, we alsoanalyze the impacts of water currents on the performance ofthe routing protocols. The characteristics of each protocolare revealed by quantities of comparison experiments. Thepurpose of this work is to provide useful insights to selectappropriate routing protocols to fulfil different applicationrequirements in UASNs.

The remainder of the paper is organized as follows.Section 2 summarizes typical routing protocols designed forUASNs. Section 3 introduces H2-DAB, GEDAR, E-PULRP,and PER in detail. A detailed analysis of simulation resultsis discussed in Section 4. Section 5 draws conclusions andpoints out the direction of future research.

2. Related Work

In the literature, most existing routing protocols can beclassified into two categories: geographic-based routing pro-tocols and hybrid-based (both energy-based and geographic-based) routing protocols. Geographic-based protocols (e.g.,[13, 14, 16–21]) leverage the location information of sensornodes to forward packets from a source node to a desti-nation node. Besides geographic information, hybrid-basedprotocols (e.g., [2, 15, 22–26]) also take into account theenergy optimization. In the rest of the section, representativegeographic-based and hybrid-based routing protocols arereviewed, respectively.

2.1. Geographic-Based Routing Protocol. In [17], Yan et al.proposed a depth-based routing (DBR). In DBR, with the

depth of each node known from pressure sensors, data pack-ets are forwarded along the upward direction. Multisinks areintroduced in DBR to improve the routing success ratio. Thepacket delivery ratio depends upon the node distribution andthe number of sinks. Based onDBR, two improved protocols,namely, adaptive mobility of courier nodes in threshold-optimized depth-based routing (AMCTD) [18] and depth-based multihop routing (DBMR) [19], are proposed.

In [18], Jafri et al. proposed an adaptive mobility of cou-rier nodes in threshold-optimized depth-based routing(AMCTD). AMCTD addresses the problems of little stabilityperiod, swift energy consumption of low-depth nodes, andpoor throughput during the instability period caused byunequal load distribution among the nodes. In order toprolong the network lifetime,AMCTDexplores the proficientamendments in depth threshold and implements the optimalweight function.

In [19], Liu and Li proposed a depth-basedmultihop rout-ing (DBMR). Since DBR uses flooding mode for transmis-sion, a large amount of redundant data forwarding and chan-nel occupancy might be caused. To reduce communicationoverhead,DBMRadoptsmultihop to sendpackets.Moreover,DBMR can take advantage of multiple-sink underwatersensor network architecture without introducing extra cost.

In [20], Chen and Lin proposed a mobicast routingprotocol (MRP) which takes the mobility of nodes into con-sideration. Packets can be gathered within a sphere centeredin an autonomous underwater vehicle (AUV). Nodes withinthe range of the sphere upload data packets to the AUV. Theother nodes are in idle state to wait for the arrival of the AUV.The sphere can be split up into slices due to the effect of watercurrents.When the nodes in a slice drift away, the slice wouldbe enlarged to contain more nodes. The slices would shrinkto reduce the number of nodes when additional nodes flowinto the slice.

In [13], Ayaz et al. proposed a hop-by-hop dynamicaddressing based (H2-DAB) routing protocol. H2-DAB tack-les the challenges of UASNs by implementing the dynamicaddressing scheme among the sensor nodes without requir-ing the localization information.

In [14], Coutinho et al. proposed a geographic andopportunistic routing with depth adjustment-based topologycontrol for communication recovery (GEDAR). GEDAR usesthe greedy opportunistic mechanism to route data packet. Toaddress the problem of void regions, GEDAR takes advantageof the depth adjustment apparatus in the current underwatersensor node technology andmoves void nodes to new depthsto adjust network topology.

In [21], Zhang et al. proposed a prediction based delay-tolerant protocol (PBDTP). PBDTP is designed to address theproblems of link reliability, long delay, and inconsistent delayin UASNs. Using data prediction and adjustment algorithms,PBDTP can predict a value for a sensor node if its data werenot received at the sink node.

In [16], Guo et al. proposed a generic prediction assistedsingle-copy routing (PASR). PASR employs an effectivegreedy algorithm ACPG which captures the features ofnetwork mobility patterns. Furthermore, an online heuristic

International Journal of Distributed Sensor Networks 3

Table 1: Characteristics of routing protocols proposed for UASNs in related work.

