ResearchArticle Hybrid Localization Approach for ...

Research ArticleHybrid Localization Approach for Underwater Sensor Networks

Pei-Hsuan Tsai Rong-Guei Tsai and Shiuan-ShiangWang

Institute of Manufacturing Information and System National Cheng Kung University Tainan City 70101 Taiwan

Correspondence should be addressed to Pei-Hsuan Tsai peihsuantsaigmailcom

Received 1 June 2017 Revised 5 September 2017 Accepted 24 September 2017 Published 2 November 2017

Academic Editor Hana Vaisocherova

Copyright copy 2017 Pei-Hsuan Tsai et alThis is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

Underwater Wireless Sensor Networks (UWSNs) are widely used to collect data in the marine environment Location and timeare essential aspects when sensors collect data particularly in the case of location-aware data Many studies on terrestrial sensornetworks consider sensor locations as the locations where data is collected and focus on sensor positioning when sensors arefixed However underwater sensors are mobile networks and the sensor locations change continuously Localization schemesdesigned for static sensor networks need to run periodically to update locations and consume considerable sensor power andincrease the communication overhead hence they cannot be applied toUWSNsThis paper presents a hybrid localization approachwith data-location correction called Data Localization Correction Approach (DLCA) which positions data without additionalcommunication overhead and power consumption on sensors Without loss of generality we simulate the ocean environmentbased on a kinematic model and meandering current mobility model and conduct extensive simulations Our results show thatDLCA can significantly reduce communication costs while maintaining relatively high localization accuracy

1 Introduction

The ocean is a vast natural resource with areas that are yetto be thoroughly explored With the development of oceanengineering and network technology the ocean has becomea focus for research and underwater sensor networks havereceived considerable attention from researchers Applica-tions of underwater sensor networks range from the oilindustry to aquaculture and include instrument monitoringpollution control climate recording prediction of naturaldisturbances search and survey missions and the study ofmarine life [1ndash4] A node must know its own location beforesending data to its neighbor The need for location arisesbecause the number of nodes is very large and it is notpossible for the base station to find the nodesrsquo positions sothe individual node is required to send location informationalong with the observed data to provide exact location to theuser which means the node must localize itself

Communication and collaboration among nodes areessential to assist node self-localization In most localizationalgorithms nodes collaborate with each other by consideringseveral aspects like limited energy resources number anddensity of nodes and existence of obstacles Compared toterrestrial sensor networks underwater sensor networks face

new communication challenges (1) underwater communica-tion systems today mostly use acoustic technology becauseelectromagnetic waves cannot propagate over long distancesin water acoustic communications offer longer ranges butwith large latency limited bandwidth and time-varyingmultipath propagation and (2) underwater sensor networksare dynamic networks and sensor locations change contin-uously because of ocean currents However observed data istypically interpretedwith reference to the sensorrsquos position Inotherwords sensor locations are considered as data locationsIn such environments localization schemes designed forstatic networks need to run periodically to update sensorlocations resulting in communication overheads and sensorpower consumption In a harsh aqueous environment sensormobility makes underwater sensor networks the most chal-lenging of all network scenarios

Although the network conditions in underwater envi-ronments make localization difficult a useful property wefound is that in many applications such as environmentalmonitoring sensors periodically report their observed datato the base station For these applications location informa-tion is only useful at those discrete time points Thereforea heuristic approach to reduce communication and energycosts while maintaining the localization accuracy is to set

HindawiJournal of SensorsVolume 2017 Article ID 5768651 13 pageshttpsdoiorg10115520175768651

2 Journal of Sensors

1T 2T 3T0

Pd

Heuristic Pl

DLCA Pl

(n minus 1)T nT

middot middot middot


Figure 1 Comparison of period length

the sensor self-localized period (119875119897) equal to the data observedperiod (119875119889) The top line and middle line in Figure 1 denotethe data observed period (119875119889) and the sensor self-localizedperiod (119875119897) of the heuristic approach respectively Howeverthe downside of the heuristic approach is that when 119875119889 issmall it results in relatively high communication costs beinggenerated Due to the limited bandwidth of acoustic channelsused by underwater sensor networks this will be a largeburden on the network

Based on this motivation we aim to design a hybridlocalization approach for underwater sensor networks calledDLCA which takes full advantage of computing the data-location at a base station when data packets are receivedinstead of relying on sensors continuously self-localizing Inother words DLCA extends 119875119897 to multiples of 119875119889 as shownby the bottom line in Figure 1The whole localization processof DLCA is divided into two parts node localization andobserved data-location correction Node localization is run bythe nodes themselves and observed data-location correctionis run by the base station Simulation results show that DLCAcan greatly reduce localization communication costs whilemaintaining relative localization accuracy

This work makes two contributions First our workis unique in aiming at localization of location-aware datainstead of sensor localization in underwater sensor networksand provides a practical system with hybrid localizationThe second contribution is the design of post facto locationcorrection on the base station that improves localizationaccuracy without additional communication overheads andpower consumption on sensor nodes

The remainder of this paper is organized as follows InSection 2 we provide background information includingseveral distance acoustic measurement models time syn-chronization and communication schemes for UWSNs InSection 3 we review related works The design of DLCAincluding its network architecture data structure and funda-mental algorithms are presented in Section 4 the simulationresults are presented in Section 5 and Section 6 contains ourconclusions

2 Background

Underwater sensor networks comprise a large number ofsmall devices deployed in a physical underwater environ-ment Each node has special capabilities such as wirelesscommunicationwith its neighbors sensing data data storageand processing Today there are more than 50 localizationalgorithms in existence [5ndash7] These can be classified on thebasis of different aspects

Centralized Localization versus Distributed Localization Incentralized localization one central base station computes

the locations of unknown nodes while in distributed local-ization computation is done by the sensor itself and nodescommunicate with each other to obtain their position in thenetwork DLCA uses hybrid localization The locations ofsensors are computed by the sensors themselves to supportreal-time monitoring but the data locations are corrected bya base station to reduce nodes power consumption

Anchor Based versusAnchorlessThe anchor-based algorithmsuse anchor nodes (which know their own position fromprior GPS data) as reference nodes for localization Themore anchor nodes the higher localization accuracy butthe cost also increases as anchor nodes are equipped withextra resourcesThe anchorless schemesmeasure the distancebetween nodes for creating a local map of the nodes How-ever the distance measurement techniques to date have notbeen accurate so global coordinates are preferred over localcoordinates for most applications Therefore DLCA utilizesthe anchor-based localization

Range-Based versus Range-Free Range-based systems usetechniques such as ultrasound to measure the distancebetween nodes and then triangulation to compute the posi-tions of nodes Range-free techniques use implicit informa-tion provided by anchor nodes to obtain positions of nodessuch as number of hops between devices or radio coveragemembership [8ndash10] Although range-free protocols do notneed additional hardware for distance measurements theycan only provide rough positional estimates DLCA is moreinterested in accurate localization and thus DLCA adoptsrange-based schemes

Several distance measuring techniques such as receivedsignal strength indication (RSSI) time of arrival (ToA) andtime difference of arrival (TDoA) are widely used in UWSNsfor localization Among them ToA in particular is the mostcommonly used for UWSNs as it is more accurate [1ndash4] It isnot affected by channel fading but has an issue in achievingsynchronization between nodes It sends a single packet fromone node to another containing the time of its transmissionassuming perfect clock synchronization between nodes Thereceiving node knows when the packet arrived and if it issynchronizedwith the sender node the distance travelled canbe calculated using the following equation

119889ToA = 119888 lowast Δ119905ToA (1)

where 119888 is the speed of sound in seawater (119888 = 1500ms) andΔ119905ToA is the time of arrival (Δ119905ToA = receiver time ndash sendertime)

However sound waves can have transmission delays andvarying transmission ratesThe delaysmake the ToA distancelonger than the actual distance and the longer the realdistance the more the error deviation of the ToA distance[11ndash13] Equations (2) and (3) simulate the difference betweenactual distance and ToA distance

119889ToA = 119889Real + 120576 (2)

120576 = 119889Real times 119873(120583 1205902) (3)

Journal of Sensors 3

where 119889Real is the real distance between two nodes and ToAdistance 119889ToA is the real distance with an error 120576 simulatedby Gaussian distribution calculated in (3)

Synchronized time is the prerequisite of using ToA toestimate distance and therefore many localization algorithmsrely on the time synchronization services A synchronizationalgorithm for UWSNs must consider additional factors suchas long propagation delays from the use of acoustic communi-cation and sensor node mobility Many time synchronizationapproaches have been proposed for different network topolo-gies [14ndash17] The approach in [17] is designed for the samenetwork topology as ours Their approach solves localizationand time synchronization jointly to save energy and toimprove the accuracy of both services by utilizing the mes-sage exchanges among the nodes We will apply and evaluatethe joint scheme performing the node localization and timesynchronization simultaneously at every 119875119897 so that the clockdrift can be limited to a tolerable range by setting the length of119875119897 to similar network scenariosThe results will be reported ina future paper In this paper we assume that all nodes are timesynchronized

Unlike terrestrial networks which mainly rely on radiowaves for communication UWSNs utilize acoustic wavesfor information exchange To provide high throughput inan energy-efficient way it is important for UWSNs to havean efficient Medium Access Control (MAC) protocol thatallows the nodes to share a common broadcast channel andto prevent simultaneous transmissions or resolve collisions ofdata packets while providing energy efficiency low channelaccess delays and fairness among the nodes However theunderwater acoustic environment poses difficulties for MACprotocol design for example high and variable propagationdelay limited bandwidth and data rate noise and energyconsumption In addition network topology and deploy-ment highly affect the performance of the MAC protocolContention-free MAC protocols are not good solutions forour network topology [18] because they are either dedicatedto source-to-destination packet exchange without supportingbroadcasting or may not yield better performance thanrandom access approaches owing to the long and varyingpropagation latency of the underwater acoustic channel Inaddition our approach DLCA can tolerate packet collisionAlthough the higher the node density the higher the prob-ability of collision as long as there are neighbors our data-location correction approach is applicable and therefore themore the neighbor nodes the higher the collision toleranceIn this paper we assume that the contention-basedMACpro-tocol (eg random access) is adopted for communications In[19] a collision-tolerant packet scheduling was proposed forunderwater acoustic localization They assumed that duringa localization period the reference nodes transmit randomlyfor example according to a Poisson distribution with anaverage transmission rate of 120582 packets per second We willapply and evaluate the collision-tolerant algorithm describedin [19] with our scheme and report the results in a futurepaper

3 Related Work

A tremendous amount of localization schemes has beenreported for terrestrial WSNs [5ndash7] However the uniquecharacteristics of the underwater sensor network environ-ment for example sensor node mobility long propagationdelay and high power consumption in transmit and receivemodes make the existing WSN algorithms inefficient forUWSNs To overcome these issues modifications to WSNlocalization schemes or different alternatives have been pro-posed [1ndash4]

According to [4] UWSN localization techniques canbe divided into two categories centralized and distributedtechniques The main advantage of the centralized techniqueis reducing the computational burden of the underwatersensor nodes The major drawback of the scheme is notsupporting real-time location information and requiring highcommunication overhead and high energy consumption insending localization related messages to an access pointwhich is underwater or on the surface In contrast themain advantage of the distributed technique is to supportapplications that need real-time location information forexample online monitoring and coordinated motion Themajor drawbacks of the scheme are high energy consumptionand the communication and computational burdens on thesensor nodes

The design of the localization techniques highly dependson the network topology and applications We describe herethe most recent localization techniques with similar networktopologies as ours

31 Distributed Localization Techniques In distributed local-ization techniques the sensor nodes need to collect anchorpositions compute the distance to anchors or neighbors andrun location estimation algorithms for self-positioning LDB[9 10] AAL [20 21] DNRL [22 23] and MobiL [24] focuson how anchors provide their positions to the sensor nodesand assume that the anchors are mobile and equipped withGPS to obtain their locations Their disadvantages includehigh localization delays and high cost as the accuracy ofthe location relies on a large number of mobile beaconsespecially for large-scale UWSNs

Large-Scale Hierarchical Localization (LSHL) exploresthe localization problem in large-scale UWSNs and providesa hierarchical approach to divide the localization processinto anchor and ordinary node localizations [25 26] Itassumes that an anchor can get perfect location estimationand focus on ordinary node localization The drawback ofLSHL is having a high energy consumption and communi-cation overhead due to beacon exchanges Many researches[27ndash30] are proposed to improve LSHL including SLMPand our work based on the same network topology andassumptions

SLMP is a prediction-based localization scheme [27]Theanchor locations are estimated by either using trilaterationwith surface buoy coordinates or runningmobility predictionalgorithms The ordinary nodes use the mobility pattern topredict their locations and the pattern is assumed valid untilan update from an anchor node is receivedThe approaches in


[28ndash30] are designed to improve the accuracy ofmobility pre-diction algorithms However the communication overheadand energy consumption depend on the mobility pattern forthe ordinary node

32 Centralized Localization Techniques Centralized tech-niques calculate the location of each sensor node in acommand center or sink and the sensor nodes do notknow their locations unless the sink node explicitly sendstheir information Therefore they are not convenient forapplications that require accurate and real-time locationinformation In addition most centralized techniques are notapplicable to large-scale UWSNs for example SLMP [27]assumes that data are sent via wired communications

Collaborative Localization (CL) [31] is anchor-freeand the nodes collaborate to determine their positioningautonomously without using surface buoys or ships Thesensor nodes are categorized as profilers or followers Aprofiler travels to a depth first and its trajectory is used as aprediction of the future location of the followers Howeverit initially assumes that all nodes are localized by GPSand the coverage and accuracy of localization depend onthe trajectories of the followers that have to be the sameas the profilers which are not applicable to large-scaleUWSNs

Furthermore [32 33] target applications where therelation between the data and location is resolved at thepostprocessing stage by a central station The sensor nodecollects distance estimates between itself and its neighborsand then all distance estimations are sent to a centralstation and processed offline An iteration algorithm is usedto obtain the positions At each iteration the algorithmrefines the positions of the sensor nodes The drawbacksare long localization time and huge energy of the sink Inthis paper we propose a vector-based approach to improvethese

In this paper we propose a hybrid localization approachto support real-time location information when it is neededand to reduce the computational burden on sensor nodesby post facto correcting the location of observed data Thecommunication overhead and energy consumption dependon the accuracy and promptness of location demanded bythe applications For an acceptable location error toleranceour approach DLCA decreases the number of localizationupdates and consequently its communication overhead andenergy consumption is low

4 Design of DLCA

In this section we describe the design of DLCA a data-location correction approach for underwater sensor net-worksWe first present the network architecture and then thealgorithms governing the model followed by an example

41 Network Topology TheDLCAnetwork architecture com-prises four types of nodes as shown in Figure 2 surfacebuoys anchor nodes sensor nodes and a base station Surfacebuoys are nodes deployed on the sea surface and equippedwith GPS to obtain their absolute location from GPS The

Sensor node

Base station

Anchor node

Surface buoy

Figure 2 Network architecture of underwater sensor networks

anchor nodes communicate with surface buoys and obtaintheir locations through the GPS intelligent buoy system [3435] Therefore we assume that the locations of surface buoysand anchor nodes are sufficiently accurate in this paperNeither surface buoys nor anchor nodes are equipped withsensors and their main functions are assisting the localizingof sensor nodes and transferring data from sensor nodesThe sensor nodes are those nodes that cannot communicatedirectly with the surface buoys because of distance or otherconstraints but can communicate with the anchor nodesto estimate their own locations Through message passinglocalized sensor nodes can assist other nonlocalized sensornodes to estimate their locations [8 36] The base station isthe node to centralize and to integrate all the data transferredby surface buoys

42 Overview of DLCA Here are the assumptions made inthe design of the DLCA

(1) All nodes are time synchronized initially Then atevery 119875119897 the time synchronization is jointed withnode localization by adopting [17] and therefore weassume that all nodes are time synchronized

(2) To avoid packet collision all nodes transmit sepa-rately for example according to a Poisson distribu-tion with an average transmission rate of 120582 packetsper second Collision-tolerant packet scheduling [19]can be exploited to achieve a desired probability ofsuccessful self-localization for a given number 119873 ofanchors while 120582 and the minimum localization timecan be determined

(3) To avoid broadcast storming the sensor node onlybroadcasts the packet with its observed data but


forwards the received data packets to its referencenodes The reference nodes are updated every 119875119897

(4) Initially only surface buoys and anchor nodes canfix their locations using GPS a GPS intelligent buoysystem or other means

(5) All the sensor nodes are functionally identical

421 Node Localization Again DLCA is a hybrid approachcomprising two parts node localization which is run in thenodes and data-location correction which is run at the basestation The sensor nodes initialize and periodically updatetheir locations as follows

To estimate their locations the sensor nodes stamp thesending time 1198791 and then immediately broadcast a Reqmessage to their neighboring reference nodes The sensornodes then listen to the localization messages from any otherlocalized nodes

Upon receiving the Req message each reference nodeimmediately marks its local time as 1199052 and then after a timeinterval 119905119903 (avoiding collision) each of the reference nodesends back a Resmessage containing its location information1199052 and sending time 1199053 When receiving the Resmessage theordinary node marks its receiving time 1198794

In each period 119875119897 the unlocalized or location-expiredsensor nodes that receivedmore than three locationmessagesperform self-localization by multilateration [24 25 27] andtime synchronization by Mobi-Sync [14] The successfullylocalized and time synchronized sensor nodes start to observedata and then transmit data packets with their locationinformation

422 Data Packet Format for Data-Location CorrectionDLCA relies on the information embedded in data packetsto do data-location correction To achieve a good balancebetween packet length and sufficient localization informa-tion the data packet format is only slightly changed as shownin Figure 3

(i) Observer ID stores the ID of the sensor that observesthe data and broadcasts this data packet

(ii) Observed data stores the data content observed by thesensor node at each 119875119889 We assume here that sensorsobserve data periodically

(iii) Location stores the location of the observer node

(iv) Observed time stores the time when the data isobserved We assume here that the sensor broadcaststhe data packet right after the data is observed

The former four fields are essential for all UWSN appli-cations Two extra fields are added to support data-locationcorrection ldquoreceiver IDrdquo and ldquoreceived timerdquo

(i) Receiver ID stores the ID of the node that is the firstreceiver of this data packet

(ii) Received time stores the time when the first receiverreceives this data packet

ObserverID

Observeddata Location Observed

timeReceiver

IDReceived

time

Figure 3 Data packet format

Anchor nodes

Sensor nodes

48

6

7

9

5

2

1

3

Figure 4 An example of DLCA graph

Table 1 Data packets grouped with identical send time which isequal to 3

ObserverID


timeReceiver

IDReceivedtime

4 (52 29) 3 1 3006574 (52 29) 3 2 3007124 (52 29) 3 5 3007804 (52 29) 3 6 3005554 (52 29) 3 8 3002275 (56 35) 3 1 3007175 (56 35) 3 2 3002255 (56 35) 3 3 3005976 (44 31) 3 7 3003527 (39 36) 3 8 3006558 (43 29) 3 6 3003869 (41 45) 3 3 300757

After several transmissions data packets will be transmit-ted to the base stationThebase station then classifies receiveddata packets according to their observed time Packets withthe same or close observed time are grouped together Table 1illustrates an example of grouping received data packets withidentical observed time= 3When the number of data packetsin a group is sufficient or a predetermined correct timeis achieved [32] the base station starts the data-locationcorrection procedure

423 Data Structure of DLCA Data packets within the samegroup can be illustrated schematically on a DLCA graph Forexample Figure 4 shows the DLCA graph of Table 1 In theDLCAgraph there are sensor nodes and anchor nodeswithinthe same packet table For observer node 119901 and receiver node


Table 2 DLCA table

119899id Location cvloc(119899id) Neighbor list119899119887id 119889ToA(119899id 119899119887id) cvToA(119899id 119899119887id) 119889Euc(119899id 119899119887id)

119902 in a data packet we denote the two nodes 119901 and 119902 asneighbor nodes to each other and use an undirected edge119890(119901 119902) to show their relationship in the DLCA graph

A DLCA table is used to maintain information of theDLCA graph as shown by Table 2 The fields of the DLCAtable from left to right are node ID data packet locationaccuracy of the data packet location information of theirneighbor nodes including the neighbor nodersquos ID ToAdistance between the node and its neighbor accuracy ofToA distance and Euclidean distance between the node andits neighbor We assume that the base station has all theinformation of anchor nodes including the locations andaccuracies