Routing protocols Timesynchronization

Multipathestablishment

Collisionavoidance

Routing voiddetection Routing update Energy

efficiencyQELAR [2] No No MAC No Partial update YesDBR [10] No Yes MAC No Periodical update YesAMCTD [11] No Yes CSMA/CA No Periodical update YesDBMR [12] No Yes MAC Yes Periodical update YesMRP [17] No Yes MAC Yes Partial update YesH2-DAB [18] No Yes MAC No Periodical update YesGEDAR [19] No Yes CSMA Yes Full update YesPBDTP [20] No Yes CSMA No Full update YesPASR [13] No No CSMA Yes Partial update YesTSBNC [14] Yes No TDMA Yes Periodical update YesE-PULRP [21] No Yes CDMA No Full update YesPER [16] No Yes MAC No Full update YesL2-ABF [22] No Yes MAC No Full update YesLB-AGR [15] No No MAC Yes Partial update YesEEDBR [23] No Yes MAC Yes Full update Yes

protocol is designed by choosing appropriate historical infor-mation and forwarding criteria based on the guidance fromACPG.

2.2. Hybrid-Based Routing Protocol. In the design of routingprotocol, energy optimization is also a major concern. In [2],Hu and Fei used an adaptive, energy-efficient, and lifetime-aware routing protocol based on reinforcement learningrouting protocol (QELAR). In QELAR, the residual energy ofeach node, as well as the energy distribution among a groupof nodes, is factored in throughout the routing process tocalculate the reward function. The reward function aids inselecting the adequate forwarders for packets. Each node inthe network is responsible for learning the environment, aim-ing to take the optimal action and improving the performanceof the whole network gradually.

In [22], Wu et al. proposed a time-slot-based routingalgorithm (TSR). A probability balanced mechanism namedTSBR is applied to TSR to eliminate redundancy and reducebit error ratio. The theory of network coding is also intro-duced to TSBR to meet the requirements of further reducingnode energy consumption and extending the network life-time.

In [15], Gopi et al. proposed an energy optimized pathunaware layered routing protocol (E-PULRP). E-PULRPconsists of two phases: layering phase and communicationphase. In the layering phase, a set of concentric shells (layers)are formed around the central sink node. The layeringstructure ensures that the packet is forwarded towards thesink node. In the communication phase, one relay node isidentified from each layer to forward the packet.

In [23], Huang et al. proposed a power-efficient routing(PER) protocol. Based on the fuzzy logic inference system,a forwarding node selector is employed to determine theappropriate sensors to forward the packets to the destination.

Also, a forwarding tree trimming mechanism is adopted toprevent excess spread of forwarded packets.

In [24], Ali et al. proposed a layer-by-layer angle basedflooding (L2-ABF) routing protocol. L2-ABF addresses theproblem of nodes scatter in UASNs. In L2-ABF, without usingany explicit configuration and location information, eachnode can calculate its flooding angle and then forward datapackets to the next upper layer toward surface sinks. Thenumber of nodes which flood the data packets is controlledby the angle for flooding cone to prevent flooding over thewhole network. The flooding cone is adjusted in layer-by-layer manner by using the angle based technique among theupper layer nodes.

In [25], Du et al. proposed a level based adaptivegeographic routing (LB-AGR) protocol. Based on availableenergy, density, location, and level-difference between neigh-bor nodes, LB-AGR defines an integrated forwarding factorfor each candidate node. The integrated forwarding factor isused to determine the best next-hop amongmultiple qualifiedcandidates.

In [26], Wahid and Kim proposed an energy-efficientdepth-based routing (EEDBR) protocol. EEDBR utilizes thedepth and the residual energy of sensor nodes as a routingmetric. A sender-based approach is employed for routingwhere the sender decides a set of next forwarding nodesin order to reduce redundant transmissions from multipleforwarders.

We conclude the characteristics of routing protocolsmen-tioned above in Table 1.

3. Typical Routing Protocols for UASNs

Since routing protocols might differ depending on the appli-cation and network architecture, it is unfair and misleadingto compare them without considering their assumptions and


Node 1

Node 24

Node 2Node 3 Node 4 Node 5

Node 11

Node 20

Node 15

Node 10

Node 19Node 21

Node 14 Node 16

Node 25

Node 8

Node 26

Node 9

Node 23

Node 18

Node 13

Node 22

Node 12

Node 7

19

29

38

56

11

22

33

44

55

16

25

34

45

56

19

29

39

49

59

39

49

59

19

29

Sink 1 Sink 2 Sink 3 Sink 4 Sink 5

Node 17 47

Surface sink

Acoustic link(with surface sink)

Ordinary node

Radio link

Figure 1: An example of dynamic addressing.

objectives. A comparative analysis based on a set of samplealgorithms which share the same design assumptions andobjectives is reasonable. In this section, four typical routingprotocols proposed for UASNs, namely, H2-DAB, GEDAR,E-PULRP, and PER, are introduced in detail.

3.1. H2-DAB. To deal with the node mobility in UASNs,H2-DAB assigns a dynamic address (called HopID) to eachsensor node based on the hop count from the sink node.The process is as follows. The sink node broadcasts a Hellopacket. Each receiving node is assigned a HopID. Then, thereceiving nodes increment the HopID and rebroadcast theHello packet including the updated HopID. Since the HopIDis increased hop by hop, the nodes closer to the sink willbe assigned smaller HopIDs than the nodes away from thesink. During the forwarding of data packets, the nodes havingsmall HopIDs are selected for forwarding the data packets.Multisink is employed to balance the energy consumption.When a routing path is invalid, the source node can chooseother sink nodes to upload data. All the routing paths will beperiodically updated when their deadline expires.

3.1.1. Addressing Schemes. As shown in Figure 1, each nodecalculates the hop count to two sinks and stores the hop countinformation. TakingNode 1 as an example, the least hop countit stores is one; it means that Node 1 can transmit its packetsto a sink node (Sink 1) in one hop. And the second lowest hopcount is nine; it means that there exists a sink node which isnine hops away from Node 1. The routing path which has theleast hop count is selected as primary path while the otherwhich has the second lowest hop count works as an alternatepath.

3.1.2. Route Updating and Maintenance. Before the deadlineof the route expires, if the whole network is affected by low-speed water current for a short time, the previous routing

path is regarded as valid. Otherwise, hop count of eachnode would be recalculated according to updated networktopology information.

3.2. GEDAR. GEDAR takes the advantage of greedy oppor-tunistic forwarding to improve data delivery ratio. With thecapability of depth adjustment, the problem of void regioncan be addressed. Therefore, the connectivity between thesource node and the sink can be ensured. GEDAR usesmultisink to balance energy consumption in UASNs.

3.2.1. Greedy Forwarding Strategy. A greedy opportunis-tic forwarding strategy is applied for next-hop forwarderselection. The main idea is to advance the packet towardssome sonobuoy in each hop. Given the Euclidean distancebetween a source node and the candidate forwarder nodeand the packet delivery probability of packets, the normal-ized advance (NADV) [27] is defined as the product ofthe Euclidean distance and the delivery probability. Besidesachieving trade-off between proximity and link cost, NADVis used to determine the priorities of the candidate nodes.The neighbor which has the largest NADV is selected as theprimary forwarding node while the neighbor which has thesecond largest NADV is selected as an alternative forwardingnode. In general case, packets are propagated by forwarderswhich have the highest priorities. When predefined waitingtime expires, primary forward nodes still fail to deliver thepackets; alternative forwarding nodes can replace the primaryforward nodes and complete the propagation of the packets.

3.2.2. Recovery Mode. If a node cannot forward a packetusing the greedy forwarding strategy, GEDAR switches tothe void node recovery mode. GEDAR allows void nodes tobroadcast announcement messages to their neighbors. Theneighbors that receive the void node announcement messagewill remove the void node from their routing table. Throughmapping a 3D environment into a 2D plane, the void nodethen determines the new depth such that the distance to itsclosest sonobuoy is greater than the distance of the neighborto the nearest sink. And then, void nodes adjust their depthwhich results in the minimum displacement. If the nodecannot determine a new depth, the recovery mode functionis called again.

3.3. E-PULRP. E-PULRP is an improved protocol based onpath unaware layered routing protocol (PULRP) [28]. E-PULRP consists of two phases, namely, layering phase andcommunication phase. Different fromH2-DAB andGEDAR,only one sink node is allowed in E-PULRP.

3.3.1. Layering Phase. In E-PULRP, a set of concentric shells(layers) are formed around the central sink node. Figure 2is a planar graph of the layered network, where 𝑅

𝐿is the

maximum attainable transmission range of a node in layer 𝐿and 𝑎𝐿is the width of layer 𝐿. Layer formation is explained as

follows: a probe of energy Ep1 is initiated at the sink node(layer 0) and all nodes that receive the probe with energyat least equal to Ed (the detection threshold) will assign


Sink node

Layer2

Layer3

Layer4

Layer5

Range circle

NodeSink node

Layer 1

RLaL

T

Figure 2: An example of layered network.

themselves as layer 1 nodes. Layer 1 nodes can communicatewith the sink node in single hop. Following the same step,the whole network can be covered by a certain number oflayers and data can be transmitted to the sink layer by layer.In addition, the transmission power of each node is adjustableso that nodes within each layer are able to bear suitable trafficload.

3.3.2. Communication Phase. Packets are forwarded fromouter layer to the sink. When a source node uploads datato the sink, it enquires its neighbors which belong to aninner layer. Based on neighbor nodes’ response time and theirresidual energy, the best next forwarding node can be chosen.

3.4. PER. The architecture of PER consists of two modules,namely, a forwarding node selector and a forwarding treetrimming mechanism. As shown in Figure 3(a), the for-warding node selector selects at most two candidate sensornodes to forward the packets. In Figure 3(b), a forwardingtree trimming mechanism is used to prevent unnecessarypower consumption of the sensor nodes from fast spreadingof packet forwarding over the UASNs.