424 DLCA Table Initialization Each node on the DLCAgraph is an entry in the DLCA table Initially the data packetlocation is set as the location after node self-localization every119875119897 The accuracy of every 119875119889 which is the confidence valueof the data packet location at every 119875119889 is given by 1(1 + 119899)where 119899 ranges from 0 to (119875119897119875119889 minus 1) as the accuracy of thelocation will decrease with time and 119875119897 can be a multiple of119875119889 119889ToA(119899id 119899119887id) represents the ToA distance between 119899idand its neighbor 119899119887id The length is computed based on (1)cvToA(119899id 119899119887id) represents the confidence value of the ToAdistance The confidence value of each ToA distance is setbased on the following equation

cvToA (119899id 119899119887id)

= 0 119889ToA (sdot sdot sdot ) gt 1198771 minus 119889ToA (119899id 119899119887id)119877 otherwise

(4)

where 119877 is the communication radius of the sensor nodeThe confidence value of the ToA distance is set to zero whenthe distance is larger than 119877 as it is unreasonable that themeasured distance is larger than the communication radius

The Euclidean distance 119889Euc(119899id 119899119887id) between 119899id and119899119887id is computed based on the two nodesrsquo locations andfollows (5) The 119889Euc(119899id 119899119887id) is updated when the locationof the data packet is updated

119889 (119901 119902) = 119889 (119902 119901)= radic(1199021 minus 1199011)2 + (1199022 minus 1199012)2 + (1199023 minus 1199013)2

(5)

Here we use the same example to illustrate the DLCAtable initialization In Table 3 nodes with number (1 2 3) areanchor nodes Their confidence values are 1 because they areassumed to be accurate Their neighbor lists are null because

they do not need other nodes to correct their locations Theremaining nodes in the DLCA table are sensor nodes In thisexample 119875119889 is set as 1The confidence value of the data packetlocation is 025 because the last sensor self-localized time is 0and the data packet observed time is 3 The neighbor nodesof node 4 include nodes 1 2 5 6 and 8 The ToA distancebetween nodes 1 and 4 is 986 because the data packet is sentat time = 3 and the data packet was first received by node 1at 300659 and the speed of sound in seawater is 1500msThe communication radius is 12 so the confidence value of theToA distance is 018 The Euclidean distance between node 4and node 1 is the distance between (52 29) and (5724 2768)

43 Recursively Correcting Data Packet Locations

431 Find Victim Node 119904V Entries with a confidence valueless than a designated threshold 120579victim are considered ascorrectable candidates The base station will continuallychoose one of the candidates to see if it is qualified to be thevictim node 119904V in short

From the example of Figure 4 the base station first selectsnode 4 which has sufficient neighbors For each neighbornode of node 4 the base station computes its reliable valueThe reliable value computed by (6) is used to quickly choosethe four most reliable neighbor nodes as reference nodes 119903id

reliablevalue = cvloc (119899119887id) lowast cvToA (119899id 119899119887id) (6)

To see if the four reference nodes are reliable enoughto correct the data packet location the base station thencomputes the new confidence value of node 4 after beingcorrected by reference nodes based on

cvloc (119904V)= 1

4 sum[cvloc (119903id) lowast (1 minus 120572) + cvToA (119899id 119903id) lowast 120572] (7)

120572 = cvToA (119899id 119903id)cvloc (119903id) + cvToA (119899id 119903id) (8)

DLCA adopts the average confidence values of the fourreferences nodes as the confidence value of 119904V as the locationof 119904V will be computed according to the four reference nodes120572 which is called adjustment parameter is used to adjustthe weight of the ToA distance confidence value When theToA distance is more accurate than the locations of neighbornodes 120572 will be greater than 05

If cvloc(119904V) is higher than cvloc(119899id) the node is classifiedas a victimnode 119904V and its location is computed by themethoddescribed in the next section Otherwise the base station


Table 3 Initialization of DLCA table


1 (5724 2768) 1 NA2 (5447 4040) 1 NA3 (5583 4711) 1 NA

4 (52 29) 025

1 986 018 542 1068 011 11665 117 002 7216 832 031 8258 341 072 9

5 (56 35) 025

1 1076 01 7422 337 072 5613 896 025 12114 117 002 721

6 (44 31) 0254 832 031 8257 528 056 7078 579 052 224

7 (39 36) 025 6 528 056 7078 982 018 806

8 (43 29) 0254 341 072 96 579 052 2247 982 018 806

9 (41 45) 025 3 1136 005 1498

14

2

6

8Shi_vector

d４Ｉ (s r1 )

dＯ＝ (s r1 )

Figure 5 119889ToA(119904V 119903id) gt 119889Euc(119904V 119903id)

skips this node and selects another candidate to repeat theprocedure of selecting the candidate 119904V The recursive victimnode selection stopswhen there are no candidates left Table 4shows the result of victim selection

432 Compute Location of Victim Node by Shift VectorsDLCA uses the ToA distance and the Euclidean distance todecide the shift vector of the victim node However there aretwo conditions of the shift vector

Condition 1 119889ToA(119904V 119903id) gt 119889Euc(119904V 119903id) is shown in Figure 5In this condition the victim node 119904V (node 4) will be adjustedfar from its reference node 1

2

41

6

8dＯ＝(s r8)

d４Ｉ(s r8)

Shi_vector

Figure 6 119889ToA(119904V 119903id) lt 119889Euc(119904V 119903id)

Condition 2 119889ToA(119904V 119903id) lt 119889Euc(119904V 119903id) is shown in Figure 6In this condition the victim node 119904V (node 4) will be adjustedcloser to its reference node 8

Equation (9) calculates the shift vector

997888V119894 = 120572 lowast (119889ToA (119904V 119903id) minus 119889Euc (119904V 119903id)) lowast997888997888997888997888997888997888(119903id 119904V)1003816100381610038161003816100381610038161003816997888997888997888997888997888997888(119903id 119904V)1003816100381610038161003816100381610038161003816

(9)

where subscript 119894 is the reference node id and 120572 is theadjustment parameter which is defined by (8) to control thelength to adjust 119904V

Each reference node of 119904V generates one shift vector asshown in Figure 7


Table 4 The result of victim selection

119904V 119899119887id 119889ToA(119899V 119899119887id) cvToA(119899V 119899119887id) 119889Euc(119899V 119899119887id) Reliable value Reference node

4

1 986 018 54 018 v2 1068 011 1166 011 v5 117 002 721 0016 832 031 825 008 v8 341 072 9 018 v

Table 5 DLCA table after DLCA


1 (5724 2768) 1 NA2 (5447 4040) 1 NA3 (5583 4711) 1 NA

4 (4728 2916) 067

1 986 018 10072 1068 011 13355 117 002 11456 832 031 3768 341 072 428

5 (5573 3688) 082

1 1076 01 9322 337 072 3743 896 025 10234 117 002 1145

6 (44 31) 0254 832 031 3767 528 056 7078 579 052 224

7 (39 36) 025 6 528 056 7078 982 018 806

8 (43 29) 0254 341 072 4286 579 052 2247 982 018 806

9 (41 45) 025 3 1136 005 1498

41

2

6

8

1

2

8

6

Figure 7 Shift vectors produced by four reference nodes of 119904V

The corrected location of 119904V which is denoted as 119904Vlowast isdecided by the resultant of the four shift vectors generated bythe four reference nodes as calculated in (10) and Figure 8

997888119904Vlowast = 997888119904V + sum997888V119894 (10)

2

14

46

8

２ＭＯＦＮＨＮ

slowast

s

Figure 8 Corrected location of 119904V

433 Recursively Correct Data Locations When there are nocandidates left the mean value of total cvloc(119899id) is computedWhen the variance of themean value of cvloc(119899id) is less than athreshold 120579cv the DLCA process is complete Table 5 presentsthe final result of DLCA


Table 6 Simulation settings

Settings ValuesNumber of nodes 500Anchor ratio 10Region 100m lowast 100m lowast 100mCommunication radius 119877 25mData observed period 119875119889 1 sDLCA discontinue threshold 120579cv 0001

5 Simulation Results

In this section we evaluate the performance of DLCA usingsimulations

51 Simulation Settings In our simulations unless otherwisespecified we use the settings shown in Table 6 Five hundrednodes comprising 50 anchor nodes and 450 sensor nodes arerandomly distributed in a 100m times 100m times 100m region witha communication radius of 25m 119875119889 is set to 1 s and 120579cv is setto 0001 All the results are the mean value of 100 simulationsThe simulation was run on a personal computer Intel core i7-3770 34GHz with 16GB RAM and 64 bit Windows 7 Thecomputing time is approximately 25ms for 500 nodes

Table 7 presents thememory needed tomaintain a DLCAtable in a timing group For each node we use a 4-byte integerto hold the node ID For location information three 4-bytefloats are used due to the three dimensions The confidencevalue is also held by a 4-byte float All nodes excluding theanchors need tomaintain a list of neighborsThe informationof a neighbor is a node ID which can be held by a pointerand the other three fields can each be held by a 4-byte floatTherefore the total memory usage is 20 bytes for a node and12 bytes for a neighborrsquos information Taking our simulationas an example it needs (500 nodes times 20 bytes) + (12 bytes timesaverage degree times 450 sensors) which is approximately 64Kbytes of memory

As for the nodemobility pattern twomodels are adoptedOne is the kinematic model in [37] The current field isassumed to be a superposition of a tidal and a residual currentfield The tidal field is assumed to be a spatially uniformoscillating current in one direction and the residual currentfield is assumed to be an infinite sequence of clockwiseand anticlockwise rotating eddiesThe dimensionless velocityfield in the kinematical model can be approximated by

119881119909 = 1198961120582V sin (1198962119909) cos (1198963119910) + 1198961120582 cos (21198961119905) + 1198964119881119910 = minus120582V cos (1198962119909) sin (1198963119910) + 1198965 (11)

where 119881119909 is the speed in the 119909-axis 119881119910 is the speed inthe 119910-axis and 1198961 1198962 1198963 120582 and V are variables whichare closely related to environmental factors such as tidesand bathymetry These parameters will change in differentenvironments

We use node density as the node degree which is theexpected number of nodes in a nodersquos neighborhood Nodedensity impact on localization coverage is defined as the ratio

0010203040506070809

1

Loca

lizat

ion

cove

rage

10 20 30 40 50 600Node density

Figure 9 Impact of node density on localization coverage

of localization nodes to the total nodes In Figure 9 thenode density is varied by changing the communication rangeof all the nodes from 5m to 35m Figure 9 shows thatfor localization methods based on trilateral positioning theminimum node density to achieve 08 localization coverageis 10

By simulating the node mobility based on the kinematicmodel we found that the node density decreases quickly astime increases especially when the speed of ocean current ishigh To keep the area of deployment the same and maintainhigh localization coverage (gt08) in this paper we simulatethe ocean current by assuming that all the parameters arerandom variables subject to a normal distribution with thesettings presented in Table 8

ToA distance measurements between nodes are assumedto follow (2) and (3) which are subject to the normaldistributions with 001 as mean values and 005 as the stan-dard deviations This is a reasonable assumption and can beeasily justified by existing underwater distance measurementtechnologies [11ndash13]

In addition to the kinematic model [37] the meanderingmodel [38] is also considered in this paper The nondimen-sional form of the meandering jet model is

120595 (119909 119910 119905)= minus tanh[ 119910 minus 119861 (119905) sin (119896 (119909 minus 119888119905))

radic1 + 11989621198612 (119905) cos2 (119896 (119909 minus 119888119905))]

119906 = minus120597120595120597119910

V = minus120597120595120597119909

= minus120597119910120595 (119909 119910 119905) = minus120597119909120595 (119909 119910 119905)

(12)

where 119861(119905) = 119860+ 120598 cos(120596119905)The parameter 119896 sets the numberof meanders in the unit length 119888 is the phase speed withwhich they shift downstream The time-dependent function119861 modulates the width of the meanders 119860 determines theaverage meander width 120598 is the amplitude of themodulation


Table 7 Memory usage of a DLCA table


Bytes 4 4 lowast 3 dimensions 4 Pointer 4 4 4

Table 8 Ocean current parameter settings

Parameters Mean std1198961 1198962 120587 011205871198963 2120587 021205871198964 1198965 1 01V 1 01120582 05sim30 005sim03

Table 9 Meandering model parameter settings

Parameters ValuesAverage meander width 119860 12Phase speed 119888 012Number of meanders in the unit length 119896 212058775Frequency 120596 04Amplitude of the modulation 120598 2 3 4 5

and 120596 is its frequency We use the meandering model tosimulate the ocean current based on the settings in [38]presented in Table 9

We consider two performance metrics localization errorand average communication cost Localization error is theaverage distance between the estimated positions and the realpositions of all nodes As in [25 27] for our simulations wenormalize this absolute localization error to the node com-munication range 119877 Average communication cost is definedas the overall messages exchanged in the network divided bythe number of localized nodes which is normalized to thesize of the beacon message (16 bytes in our simulations)

52 Results and Analysis

521 Performance with Varying Average Moving Speed Inthis simulation we compare our scheme to three schemesLSHL scheme [25 26] Euclidean scheme [39] and recursivescheme [40]119875119897 is set as 2119875119889 Figures 10 and 11 clearly show theeffect of node mobility on the localization performance Wecan see that DLCA successfully corrects data packet locationsand decreases the communication costThis is becauseDLCAis executed in the base station therefore the period ofsending localization message can be extended to decreasethe communication costs The localization error of all theschemes increases with the speed at which the node movesThis is mainly because the distance measurement errorincreases with the average moving speed Correspondinglythe final localization error will increase as well

522 Performance with Varying Times of 119875119889 In this set ofsimulations we set the length of 119875119897 to 11 multiples of 119875119889 and

DLCARecursive

EuclideanLSHL

2 25 315Average moving speed (ms)

0

01

02

03

04

05

Loca

lizat

ion

erro

r (R)

Figure 10 DLCA performance on localization error compared toother schemes

DLCA with n = 2Recursive

EuclideanLSHL


05

1015202530354045

Aver

age

com

mun

icat

ion

cost

Figure 11 DLCA performance on average communication costcompared to other schemes

change 119899 from 0 to 10 For kinematic model we considerdifferent marine environments by setting 120582 to be 05 1 2and 3 and the corresponding speeds of ocean current are254 395 644 and 886 (ms) respectively For meanderingmodel we set 120598 to be 2 3 4 5 and the corresponding speeds ofocean current are 086 161 205 and 325 (ms) respectively

It is shown in [25] that range-based localization schemesplace large requirements on the node density of networksFigure 9 also shows that the node density needs to be at least10 in order to localize 80 of the nodes

Figures 12 and 13 provide references for UWSN applica-tions to decide a suitable 119899 to fit their performance require-ment in different marine environments based on kinematicmodel and meandering model respectively For example themaximum 119899 is 10 to localize 99 nodes with less than 05119877


0

05

1

15

2Lo

caliz

atio

n er

ror (

R)

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 05

= 1

= 2

= 3

(a)

09

092

094

096

098

1

Loca

lizat

ion

cove

rage

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 05

= 1

= 2

= 3

(b)

Figure 12 DLCA performance on (a) localization error and (b) coverage with varying 120582 and 119899 (times of 119875119889) based on kinematic model

1 2 3 4 5 6 7 8 9 100n (times of Pd)

0

05

1

15

2

Loca

lizat

ion

erro

r (R)

= 2

= 3

= 4

= 5

(a)

1

096

097

098

099

Loca

lizat

ion

cove

rage

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 2

= 3

= 4

= 5

(b)

Figure 13 DLCA performance on (a) localization error and (b) coverage with varying 120576 and 119899 (times of 119875119889) based on meandering currentmobility model

localization error when 10 anchor nodes and less than 3msspeed are present in the network

523 Performance with Varying Anchor Percentage In thesimulation of Figure 14 120582 is set to 05 and 120576 is set to5 We can see that the more the anchors the lower thelocalization error For example if the anchor percentageis 5 the slope is almost 06 but if the anchor per-centage is 20 the slope can reach 045 This suggeststhat in sparse networks we can increase the number ofanchor nodes to achieve higher localization accuracy Itshould be noted that for all range-based localization meth-ods their localization accuracy increases with the anchorpercentage

6 Summary

In summary the main contributions of this paper are asfollows (1) we are the first to propose a hybrid localizationapproach which post facto corrects data locations at a basestation to improve the overall network communication costand sensor power With our hybrid localization approachthe period of sensor self-localization can be extended andthus decrease the computational overheads and high energyrequirements of sensors (2) DLCA is easily implementedand is cost-efficient in both computing time and memoryspace and (3) we analyze DLCA performance under dif-ferent marine environments by simulating ocean-currentspeeds based on kinematic models and also compare theresults to several range-based localization approaches The


Without DLCADLCA with anchor ratio = 5DLCA with anchor ratio = 10DLCA with anchor ratio = 20

0

01

02

03

04

05

06

07

08Lo

caliz

atio

n er

ror (

R)

1 2 3 4 5 6 7 8 9 100n (times of Pd)

(a)


1 2 3 4 5 6 7 8 9 100n (times of Pd)

0

05

1

15

2

Loca

lizat

ion

erro

r (R)

(b)

Figure 14 Performance on localization error with varying anchor ratio and 119899 (times of 119875119889) based on (a) kinematic model and (b) meanderingcurrent mobility model

results indicate satisfactory performance of our proposedscheme

Conflicts of Interest

The authors declare that they have no conflicts of interest

References

[1] H-P Tan R Diamant W K G Seah and M Waldmeyer ldquoAsurvey of techniques and challenges in underwater localizationrdquoOcean Engineering vol 38 no 14-15 pp 1663ndash1676 2011

[2] M Beniwal and R Singh ldquoLocalization techniques and theirchallenges in underwater wireless sensor networksrdquo Interna-tional Journal of Computer Science and Information Technolo-gies vol 5 pp 4706ndash4710 2014

[3] G Han J Jiang L Shu Y Xu and F Wang ldquoLocalizationalgorithms of underwater wireless sensor networks a surveyrdquoSensors vol 12 no 2 pp 2026ndash2061 2012

[4] M Erol-Kantarci H T Mouftah and S Oktug ldquoA surveyof architectures and localization techniques for underwateracoustic sensor networksrdquo IEEE Communications Surveys ampTutorials vol 13 no 3 pp 487ndash502 2011

[5] V Garg and M Jhamb ldquoA review of wireless sensor networkon localization techniquesrdquo International Journal of EngineeringTrends and Technology vol 4 pp 1049ndash1053 2013

[6] T J S Chowdhury C Elkin V Devabhaktuni D B Rawatand J Oluoch ldquoAdvances on localization techniques for wirelesssensor networks a surveyrdquo Computer Networks vol 110 pp284ndash305 2016

[7] G Han H Xu T Q Duong J Jiang and T Hara ldquoLocalizationalgorithms of wireless sensor networks a surveyrdquo Telecommu-nication Systems vol 52 pp 1ndash18 2011

[8] S Lee and K Kim ldquoLocalization with a mobile beacon inunderwater sensor networksrdquo in Proceedings of the IEEEIFIP

8th International Conference on Embedded and UbiquitousComputing EUC rsquo10 pp 316ndash319 December 2010

[9] H Luo Z Guo W Dong F Hong and Y Zhao ldquoLDBlocalization with directional beacons for sparse 3D underwateracoustic sensor networksrdquo Journal of Networks vol 5 no 1 pp28ndash38 2010

[10] H Luo Y Zhao Z Guo S Liu P Chen and L M Ni ldquoUDBusing directional beacons for localization in underwater sensornetworksrdquo in Proceedings of 14th IEEE International Conferenceon Parallel and Distributed Systems ICPADS rsquo08 pp 551ndash558December 2008

[11] S A Golden and S S Bateman ldquoSensor measurements for Wi-Fi location with emphasis on time-of-arrival rangingrdquo IEEETransactions on Mobile Computing vol 6 no 10 pp 1185ndash11982007

[12] H Xiong Z Chen W An and B Yang ldquoRobust TDOA local-ization algorithm for asynchronous wireless sensor networksrdquoInternational Journal of Distributed Sensor Networks vol 2015Article ID 598747 10 pages 2015