3.4.1. Fuzzy Logic System and Decision Tree. Fuzzy logicsystem is adopted as a candidate as the forwarding nodeselector. The forwarding node selector utilizes the threeparameters, including the distance and the angle between twoneighboring sensor nodes, and the remaining energy left inthe sensor node as input of fuzzy logic system. In PER, atypical C4.5 decision tree [29] is adopted to cooperate withfuzzy logic system; a decision tree and fuzzy logic systemcollaborate to determine the next forwarder.

3.4.2. Forwarding Tree Trimming Mechanism. An intermedi-ate sensor can employ the trimming mechanism to forward

Table 2: Simulation parameters.

Parameters ValueThe number of ordinary nodes 350Transmission range of ordinary nodes 100mTransmission range of a sink 150mFrequency 10 kHzTransmission energy consumption 50 nJ/bitWater velocity (V) 0.5m/s and 0.8m/sData packet size 218 bytesControl packet size 40 bytes

the packets to the top-selected forwarding node.The interme-diate sensor then transmits the packets to the second selectedforwarding node if the number of duplicated packets receivedby the intermediate sensor does not exceed a preset threshold.This mechanism not only resolves the problem of broadcastpacket, but also prevents the disruption of packet forwardingfrom the inappropriate selection of forwarding nodes.

4. Performance Analysis and Comparison

4.1. Energy Consumption Model. We use the energy con-sumption model proposed in [30]. The energy dissipation ofeach node consists of transmission and receiving. AccordingtoThorps expression [31], the absorption coefficient𝐴(𝑓) canbe calculated by the following formula:

𝐴 (𝑓) =0.1𝑓2

1 + 𝑓2+44𝑓2

4100 + 𝑓2+ 2.75 × 10

−4𝑓2+ 0.003, (1)

where 𝑓 is the center frequency of the acoustic signal.

4.2. Simulation Setup. In this paper, we use MATLAB tosimulate the performance of H2-DAB, GEDAR, E-PULRP,and PER. Sensor nodes are randomly deployed in a 500m ×500m × 500m area. Since H2-DAB and GEDAR work ina network with multisink while E-PULRP and PER use onesink, in Figure 4, we show two examples of underwaternetwork topologies. Figure 4(a) is a network consisting ofone sink node and 150 ordinary nodes while Figure 4(b) isa network consisting of multisinks and 150 ordinary nodes.The sink node is assumed to be static deployed on the surfaceof water. The generation of packets in the network followsPoisson process with parameter 𝜆 = 0.0033. Ordinary nodescan drift with the water currents in horizontal direction.Mobility in other directions is not taken into account.IEEE 802.11 standard is adopted in MAC layer. Simulationparameters are listed in Table 2.

4.3. Performance Metrics. The average number of controlpackets per node is defined as the ratio of control packetsto the number of relay nodes. The end-to-end delay is thetime span from the routing establishment to the instant whenthe packets are received by sink. The data delivery ratio isthe percentage of packets which are successfully uploaded tothe sink. The total energy consumption is the sum of energy


Regular forwarding node

Source

Destination

(a) A packet forwarding tree without trim-ming mechanism

Source

Destination

Regular forwarding nodeThe only single forwarding node

(b) A packet forwarding tree with trimmingmechanism

Figure 3: Packet forwarding tree.

0 100 200 300 400500

0100200

300400

5000

100200300400500

Ordinary nodesSink

(a) A network consists of 1 sink node and 150 nodes

Ordinary nodesSink

0 100 200 300 400 500

0100

200300400

5000

100200300400500

(b) A network consists of 9 sink nodes and 150 nodes

Figure 4: Sensor node deployment in UASNs.

consumed for transmission and receiving in the process ofrouting establishment.

In our simulations, we design two groups of experimentsfor each evaluation criterion. In the first group, we simulatea stable underwater environment. In the second group, weintroduce the mobility and research the impact of watercurrents on the protocols.

4.4. Simulation Results

4.4.1. Static Environment. Figure 5 highlights the effect ofnodes on the average number of control packets. We changethe number of nodes from 150 to 350 with an incrementof 25. It can be observed that, with the increase of nodes,the average number of control packets of H2-DAB, GEDAR,and E-PULRP increases slowly. It can be explained that theseprotocols forward packets to the direction of destinationrather than in a random way so that the number of controlpackets can be restricted. PER shows a largest growth onthe average number of control packets because it does not

have definite forwarding direct so that additional controlpackets are needed to be used to establish the routing path.In addition, with the increase of nodes, muchmore neighborsshould be enquired.