[13] K K Chintalapudi A Dhariwal R Govindan and G Sukhat-me ldquoAd-hoc localization using ranging and sectoringrdquo in Pro-ceedings of the IEEE INFOCOM 2004 - Conference on ComputerCommunications - Twenty-Third Annual Joint Conference of theIEEE Computer and Communications Societies pp 2662ndash2672March 2004

[14] J Liu Z Zhou Z Peng J-H Cui M Zuba and L Fion-della ldquoMobi-sync efficient time synchronization for mobileunderwater sensor networksrdquo IEEETransactions on Parallel andDistributed Systems vol 24 no 2 pp 406ndash416 2013

[15] J Liu Z Wang M Zuba Z Peng J-H Cui and S ZhouldquoDA-Sync a doppler-assisted time-synchronization scheme formobile underwater sensor networksrdquo IEEE Transactions onMobile Computing vol 13 no 3 pp 582ndash595 2014

[16] Z Li Z Guo F Hong and L Hong ldquoE2DTS an energy effi-ciency distributed time synchronization algorithm for under-water acoustic mobile sensor networksrdquo Ad Hoc Networks vol11 no 4 pp 1372ndash1380 2013


[17] J Liu Z Wang J-H Cui S Zhou and B Yang ldquoA joint timesynchronization and localization design for mobile underwatersensor networksrdquo IEEE Transactions on Mobile Computing vol15 no 3 pp 530ndash543 2016

[18] K Chen M Ma E Cheng F Yuan and W Su ldquoA survey onMACprotocols for underwater wireless sensor networksrdquo IEEECommunications Surveys Tutoials vol 16 pp 1433ndash1447 2014

[19] H Ramezani F Fazel M Stojanovic and G Leus ldquoCollisiontolerant and collision free packet scheduling for underwateracoustic localizationrdquo IEEE Transactions on Wireless Commu-nications vol 14 no 5 pp 2584ndash2595 2015

[20] M Erol L F M Vieira and M Gerla ldquoAUV-aided localizationfor underwater sensor networksrdquo in Proceedings of the Pro-ceeding of the 2nd Annual International Conference on WirelessAlgorithms Systems and Applications (WASA rsquo07) pp 44ndash54Chicago Ill USA August 2007

[21] MWaldmeyerH-P Tan andWKG Seah ldquoMulti-stageAUV-aided localization for underwater wireless sensor networksrdquo inProceedings of the IEEE International Conference on AdvancedInformation Networking and Applications Workshops (WAINArsquo11) pp 908ndash913 March 2011

[22] M Erol L F M Vieira and M Gerla ldquoLocalization withDiversquoNrsquoRise (DNR) beacons for underwater acoustic sensornetworksrdquo in Proceedings of the 2007 International Conferenceon Mobile Computing and Networking MobiCom rsquo07 - SecondWorkshop on Underwater Networks WUWNetrsquo07 pp 97ndash100September 2007

[23] M Erol L F M Vieira A Caruso F Paparella M Gerlaand S Oktug ldquoMulti stage underwater sensor localizationusing mobile beaconsrdquo in Proceedings of the 2nd InternationalConference on Sensor Technologies andApplications pp 710ndash714Cap Esterel France August 2008

[24] T Ojha and S Misra ldquoMobiL a 3-dimensional localizationscheme for Mobile Underwater Sensor Networksrdquo in Proceed-ings of 2013 National Conference on Communications NCC rsquo13February 2013

[25] Z Zhou J-H Cui and S Zhou ldquoEfficient localization for large-scale underwater sensor networksrdquoAdHoc Networks vol 8 no3 pp 267ndash279 2010

[26] G Han A Qian C Zhang Y Wang and J J P C RodriguesldquoLocalization algorithms in large-scale underwater acousticsensor networks a quantitative comparisonrdquo International Jour-nal of Distributed Sensor Networks vol 2014 Article ID 37938211 pages 2014

[27] Z Zhou Z Peng J-H Cui Z Shi and A Bagtzoglou ldquoScalablelocalization with mobility prediction for underwater sensornetworksrdquo IEEE Transactions on Mobile Computing vol 10 no3 pp 335ndash348 2011

[28] Y Zhang J Liang S Jiang andWChen ldquoA localizationmethodfor underwater wireless sensor networks based onmobility pre-diction and particle swarm optimization algorithmsrdquo Sensorsvol 16 no 2 2016

[29] M Liu X Guo and S Zhang ldquoLocalization based on bestspatial correlation distance mobility prediction for underwaterwireless sensor networksrdquo in Proceedings of the 34th ChineseControl Conference CCC rsquo15 pp 7827ndash7832 July 2015

[30] G Zhu R Jiang L Xie and Y Chen ldquoA distributed localizationscheme based on mobility prediction for underwater wirelesssensor networksrdquo in Proceedings of the 26th Chinese Control andDecision Conference CCDC rsquo14 pp 4863ndash4867 June 2014

[31] D Mirza and C Schurgers ldquoCollaborative localization for fleetsof underwater driftersrdquo in Proceedings of the OCEANS pp 1ndash6Vancouver Canada October 2007

[32] D Mirza and C Schurgers ldquoMotion-aware self-localizationfor underwater networksrdquo in Proceedings of the 3rd ACMInternational Workshop on Underwater Networks pp 51ndash582008

[33] D Mirza and C Schurgers ldquoEnergy-efficient ranging for post-facto self-localization in mobile underwater networksrdquo IEEEJournal on Selected Areas in Communications vol 26 no 9 pp1697ndash1707 2008

[34] C Bechaz and H Thomas ldquoGIB System the underwater GPSsolutionrdquo Proceedings of the 5th Europe Conference on Under-water Acoustics 2000

[35] T C Austin R P Stokey and K M Sharp ldquoPARADIGM Abuoy-based system for AUV navigation and trackingrdquo OceansConference Record (IEEE) vol 2 pp 935ndash938 2000

[36] J Callmer M Skoglund and F Gustafsson ldquoSilent localizationof underwater sensors usingmagnetometersrdquoEURASIP Journalon Advances in Signal Processing vol 2010 Article ID 709318 8pages 2010

[37] S P Beerens H Ridderinkhof and J T F Zimmerman ldquoAnanalytical study of chaotic stirring in tidal areasrdquoChaos Solitonsamp Fractals vol 4 no 6 pp 1011ndash1029 1994

[38] A Caruso F Paparella L F M Vieira M Erol and M GerlaldquoThe meandering current mobility model and its impact onunderwater mobile sensor networksrdquo in Proceedings of the27th IEEE Communications Society Conference on ComputerCommunications (INFOCOM rsquo08) pp 221ndash225 Phoenix ArizUSA April 2008

[39] D Niculescu and B Nath ldquoAd hoc positioning system (APS)using AOArdquo in Proceedings of the 22nd Annual Joint Conferenceon the IEEE Computer and Communications Societies (INFO-COM rsquo03) vol 3 pp 1734ndash1743 San Francisco Calif USA April2003

[40] J Albowicz A Chen and L Zhang ldquoRecursive position estima-tion in sensor networksrdquo in Proceedings of the 2001 InternationalConference on Network Protocols ICNP pp 35ndash41 November2001

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Active and Passive Electronic Components

Control Scienceand Engineering

Journal of


International Journal of

RotatingMachinery


Hindawi Publishing Corporation httpwwwhindawicom

Journal of

Volume 201

Submit your manuscripts athttpswwwhindawicom

VLSI Design



Shock and Vibration


Civil EngineeringAdvances in

Acoustics and VibrationAdvances in



Electrical and Computer Engineering

Journal of

Advances inOptoElectronics


Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of


Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014


Chemical EngineeringInternational Journal of Antennas and

Propagation




Navigation and Observation



DistributedSensor Networks



1T 2T 3T0

Pd

Heuristic Pl

DLCA Pl

(n minus 1)T nT



Figure 1 Comparison of period length

the sensor self-localized period (119875119897) equal to the data observedperiod (119875119889) The top line and middle line in Figure 1 denotethe data observed period (119875119889) and the sensor self-localizedperiod (119875119897) of the heuristic approach respectively Howeverthe downside of the heuristic approach is that when 119875119889 issmall it results in relatively high communication costs beinggenerated Due to the limited bandwidth of acoustic channelsused by underwater sensor networks this will be a largeburden on the network

Based on this motivation we aim to design a hybridlocalization approach for underwater sensor networks calledDLCA which takes full advantage of computing the data-location at a base station when data packets are receivedinstead of relying on sensors continuously self-localizing Inother words DLCA extends 119875119897 to multiples of 119875119889 as shownby the bottom line in Figure 1The whole localization processof DLCA is divided into two parts node localization andobserved data-location correction Node localization is run bythe nodes themselves and observed data-location correctionis run by the base station Simulation results show that DLCAcan greatly reduce localization communication costs whilemaintaining relative localization accuracy

This work makes two contributions First our workis unique in aiming at localization of location-aware datainstead of sensor localization in underwater sensor networksand provides a practical system with hybrid localizationThe second contribution is the design of post facto locationcorrection on the base station that improves localizationaccuracy without additional communication overheads andpower consumption on sensor nodes

The remainder of this paper is organized as follows InSection 2 we provide background information includingseveral distance acoustic measurement models time syn-chronization and communication schemes for UWSNs InSection 3 we review related works The design of DLCAincluding its network architecture data structure and funda-mental algorithms are presented in Section 4 the simulationresults are presented in Section 5 and Section 6 contains ourconclusions

2 Background

Underwater sensor networks comprise a large number ofsmall devices deployed in a physical underwater environ-ment Each node has special capabilities such as wirelesscommunicationwith its neighbors sensing data data storageand processing Today there are more than 50 localizationalgorithms in existence [5ndash7] These can be classified on thebasis of different aspects

Centralized Localization versus Distributed Localization Incentralized localization one central base station computes

the locations of unknown nodes while in distributed local-ization computation is done by the sensor itself and nodescommunicate with each other to obtain their position in thenetwork DLCA uses hybrid localization The locations ofsensors are computed by the sensors themselves to supportreal-time monitoring but the data locations are corrected bya base station to reduce nodes power consumption

Anchor Based versusAnchorlessThe anchor-based algorithmsuse anchor nodes (which know their own position fromprior GPS data) as reference nodes for localization Themore anchor nodes the higher localization accuracy butthe cost also increases as anchor nodes are equipped withextra resourcesThe anchorless schemesmeasure the distancebetween nodes for creating a local map of the nodes How-ever the distance measurement techniques to date have notbeen accurate so global coordinates are preferred over localcoordinates for most applications Therefore DLCA utilizesthe anchor-based localization

Range-Based versus Range-Free Range-based systems usetechniques such as ultrasound to measure the distancebetween nodes and then triangulation to compute the posi-tions of nodes Range-free techniques use implicit informa-tion provided by anchor nodes to obtain positions of nodessuch as number of hops between devices or radio coveragemembership [8ndash10] Although range-free protocols do notneed additional hardware for distance measurements theycan only provide rough positional estimates DLCA is moreinterested in accurate localization and thus DLCA adoptsrange-based schemes

Several distance measuring techniques such as receivedsignal strength indication (RSSI) time of arrival (ToA) andtime difference of arrival (TDoA) are widely used in UWSNsfor localization Among them ToA in particular is the mostcommonly used for UWSNs as it is more accurate [1ndash4] It isnot affected by channel fading but has an issue in achievingsynchronization between nodes It sends a single packet fromone node to another containing the time of its transmissionassuming perfect clock synchronization between nodes Thereceiving node knows when the packet arrived and if it issynchronizedwith the sender node the distance travelled canbe calculated using the following equation

119889ToA = 119888 lowast Δ119905ToA (1)

where 119888 is the speed of sound in seawater (119888 = 1500ms) andΔ119905ToA is the time of arrival (Δ119905ToA = receiver time ndash sendertime)

However sound waves can have transmission delays andvarying transmission ratesThe delaysmake the ToA distancelonger than the actual distance and the longer the realdistance the more the error deviation of the ToA distance[11ndash13] Equations (2) and (3) simulate the difference betweenactual distance and ToA distance

119889ToA = 119889Real + 120576 (2)

120576 = 119889Real times 119873(120583 1205902) (3)





3 Related Work













4 Design of DLCA



Sensor node

Base station

Anchor node

Surface buoy























ObserverID


timeReceiver

IDReceived

time


Anchor nodes

Sensor nodes

48

6

7

9

5

2

1

3



ObserverID


timeReceiver

IDReceivedtime

4 (52 29) 3 1 3006574 (52 29) 3 2 3007124 (52 29) 3 5 3007804 (52 29) 3 6 3005554 (52 29) 3 8 3002275 (56 35) 3 1 3007175 (56 35) 3 2 3002255 (56 35) 3 3 3005976 (44 31) 3 7 3003527 (39 36) 3 8 3006558 (43 29) 3 6 3003869 (41 45) 3 3 300757




Table 2 DLCA table





cvToA (119899id 119899119887id)


(4)




(5)








cvloc (119904V)= 1








1 (5724 2768) 1 NA2 (5447 4040) 1 NA3 (5583 4711) 1 NA

4 (52 29) 025

1 986 018 542 1068 011 11665 117 002 7216 832 031 8258 341 072 9

5 (56 35) 025

1 1076 01 7422 337 072 5613 896 025 12114 117 002 721

6 (44 31) 0254 832 031 8257 528 056 7078 579 052 224

7 (39 36) 025 6 528 056 7078 982 018 806

8 (43 29) 0254 341 072 96 579 052 2247 982 018 806

9 (41 45) 025 3 1136 005 1498

14

2

6

8Shi_vector

d４Ｉ (s r1 )

dＯ＝ (s r1 )





2

41

6

8dＯ＝(s r8)

d４Ｉ(s r8)

Shi_vector





(9)






4

1 986 018 54 018 v2 1068 011 1166 011 v5 117 002 721 0016 832 031 825 008 v8 341 072 9 018 v



1 (5724 2768) 1 NA2 (5447 4040) 1 NA3 (5583 4711) 1 NA

4 (4728 2916) 067

1 986 018 10072 1068 011 13355 117 002 11456 832 031 3768 341 072 428

5 (5573 3688) 082

1 1076 01 9322 337 072 3743 896 025 10234 117 002 1145

6 (44 31) 0254 832 031 3767 528 056 7078 579 052 224

7 (39 36) 025 6 528 056 7078 982 018 806

8 (43 29) 0254 341 072 4286 579 052 2247 982 018 806

9 (41 45) 025 3 1136 005 1498

41

2

6

8

1

2

8

6



997888119904Vlowast = 997888119904V + sum997888V119894 (10)

2

14

46

8


slowast

s














0010203040506070809

1

Loca

lizat

ion

cove

rage

10 20 30 40 50 600Node density







radic1 + 11989621198612 (119905) cos2 (119896 (119909 minus 119888119905))]

119906 = minus120597120595120597119910

V = minus120597120595120597119909

= minus120597119910120595 (119909 119910 119905) = minus120597119909120595 (119909 119910 119905)

(12)















DLCARecursive

EuclideanLSHL


0

01

02

03

04

05

Loca

lizat

ion

erro

r (R)



EuclideanLSHL


05

1015202530354045

Aver

age

com

mun

icat

ion

cost






0

05

1

15

2Lo

caliz

atio

n er

ror (

R)

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 05

= 1

= 2

= 3

(a)

09

092

094

096

098

1

Loca

lizat

ion

cove

rage

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 05

= 1

= 2

= 3

(b)


1 2 3 4 5 6 7 8 9 100n (times of Pd)

0

05

1

15

2

Loca

lizat

ion

erro

r (R)

= 2

= 3

= 4

= 5

(a)

1

096

097

098

099

Loca

lizat

ion

cove

rage

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 2

= 3

= 4

= 5

(b)




6 Summary




0

01

02

03

04

05

06

07

08Lo

caliz

atio

n er

ror (

R)

1 2 3 4 5 6 7 8 9 100n (times of Pd)

(a)


1 2 3 4 5 6 7 8 9 100n (times of Pd)

0

05

1

15

2

Loca

lizat

ion

erro

r (R)

(b)





References











































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation













3 Related Work













4 Design of DLCA



Sensor node

Base station

Anchor node

Surface buoy























ObserverID


timeReceiver

IDReceived

time


Anchor nodes

Sensor nodes

48

6

7

9

5

2

1

3



ObserverID


timeReceiver

IDReceivedtime

4 (52 29) 3 1 3006574 (52 29) 3 2 3007124 (52 29) 3 5 3007804 (52 29) 3 6 3005554 (52 29) 3 8 3002275 (56 35) 3 1 3007175 (56 35) 3 2 3002255 (56 35) 3 3 3005976 (44 31) 3 7 3003527 (39 36) 3 8 3006558 (43 29) 3 6 3003869 (41 45) 3 3 300757




Table 2 DLCA table





cvToA (119899id 119899119887id)


(4)




(5)








cvloc (119904V)= 1








1 (5724 2768) 1 NA2 (5447 4040) 1 NA3 (5583 4711) 1 NA

4 (52 29) 025

1 986 018 542 1068 011 11665 117 002 7216 832 031 8258 341 072 9

5 (56 35) 025

1 1076 01 7422 337 072 5613 896 025 12114 117 002 721

6 (44 31) 0254 832 031 8257 528 056 7078 579 052 224

7 (39 36) 025 6 528 056 7078 982 018 806

8 (43 29) 0254 341 072 96 579 052 2247 982 018 806

9 (41 45) 025 3 1136 005 1498

14

2

6

8Shi_vector

d４Ｉ (s r1 )

dＯ＝ (s r1 )





2

41

6

8dＯ＝(s r8)

d４Ｉ(s r8)

Shi_vector





(9)






4

1 986 018 54 018 v2 1068 011 1166 011 v5 117 002 721 0016 832 031 825 008 v8 341 072 9 018 v



1 (5724 2768) 1 NA2 (5447 4040) 1 NA3 (5583 4711) 1 NA

4 (4728 2916) 067

1 986 018 10072 1068 011 13355 117 002 11456 832 031 3768 341 072 428

5 (5573 3688) 082

1 1076 01 9322 337 072 3743 896 025 10234 117 002 1145

6 (44 31) 0254 832 031 3767 528 056 7078 579 052 224

7 (39 36) 025 6 528 056 7078 982 018 806

8 (43 29) 0254 341 072 4286 579 052 2247 982 018 806

9 (41 45) 025 3 1136 005 1498

41

2

6

8

1

2

8

6



997888119904Vlowast = 997888119904V + sum997888V119894 (10)

2

14

46

8


slowast

s














0010203040506070809

1

Loca

lizat

ion

cove

rage

10 20 30 40 50 600Node density







radic1 + 11989621198612 (119905) cos2 (119896 (119909 minus 119888119905))]

119906 = minus120597120595120597119910

V = minus120597120595120597119909

= minus120597119910120595 (119909 119910 119905) = minus120597119909120595 (119909 119910 119905)

(12)















DLCARecursive

EuclideanLSHL


0

01

02

03

04

05

Loca

lizat

ion

erro

r (R)



EuclideanLSHL


05

1015202530354045

Aver

age

com

mun

icat

ion

cost






0

05

1

15

2Lo

caliz

atio

n er

ror (

R)

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 05

= 1

= 2

= 3

(a)

09

092

094

096

098

1

Loca

lizat

ion

cove

rage

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 05

= 1

= 2

= 3

(b)


1 2 3 4 5 6 7 8 9 100n (times of Pd)

0

05

1

15

2

Loca

lizat

ion

erro

r (R)

= 2

= 3

= 4

= 5

(a)

1

096

097

098

099

Loca

lizat

ion

cove

rage

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 2

= 3

= 4

= 5

(b)




6 Summary




0

01

02

03

04

05

06

07

08Lo

caliz

atio

n er

ror (

R)

1 2 3 4 5 6 7 8 9 100n (times of Pd)

(a)


1 2 3 4 5 6 7 8 9 100n (times of Pd)

0

05

1

15

2

Loca

lizat

ion

erro

r (R)

(b)





References











































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation















4 Design of DLCA



Sensor node

Base station

Anchor node

Surface buoy























ObserverID


timeReceiver

IDReceived

time


Anchor nodes

Sensor nodes

48

6

7

9

5

2

1

3



ObserverID


timeReceiver

IDReceivedtime

4 (52 29) 3 1 3006574 (52 29) 3 2 3007124 (52 29) 3 5 3007804 (52 29) 3 6 3005554 (52 29) 3 8 3002275 (56 35) 3 1 3007175 (56 35) 3 2 3002255 (56 35) 3 3 3005976 (44 31) 3 7 3003527 (39 36) 3 8 3006558 (43 29) 3 6 3003869 (41 45) 3 3 300757