Figure 6 shows the change of the end-to-end delay withdifferent number of nodes. Notably, when the network issparse (e.g., only 150 nodes), each protocol achieves high end-to-end delay. Because the bandwidth of acoustic channel islimited, the number of relay nodes is small and data packetsare accumulated in the forward queue which results in a highend-to-end delay. With the increase of nodes, such delay canbe reduced. Muchmore nodes function as relay nodes so thatpackets can be forwarded without waiting for a long time. Itcan be observed that, under the same circumstance, the E-PULRP and GEDAR achieve higher delay than that of PERand H2-DAB. This is because E-PULRP and GEDAR presetwaiting time to find forwarding nodes which increases thetime of routing discovery and establishment.

Figure 7 shows the change of the data delivery ratio withdifferent number of nodes. It can be observed that the data

International Journal of Distributed Sensor Networks 7Av

erag

e num

ber o

f con

trol p

acke

ts

GEDARE-PULRPPER

Number of nodes350325300275250225200175150

90

80

70

60

50

40

30

20

10

H2_DAB

Figure 5: Comparison of the average number of control packets pernode.

4

5

6

7

8

9

10

11

12

End-

to-e

nd d

elay

(s)

H2-DABGEDAR

E-PULRPPER

Number of nodes350325300275250225200175150

Figure 6: Comparison of end-to-end delay.

delivery ratio is small when the network consists of few sensornodes. With the supplement of nodes, much more nodesin the network mean many alternative routing paths can beestablished to ensure the success rate for data transmission.Moreover, the package loss ratio and the retransmission ratiocan be also reduced. We find out that E-PULRP outperformsother protocols because it sacrifices the route establishingtime to achieve reliable data forwarding. A backup route canbe established inH2-DAB to deal with the failed transmissionso that the performance of H2-DAB in terms of data deliveryratio is second only to E-PULRP.

Figure 8 shows the relationship between the total energyconsumption and the number of nodes. The total energyconsumption increases with the increase of nodes. Reasonscan be explained as follows. First, much more packets are

0.70

0.75

0.80

0.85

0.90

0.95

1.00

Dat

a del

iver

y ra

tio

H2-DABGEDAR

E-PULRPPER

Number of nodes350325300275250225200175150

Figure 7: Comparison of data delivery ratio.

H2-DABGEDAR

E-PULRPPER

Number of nodes350325300275250225200175150

600

500

400

300

200

100

0

Ener

gy co

nsum

ptio

n

Figure 8: Comparison of total energy consumption.

generated with the increase of nodes. It means that largeamount of energy is consumed to upload the additional pack-ets. Also, since a nodemight have muchmore neighbors withthe increase of nodes, the number of broadcast packets as wellas control packets increases. In this regard, communicationoverhead increases with the increase of nodes.

4.4.2. Dynamic Environment. We simulate a dynamic envi-ronment with different water velocities. Nodes are assumedto drift back to the network when they reach the boundary ofthe environment. In our simulation, we study the impact ofhorizontal movement on the protocols.

Figure 9 depicts the influence of the number of nodeson the average number of control packets in two dynamic


Aver

age n

umbe

r of c

ontro

l pac

kets

H2-DABGEDAR

E-PULRPPER

Number of nodes350325300275250225200175150

100

90

80

70

60

50

40

30

20

10

(a) Average number of control packets when V = 0.5m/s

H2-DABGEDAR

E-PULRPPER

Number of nodes350325300275250225200175150

Aver

age n

umbe

r of c

ontro

l pac

kets

100

90

80

70

60

50

40

30

20

10

(b) Average number of control packets when V = 0.8m/s

Figure 9: Comparison of the average number of control packets for each node at different water velocities.

H2-DABGEDAR

E-PULRPPER

End-

to-e

nd d

elay

(s)

Number of nodes350325300275250225200175150

14

13

12

11

10

9

8

7

6

5

(a) End-to-end delay when V = 0.5m/s

H2-DABGEDAR

E-PULRPPER

Number of nodes350325300275250225200175150

End-

to-e

nd d

elay

(s)

14

13

12

11

10

9

8

7

6

5

(b) End-to-end delay when V = 0.8m/s

Figure 10: Comparison of the end-to-end delay at different water velocities.

environments. It can be observed from Figures 9(a) and 9(b)that the average number of control packets increases with theincrease of nodes. Compared with static environment, morecontrol packets are required in a dynamic environment todescribe the change of the topology caused by node mobility.When the network consists of the same number of nodes,high-speed water currents contribute to a great increase inthe average number of control packets.

Figure 10 depicts the relationship between the end-to-end delay and the number of nodes in two dynamic envi-ronments. The end-to-end delay is reduced with the increaseof nodes. It can be observed that when the network consistsof the same number of nodes, low-speed water currents do

not bring about great increase in the end-to-end delay.This isbecause the location of nodes fluctuates slightly by the low-speed water; previous routes might be still valid. Thus, theend-to-end delay achieved by each protocol does not changegreatly.