Table 2 DLCA table





cvToA (119899id 119899119887id)


(4)




(5)








cvloc (119904V)= 1








1 (5724 2768) 1 NA2 (5447 4040) 1 NA3 (5583 4711) 1 NA

4 (52 29) 025

1 986 018 542 1068 011 11665 117 002 7216 832 031 8258 341 072 9

5 (56 35) 025

1 1076 01 7422 337 072 5613 896 025 12114 117 002 721

6 (44 31) 0254 832 031 8257 528 056 7078 579 052 224

7 (39 36) 025 6 528 056 7078 982 018 806

8 (43 29) 0254 341 072 96 579 052 2247 982 018 806

9 (41 45) 025 3 1136 005 1498

14

2

6

8Shi_vector

d４Ｉ (s r1 )

dＯ＝ (s r1 )





2

41

6

8dＯ＝(s r8)

d４Ｉ(s r8)

Shi_vector





(9)






4

1 986 018 54 018 v2 1068 011 1166 011 v5 117 002 721 0016 832 031 825 008 v8 341 072 9 018 v



1 (5724 2768) 1 NA2 (5447 4040) 1 NA3 (5583 4711) 1 NA

4 (4728 2916) 067

1 986 018 10072 1068 011 13355 117 002 11456 832 031 3768 341 072 428

5 (5573 3688) 082

1 1076 01 9322 337 072 3743 896 025 10234 117 002 1145

6 (44 31) 0254 832 031 3767 528 056 7078 579 052 224

7 (39 36) 025 6 528 056 7078 982 018 806

8 (43 29) 0254 341 072 4286 579 052 2247 982 018 806

9 (41 45) 025 3 1136 005 1498

41

2

6

8

1

2

8

6



997888119904Vlowast = 997888119904V + sum997888V119894 (10)

2

14

46

8


slowast

s














0010203040506070809

1

Loca

lizat

ion

cove

rage

10 20 30 40 50 600Node density







radic1 + 11989621198612 (119905) cos2 (119896 (119909 minus 119888119905))]

119906 = minus120597120595120597119910

V = minus120597120595120597119909

= minus120597119910120595 (119909 119910 119905) = minus120597119909120595 (119909 119910 119905)

(12)















DLCARecursive

EuclideanLSHL


0

01

02

03

04

05

Loca

lizat

ion

erro

r (R)



EuclideanLSHL


05

1015202530354045

Aver

age

com

mun

icat

ion

cost






0

05

1

15

2Lo

caliz

atio

n er

ror (

R)

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 05

= 1

= 2

= 3

(a)

09

092

094

096

098

1

Loca

lizat

ion

cove

rage

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 05

= 1

= 2

= 3

(b)


1 2 3 4 5 6 7 8 9 100n (times of Pd)

0

05

1

15

2

Loca

lizat

ion

erro

r (R)

= 2

= 3

= 4

= 5

(a)

1

096

097

098

099

Loca

lizat

ion

cove

rage

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 2

= 3

= 4

= 5

(b)




6 Summary




0

01

02

03

04

05

06

07

08Lo

caliz

atio

n er

ror (

R)

1 2 3 4 5 6 7 8 9 100n (times of Pd)

(a)


1 2 3 4 5 6 7 8 9 100n (times of Pd)

0

05

1

15

2

Loca

lizat

ion

erro

r (R)

(b)





References











































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation

























ObserverID


timeReceiver

IDReceived

time


Anchor nodes

Sensor nodes

48

6

7

9

5

2

1

3



ObserverID


timeReceiver

IDReceivedtime

4 (52 29) 3 1 3006574 (52 29) 3 2 3007124 (52 29) 3 5 3007804 (52 29) 3 6 3005554 (52 29) 3 8 3002275 (56 35) 3 1 3007175 (56 35) 3 2 3002255 (56 35) 3 3 3005976 (44 31) 3 7 3003527 (39 36) 3 8 3006558 (43 29) 3 6 3003869 (41 45) 3 3 300757




Table 2 DLCA table





cvToA (119899id 119899119887id)


(4)




(5)








cvloc (119904V)= 1








1 (5724 2768) 1 NA2 (5447 4040) 1 NA3 (5583 4711) 1 NA

4 (52 29) 025

1 986 018 542 1068 011 11665 117 002 7216 832 031 8258 341 072 9

5 (56 35) 025

1 1076 01 7422 337 072 5613 896 025 12114 117 002 721

6 (44 31) 0254 832 031 8257 528 056 7078 579 052 224

7 (39 36) 025 6 528 056 7078 982 018 806

8 (43 29) 0254 341 072 96 579 052 2247 982 018 806

9 (41 45) 025 3 1136 005 1498

14

2

6

8Shi_vector

d４Ｉ (s r1 )

dＯ＝ (s r1 )





2

41

6

8dＯ＝(s r8)

d４Ｉ(s r8)

Shi_vector





(9)






4

1 986 018 54 018 v2 1068 011 1166 011 v5 117 002 721 0016 832 031 825 008 v8 341 072 9 018 v



1 (5724 2768) 1 NA2 (5447 4040) 1 NA3 (5583 4711) 1 NA

4 (4728 2916) 067

1 986 018 10072 1068 011 13355 117 002 11456 832 031 3768 341 072 428

5 (5573 3688) 082

1 1076 01 9322 337 072 3743 896 025 10234 117 002 1145

6 (44 31) 0254 832 031 3767 528 056 7078 579 052 224

7 (39 36) 025 6 528 056 7078 982 018 806

8 (43 29) 0254 341 072 4286 579 052 2247 982 018 806

9 (41 45) 025 3 1136 005 1498

41

2

6

8

1

2

8

6



997888119904Vlowast = 997888119904V + sum997888V119894 (10)

2

14

46

8


slowast

s














0010203040506070809

1

Loca

lizat

ion

cove

rage

10 20 30 40 50 600Node density







radic1 + 11989621198612 (119905) cos2 (119896 (119909 minus 119888119905))]

119906 = minus120597120595120597119910

V = minus120597120595120597119909

= minus120597119910120595 (119909 119910 119905) = minus120597119909120595 (119909 119910 119905)

(12)















DLCARecursive

EuclideanLSHL


0

01

02

03

04

05

Loca

lizat

ion

erro

r (R)



EuclideanLSHL


05

1015202530354045

Aver

age

com

mun

icat

ion

cost






0

05

1

15

2Lo

caliz

atio

n er

ror (

R)

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 05

= 1

= 2

= 3

(a)

09

092

094

096

098

1

Loca

lizat

ion

cove

rage

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 05

= 1

= 2

= 3

(b)


1 2 3 4 5 6 7 8 9 100n (times of Pd)

0

05

1

15

2

Loca

lizat

ion

erro

r (R)

= 2

= 3

= 4

= 5

(a)

1

096

097

098

099

Loca

lizat

ion

cove

rage

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 2

= 3

= 4

= 5

(b)




6 Summary




0

01

02

03

04

05

06

07

08Lo

caliz

atio

n er

ror (

R)

1 2 3 4 5 6 7 8 9 100n (times of Pd)

(a)


1 2 3 4 5 6 7 8 9 100n (times of Pd)

0

05

1

15

2

Loca

lizat

ion

erro

r (R)

(b)





References











































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










Table 2 DLCA table





cvToA (119899id 119899119887id)


(4)




(5)








cvloc (119904V)= 1








1 (5724 2768) 1 NA2 (5447 4040) 1 NA3 (5583 4711) 1 NA

4 (52 29) 025

1 986 018 542 1068 011 11665 117 002 7216 832 031 8258 341 072 9

5 (56 35) 025

1 1076 01 7422 337 072 5613 896 025 12114 117 002 721

6 (44 31) 0254 832 031 8257 528 056 7078 579 052 224

7 (39 36) 025 6 528 056 7078 982 018 806

8 (43 29) 0254 341 072 96 579 052 2247 982 018 806

9 (41 45) 025 3 1136 005 1498

14

2

6

8Shi_vector

d４Ｉ (s r1 )

dＯ＝ (s r1 )





2

41

6

8dＯ＝(s r8)

d４Ｉ(s r8)

Shi_vector





(9)






4

1 986 018 54 018 v2 1068 011 1166 011 v5 117 002 721 0016 832 031 825 008 v8 341 072 9 018 v



1 (5724 2768) 1 NA2 (5447 4040) 1 NA3 (5583 4711) 1 NA

4 (4728 2916) 067

1 986 018 10072 1068 011 13355 117 002 11456 832 031 3768 341 072 428

5 (5573 3688) 082

1 1076 01 9322 337 072 3743 896 025 10234 117 002 1145

6 (44 31) 0254 832 031 3767 528 056 7078 579 052 224

7 (39 36) 025 6 528 056 7078 982 018 806

8 (43 29) 0254 341 072 4286 579 052 2247 982 018 806

9 (41 45) 025 3 1136 005 1498

41

2

6

8

1

2

8

6



997888119904Vlowast = 997888119904V + sum997888V119894 (10)

2

14

46

8


slowast

s














0010203040506070809

1

Loca

lizat

ion

cove

rage

10 20 30 40 50 600Node density







radic1 + 11989621198612 (119905) cos2 (119896 (119909 minus 119888119905))]

119906 = minus120597120595120597119910

V = minus120597120595120597119909

= minus120597119910120595 (119909 119910 119905) = minus120597119909120595 (119909 119910 119905)

(12)















DLCARecursive

EuclideanLSHL


0

01

02

03

04

05

Loca

lizat

ion

erro

r (R)



EuclideanLSHL


05

1015202530354045

Aver

age

com

mun

icat

ion

cost






0

05

1

15

2Lo

caliz

atio

n er

ror (

R)

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 05

= 1

= 2

= 3

(a)

09

092

094

096

098

1

Loca

lizat

ion

cove

rage

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 05

= 1

= 2

= 3

(b)


1 2 3 4 5 6 7 8 9 100n (times of Pd)

0

05

1

15

2

Loca

lizat

ion

erro

r (R)

= 2

= 3

= 4

= 5

(a)

1

096

097

098

099

Loca

lizat

ion

cove

rage

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 2

= 3

= 4

= 5

(b)




6 Summary




0

01

02

03

04

05

06

07

08Lo

caliz

atio

n er

ror (

R)

1 2 3 4 5 6 7 8 9 100n (times of Pd)

(a)


1 2 3 4 5 6 7 8 9 100n (times of Pd)

0

05

1

15

2

Loca

lizat

ion

erro

r (R)

(b)





References











































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation












1 (5724 2768) 1 NA2 (5447 4040) 1 NA3 (5583 4711) 1 NA

4 (52 29) 025

1 986 018 542 1068 011 11665 117 002 7216 832 031 8258 341 072 9

5 (56 35) 025

1 1076 01 7422 337 072 5613 896 025 12114 117 002 721

6 (44 31) 0254 832 031 8257 528 056 7078 579 052 224

7 (39 36) 025 6 528 056 7078 982 018 806

8 (43 29) 0254 341 072 96 579 052 2247 982 018 806

9 (41 45) 025 3 1136 005 1498

14

2

6

8Shi_vector

d４Ｉ (s r1 )

dＯ＝ (s r1 )





2

41

6

8dＯ＝(s r8)

d４Ｉ(s r8)

Shi_vector





(9)






4

1 986 018 54 018 v2 1068 011 1166 011 v5 117 002 721 0016 832 031 825 008 v8 341 072 9 018 v



1 (5724 2768) 1 NA2 (5447 4040) 1 NA3 (5583 4711) 1 NA

4 (4728 2916) 067

1 986 018 10072 1068 011 13355 117 002 11456 832 031 3768 341 072 428

5 (5573 3688) 082

1 1076 01 9322 337 072 3743 896 025 10234 117 002 1145

6 (44 31) 0254 832 031 3767 528 056 7078 579 052 224

7 (39 36) 025 6 528 056 7078 982 018 806

8 (43 29) 0254 341 072 4286 579 052 2247 982 018 806

9 (41 45) 025 3 1136 005 1498

41

2

6

8

1

2

8

6



997888119904Vlowast = 997888119904V + sum997888V119894 (10)

2

14

46

8


slowast

s














0010203040506070809

1

Loca

lizat

ion

cove

rage

10 20 30 40 50 600Node density







radic1 + 11989621198612 (119905) cos2 (119896 (119909 minus 119888119905))]

119906 = minus120597120595120597119910

V = minus120597120595120597119909

= minus120597119910120595 (119909 119910 119905) = minus120597119909120595 (119909 119910 119905)

(12)















DLCARecursive

EuclideanLSHL


0

01

02

03

04

05

Loca

lizat

ion

erro

r (R)



EuclideanLSHL


05

1015202530354045

Aver

age

com

mun

icat

ion

cost






0

05

1

15

2Lo

caliz

atio

n er

ror (

R)

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 05

= 1

= 2

= 3

(a)

09

092

094

096

098

1

Loca

lizat

ion

cove

rage

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 05

= 1

= 2

= 3

(b)


1 2 3 4 5 6 7 8 9 100n (times of Pd)

0

05

1

15

2

Loca

lizat

ion

erro

r (R)

= 2

= 3

= 4

= 5

(a)

1

096

097

098

099

Loca

lizat

ion

cove

rage

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 2

= 3

= 4

= 5

(b)




6 Summary




0

01

02

03

04

05

06

07

08Lo

caliz

atio

n er

ror (

R)

1 2 3 4 5 6 7 8 9 100n (times of Pd)

(a)


1 2 3 4 5 6 7 8 9 100n (times of Pd)

0

05

1

15

2

Loca

lizat

ion

erro

r (R)

(b)





References











































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation












4

1 986 018 54 018 v2 1068 011 1166 011 v5 117 002 721 0016 832 031 825 008 v8 341 072 9 018 v



1 (5724 2768) 1 NA2 (5447 4040) 1 NA3 (5583 4711) 1 NA

4 (4728 2916) 067

1 986 018 10072 1068 011 13355 117 002 11456 832 031 3768 341 072 428

5 (5573 3688) 082

1 1076 01 9322 337 072 3743 896 025 10234 117 002 1145

6 (44 31) 0254 832 031 3767 528 056 7078 579 052 224

7 (39 36) 025 6 528 056 7078 982 018 806

8 (43 29) 0254 341 072 4286 579 052 2247 982 018 806

9 (41 45) 025 3 1136 005 1498

41

2

6

8

1

2

8

6



997888119904Vlowast = 997888119904V + sum997888V119894 (10)

2

14

46

8


slowast

s














0010203040506070809

1

Loca

lizat

ion

cove

rage

10 20 30 40 50 600Node density







radic1 + 11989621198612 (119905) cos2 (119896 (119909 minus 119888119905))]

119906 = minus120597120595120597119910

V = minus120597120595120597119909

= minus120597119910120595 (119909 119910 119905) = minus120597119909120595 (119909 119910 119905)

(12)















DLCARecursive

EuclideanLSHL


0

01

02

03

04

05

Loca

lizat

ion

erro

r (R)



EuclideanLSHL


05

1015202530354045

Aver

age

com

mun

icat

ion

cost






0

05

1

15

2Lo

caliz

atio

n er

ror (

R)

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 05

= 1

= 2

= 3

(a)

09

092

094

096

098

1

Loca

lizat

ion

cove

rage

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 05

= 1

= 2

= 3

(b)


1 2 3 4 5 6 7 8 9 100n (times of Pd)

0

05

1

15

2

Loca

lizat

ion

erro

r (R)

= 2

= 3

= 4

= 5

(a)

1

096

097

098

099

Loca

lizat

ion

cove

rage

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 2

= 3

= 4

= 5

(b)




6 Summary




0

01

02

03

04

05

06

07

08Lo

caliz

atio

n er

ror (

R)

1 2 3 4 5 6 7 8 9 100n (times of Pd)

(a)


1 2 3 4 5 6 7 8 9 100n (times of Pd)

0

05

1

15

2

Loca

lizat

ion

erro

r (R)

(b)





References











































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation




















0010203040506070809

1

Loca

lizat

ion

cove

rage

10 20 30 40 50 600Node density







radic1 + 11989621198612 (119905) cos2 (119896 (119909 minus 119888119905))]

119906 = minus120597120595120597119910

V = minus120597120595120597119909

= minus120597119910120595 (119909 119910 119905) = minus120597119909120595 (119909 119910 119905)

(12)















DLCARecursive

EuclideanLSHL


0

01

02

03

04

05

Loca

lizat

ion

erro

r (R)



EuclideanLSHL


05

1015202530354045

Aver

age

com

mun

icat

ion

cost






0

05

1

15

2Lo

caliz

atio

n er

ror (

R)

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 05

= 1

= 2

= 3

(a)

09

092

094

096

098

1

Loca

lizat

ion

cove

rage

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 05

= 1

= 2

= 3

(b)


1 2 3 4 5 6 7 8 9 100n (times of Pd)

0

05

1

15

2

Loca

lizat

ion

erro

r (R)

= 2

= 3

= 4

= 5

(a)

1

096

097

098

099

Loca

lizat

ion

cove

rage

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 2

= 3

= 4

= 5

(b)




6 Summary




0

01

02

03

04

05

06

07

08Lo

caliz

atio

n er

ror (

R)

1 2 3 4 5 6 7 8 9 100n (times of Pd)

(a)


1 2 3 4 5 6 7 8 9 100n (times of Pd)

0

05

1

15

2

Loca

lizat

ion

erro

r (R)

(b)





References











































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation






















DLCARecursive

EuclideanLSHL


0

01

02

03

04

05

Loca

lizat

ion

erro

r (R)



EuclideanLSHL


05

1015202530354045

Aver

age

com

mun

icat

ion

cost






0

05

1

15

2Lo

caliz

atio

n er

ror (

R)

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 05

= 1

= 2

= 3

(a)

09

092

094

096

098

1

Loca

lizat

ion

cove

rage

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 05

= 1

= 2

= 3

(b)


1 2 3 4 5 6 7 8 9 100n (times of Pd)

0

05

1

15

2

Loca

lizat

ion

erro

r (R)

= 2

= 3

= 4

= 5

(a)

1

096

097

098

099

Loca

lizat

ion

cove

rage

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 2

= 3

= 4

= 5

(b)




6 Summary




0

01

02

03

04

05

06

07

08Lo

caliz

atio

n er

ror (

R)

1 2 3 4 5 6 7 8 9 100n (times of Pd)

(a)


1 2 3 4 5 6 7 8 9 100n (times of Pd)

0

05

1

15

2

Loca

lizat

ion

erro

r (R)

(b)





References











































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










0

05

1

15

2Lo

caliz

atio

n er

ror (

R)

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 05

= 1

= 2

= 3

(a)

09

092

094

096

098

1

Loca

lizat

ion

cove

rage

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 05

= 1

= 2

= 3

(b)


1 2 3 4 5 6 7 8 9 100n (times of Pd)

0

05

1

15

2

Loca

lizat

ion

erro

r (R)

= 2

= 3

= 4

= 5

(a)

1

096

097

098

099

Loca

lizat

ion

cove

rage

1 2 3 4 5 6 7 8 9 100n (times of Pd)

= 2

= 3

= 4

= 5

(b)




6 Summary




0

01

02

03

04

05

06

07

08Lo

caliz

atio

n er

ror (

R)

1 2 3 4 5 6 7 8 9 100n (times of Pd)

(a)


1 2 3 4 5 6 7 8 9 100n (times of Pd)

0

05

1

15

2

Loca

lizat

ion

erro

r (R)

(b)





References











































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











0

01

02

03

04

05

06

07

08Lo

caliz

atio

n er

ror (

R)

1 2 3 4 5 6 7 8 9 100n (times of Pd)

(a)


1 2 3 4 5 6 7 8 9 100n (times of Pd)

0

05

1

15

2

Loca

lizat

ion

erro

r (R)

(b)





References











































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation


































RoboticsJournal of





Journal of



RotatingMachinery



Journal of

Volume 201


VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation









ResearchArticle Hybrid Localization Approach for ...

Documents

Transcript of ResearchArticle Hybrid Localization Approach for ...