Figure 11 shows the change of the data delivery ratio withvarious numbers of nodes in two dynamic environments.Thedata delivery ratio is increased with the increase of nodes. Ina dynamic environment, the data delivery ratio is decreasedby the effect of water currents. High-speed current resultsin a great change in the position of nodes. Because of thenode mobility, the topology of UASNs changes frequently.Thus, previous routing path is invalid and should be rebuilt


H2-DABGEDAR

E-PULRPPER

Number of nodes350325300275250225200175150

Dat

a del

iver

y ra

tio1.00

0.95

0.90

0.85

0.80

0.75

0.70

0.65

0.60

(a) Data delivery ratio when V = 0.5m/s

Dat

a del

iver

y ra

tio

H2-DABGEDAR

E-PULRPPER

Number of nodes350325300275250225200175150

1.00

0.95

0.90

0.85

0.80

0.75

0.70

0.65

0.60

(b) Data delivery ratio when V = 0.8m/s

Figure 11: Comparison of data delivery ratio at different water velocities.

Ener

gy co

nsum

ptio

n

H2-DABGEDAR

E-PULRPPER

Number of nodes350325300275250225200175150

0

800

700

600

500

400

300

200

100

(a) Total energy consumption when V = 0.5m/s

H2-DABGEDAR

E-PULRPPER

Number of nodes350325300275250225200175150

Ener

gy co

nsum

ptio

n

0

800

700

600

500

400

300

200

100

(b) Total energy consumption when V = 0.8m/s

Figure 12: Comparison of total energy consumption at different water velocities.

according to the updated information of network topology. Inthis regard, water velocity has negative effects on data deliveryratio.

Figure 12 shows the change of the total energy consump-tion with various numbers of nodes in two dynamic environ-ments. As shown in Figure 12, the total energy consumptionincreases with the increase of nodes. The supplement ofnodes helps to construct more available routing paths fordata transmission. However, the energy consumption is alsoincreased because of the additional data packets generatedin the network. Also, the network which suffers from high-speed current consumes more energy than that which suffers

from low-speed current. Both the position of nodes andneighbor relationship will be changed greatly by the high-speed current; this means that most of the previous routingpaths are invalid and should be rebuilt.

4.4.3. Comparison Summary. Compared with the static envi-ronment, additional control packets are required to ensurethe operation of each protocol in the dynamic environment.With the increase of water velocity, all the protocols showa worse performance in terms of end-to-end delay, datadelivery ratio, and energy consumption. Since the network


topology is greatly changed by high-speed current, mostprevious routing paths constructed by the protocols areinvalid.The reconstruction of routing brings about additionalrouting time and energy consumption.

5. Conclusion and Future Work

Due to the unique characteristics of UASNs, traditionalrouting protocols in TWSNs expose many drawbacks inUASNs. In this paper, we introduce four routing algorithmsfor large-scale UASNs and compare their performances interms of average number of control packets, end-to-enddelay, data delivery ratio, and total energy consumption. Sim-ulation results show that E-PULRP provides high datadelivery ratio at the cost of end-to-end delay. H2-DAB hasbetter real-time performance forminimal delay transmission.GEDAR can efficiently address the problem of void regionwithout introducing extra energy. PER requires the mostcontrol packets in the process of routing establishment. Fourprotocols show similar performance in terms of total energyconsumption. We also study the impact of current on theprotocols. We find out that both H2-DAB and GEDAR takeinto account the node mobility; they can be managed easilyduring the quick routing changes where node movementsare frequent. However, high-speed current still degrades theirperformances to somedegree. E-PULRP andPER can achievehigh energy efficiency while little attention is paid to the nodemobility. Thus, they are not robust to the change of networktopology caused by node mobility. Through revealing thecharacteristics of each protocol, our work aims to provideuseful insights to select appropriate routing protocols to fulfildifferent application requirements in UASNs.

Further research should consider other network perfor-mance criteria such as the quality of service (QoS) issuesposed by real-time applications and secure routing requiredby military applications. Nonetheless, with the increasingfunctionalities available to an underwater sensor node, morecomplicated tasks which involve more energy consumptionand network overhead may be assigned to the sensor nodes.Thus, how to increase energy efficiency and scalability of thenetwork remains a challenging research area.

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper.

Acknowledgments

The work is supported by Qing Lan Project, Natural Sci-ence Foundation of Jiangsu Province of China (no.BK20131137), 2013 Special Fund of GuangdongHigher SchoolTalent Recruitment, Educational Commission of GuangdongProvince, China (Project no. 2013KJCX0131), GuangdongHigh-Tech Development Fund (no. 2013B010401035), 2013Top Level Talents Project in “Sailing Plan” of GuangdongProvince, National Natural Science Foundation of China

(Grants nos. 61572172 and 61401147), and 2014 GuangdongProvince Outstanding Young Professor Project.

References

[1] G. Acar and A. E. Adams, “ACMENet: an underwater acousticsensor network protocol for real-time environmental monitor-ing in coastal areas,” IEE Proceedings: Radar, Sonar and Nav-igation, vol. 153, no. 4, pp. 365–380, 2006.

[2] T. Hu and Y. Fei, “QELAR: a machine-learning-based adaptiverouting protocol for energy-efficient and lifetime-extendedunderwater sensor networks,” IEEE Transactions on MobileComputing, vol. 9, no. 6, pp. 796–809, 2010.

[3] J. Kong, J. Cui, D. Wu, and M. Gerla, “Building underwateradhoc networks and sensor networks for large scale real-timeaquatic applications,” in Proceedings of the Military Communi-cations Conference (MILCOM ’05), Atlantic City, NJ, USA, 2005.

[4] G. Han, C. Zhang, L. Shu, and J. J. P. C. Rodrigues, “Impacts ofdeployment strategies on localization performance in underwa-ter acoustic sensor networks,” IEEE Transactions on IndustrialElectronics, vol. 62, no. 3, pp. 1725–1733, 2015.

[5] S. He, X. Li, J. Chen, P. Cheng, Y. Sun, and D. S. Ryl, “EMD:energy-efficient P2P message dissemination in delay-tolerantwireless sensor and actor networks,” IEEE Journal on SelectedArea in Communications, vol. 31, no. 9, pp. 75–84, 2013.

[6] S. He, J. Chen, D. K. Y. Yau, and Y. Sun, “Cross-layer optimiza-tion of correlated data gathering in wireless sensor networks,”IEEE Transactions onMobile Computing, vol. 11, no. 11, pp. 1678–1691, 2012.

[7] Y. Bayrakdar,N.Meratnia, andA.Kantarci, “A comparative viewof routing protocols for underwater wireless sensor networks,”in Proceedings of the OCEANS 2011, IEEE, Santander, Spain,2011.

[8] J. Rezazadeh, M. Moradi, A. S. Ismail, and E. Dutkiewicz,“Impact of static trajectories on localization in wireless sensornetworks,”Wireless Networks, vol. 21, no. 3, pp. 809–827, 2015.

[9] Y. Wang, “Three-dimensional wireless sensor networks: geo-metric approaches for topology and routing design,” in TheArt of Wireless Sensor Networks, pp. 367–409, Springer, Berlin,Germany, 2014.

[10] G. Han, C. Zhang, T. Liu, and L. Shu, “MANCL: a multi-anchor nodes collaborative localization algorithm for under-water acoustic sensor networks,”Wireless Communications andMobile Computing, 2014.

[11] H. Yu, N. Yao, and J. Liu, “An adaptive routing protocol inunderwater sparse acoustic sensor networks,”AdHoc Networks,23 pages, 2014.

[12] E. M. Sozer, M. Stojanovic, and J. G. Proakis, “Underwateracoustic networks,” IEEE Journal of Oceanic Engineering, vol. 25,no. 1, pp. 72–83, 2000.

[13] M. Ayaz, A. Abdullah, I. Faye, and Y. Batira, “An efficientdynamic addressing based routing protocol for underwaterwireless sensor networks,” Computer Communications, vol. 35,no. 4, pp. 475–486, 2012.

[14] R. W. L. Coutinho, A. Boukerche, L. F. M. Vieira, and A. A. F.Loureiro, “GEDAR: geographic and opportunistic routing pro-tocol with depth adjustment for mobile underwater sensornetworks,” in Proceedings of the 1st IEEE International Confer-ence on Communications (ICC ’14), pp. 251–256, IEEE, Sydney,Australia, June 2014.


[15] S. Gopi, K. Govindan, D. Chander, U. B. Desai, and S. N.Merchant, “E-PULRP: energy optimized path unaware layeredrouting protocol for underwater sensor networks,” IEEE Trans-actions onWireless Communications, vol. 9, no. 11, pp. 3391–3401,2010.

[16] Z. Guo, B.Wang, and J.-H. Cui, “Prediction assisted single-copyrouting in underwater delay tolerant networks,” in Proceedingsof the 53rd IEEE Global Communications Conference (GLOBE-COM ’10), pp. 1–6, IEEE, Miami, Fla, USA, December 2010.

[17] H. Yan, Z. J. Shi, and J. Cui, “DBR: depth-based routing forunderwater sensor networks,” in NETWORKING 2008 Ad Hocand Sensor Networks, Wireless Networks, Next Generation Inter-net, pp. 72–86, Springer, Berlin, Germany, 2008.

[18] M. R. Jafri, S. Ahmed, N. Javaid, Z. Ahmad, and R. J. Qureshi,“AMCTD: adaptive mobility of courier nodes in threshold-optimized DBR protocol for underwater wireless sensor net-works,” in Proceedings of the IEEE 8th International Conferenceon Broadband, Wireless Computing, Communication and Appli-cations (BWCCA ’13), pp. 93–99, IEEE, October 2013.

[19] G. Liu and Z. Li, “Depth-based multi-hop routing protocol forunderwater sensor network,” in Proceedings of the 2nd Interna-tional Conference on Industrial Mechatronics and Automation(ICIMA ’10), pp. 268–270, IEEE, May 2010.

[20] Y.-S. Chen and Y.-W. Lin, “Mobicast routing protocol for under-water sensor networks,” IEEE Sensors Journal, vol. 13, no. 2, pp.737–749, 2013.

[21] Z. Zhang, S.-L. Lin, and K.-T. Sung, “A prediction-based delay-tolerant protocol for underwater wireless sensor networks,” inProceedings of the International Conference on Wireless Com-munications and Signal Processing (WCSP ’10), pp. 1–6, IEEE,October 2010.

[22] H.Wu,M.Chen, andX.Guan, “Anetwork coding based routingprotocol for underwater sensor networks,” Sensors, vol. 12, no.4, pp. 4559–4577, 2012.

[23] C.-J. Huang, Y.-W. Wang, H.-H. Liao, C.-F. Lin, K.-W. Hu, andT.-Y. Chang, “A power-efficient routing protocol for underwaterwireless sensor networks,” Applied Soft Computing Journal, vol.11, no. 2, pp. 2348–2355, 2011.

[24] T. Ali, L. T. Jung, and S. Ameer, “Flooding control by usingangle based cone forUWSNs,” inProceedings of the InternationalSymposium on Telecommunication Technologies (ISTT ’12), pp.112–117, IEEE, Kuala Lumpur, Malaysia, November 2012.

[25] X.-J. Du, K.-J. Huang, S.-L. Lan, Z.-X. Feng, and F. Liu,“LB-AGR: level-based adaptive geo-routing for underwatersensor network,” The Journal of China Universities of Posts andTelecommunications, vol. 21, no. 1, pp. 54–59, 2014.

[26] A. Wahid and D. Kim, “An energy efficient localization-freerouting protocol for underwater wireless sensor networks,”International Journal of Distributed Sensor Networks, vol. 2012,Article ID 307246, 11 pages, 2012.

[27] S. Lee, B. Bhattacharjee, and S. Banerjee, “Efficient geographicrouting inmultihopwireless networks,” in Proceedings of the 6thACM International Symposium on Mobile Ad Hoc Networkingand Computing, pp. 230–241, May 2005.

[28] S. Gop, G. Kannan, D. Chander, U. B. Desai, and S. N.Merchant,“Path unaware layered routing protocol (PULRP) with non-uniform node distribution for underwater sensor networks,”Wireless Communications and Mobile Computing, vol. 8, no. 8,pp. 1045–1060, 2008.

[29] X. Zhu, J. Wang, H. Yan, and S. Wu, “Research and applicationof the improved algorithmC4.5 on decision tree,” in Proceedings

of the International Conference on Test andMeasurement (ICTM’09), pp. 184–187, IEEE, December 2009.

[30] J. Jiang, G.Han,H.Guo, L. Shu, and J. J. Rodrigues, “Geographicmultipath routing based on geospatial division in duty-cycledunderwater wireless sensor networks,” Journal of Network andComputer Applications, 2015.

[31] L. M. Berkhovskikh and P. L. Lysanov, Fundamentals of OceanAcoustics, Springer Science & Business Media, 2003.

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttp://www.hindawi.com Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014


Active and Passive Electronic Components

Control Scienceand Engineering

Journal of



RotatingMachinery


Hindawi Publishing Corporation http://www.hindawi.com

Journal ofEngineeringVolume 2014

Submit your manuscripts athttp://www.hindawi.com

VLSI Design



Shock and Vibration


Civil EngineeringAdvances in

Acoustics and VibrationAdvances in



Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation http://www.hindawi.com

Volume 2014

The Scientific World JournalHindawi Publishing Corporation http://www.hindawi.com Volume 2014

SensorsJournal of


Modelling & Simulation in EngineeringHindawi Publishing Corporation http://www.hindawi.com Volume 2014


Chemical EngineeringInternational Journal of Antennas and

Propagation




Navigation and Observation



DistributedSensor Networks


research article routing protocols in underwater acoustic...

Documents