A Hierarchical Localization Scheme for Large ScaleUnderwater Wireless Sensor Networks

Yi Zhou†, Kai Chen§, Jianhua He‡,Jianbo Chen†, Alei Liang�

†School of Electronic, Information and Electrical Engineering, Shanghai Jiaotong University, China.‡Institute of Advanced Telecommunications, Swansea University, UK.

§School of Information Security and Engineering, Shanghai Jiaotong University, China�School of Software, Shanghai Jiaotong University, China

∗Corresponding author: Kai Chen, kchen@sjtu.edu.cn

Abstract—In this paper, we study the localization problemin large-scale Underwater Wireless Sensor Networks (UWSNs).Unlike in the terrestrial positioning, the global positioning system(GPS) can not work efficiently underwater. The limited band-width, the severely impaired channel and the cost of underwaterequipment all makes the localization problem very challenging.Most current localization schemes are not well suitable for deepunderwater environment. We propose a hierarchical localizationscheme to address the challenging problems. The new schememainly consists of four types of nodes, which are surface buoys,Detachable Elevator Transceivers (DETs), anchor nodes andordinary nodes. Surface buoy is assumed to be equipped withGPS on the water surface. A DET is attached to a surface buoyand can rise and down to broadcast its position. The anchor nodescan compute their positions based on the position informationfrom the DETs and the measurements of distance to the DETs.The hierarchical localization scheme is scalable, and can be usedto make balances on the cost and localization accuracy. Initialsimulation results show the advantages of our proposed scheme.

I. INTRODUCTION

The powerful computing capabilities together with dis-tributed wireless networking techniques enable creation ofinnovative self-organized and peer-to-peer large scale wirelessnetworks. Wireless sensor networks (WSN) has appeared asone of the such networks in the last few years [1]. In aWSN, a large number of powerful sensors are networked bywireless communication technologies and send measurementdata to sink(s) for further processing. The networked sensorscoordinate to perform distributed sensing of environmentalphenomena over large scale of physical space and enable reli-able monitoring and control in various applications. Wirelesssensor networks provide bridges between the virtual worldof information technology and the real physical world. Theyrepresent a fundamental paradigm shift from traditional inter-human personal communications to autonomous inter-devicecommunications. They promise unprecedented new abilities toobserve and understand large-scale, real-world phenomena ata fine spatial-temporal resolution.In recent several years, there has been a rapidly growing

interest in Underwater Wireless Sensor Networks (UWSNs).UWSNs can be used for a broad range scientific explo-ration, including ocean sampling, environmental monitoring,

undersea Explorations, disaster prevention, assisted navigation,distributed Tactical Surveillance and mine reconnaissance [1].Although many network protocols have been developed re-cently for wireless sensor networks, the unique characteristicsof the underwater communication channel, make the designof efficient and reliable data communication protocols verychallenging. There are still many issues unsolved for the largescale UWSNs, such as reliable transport, routing, MAC andlocalization [2] [3] [4] [5] [6]. Major design challenges inthe design of underwater acoustic networks are due to thelimited bandwidth, and high and variable propagation delays,severely impaired underwater channel, limited battery power,high bit error rates and temporary losses of connectivity, andnode failure due to fouling and corrosion.In those general wireless sensor networks and applications,

as well as the source detection and tracking applications ofour interest, location estimation is a vital component. Withoutthe node location information, the data received in the sinknode can not be identified where it comes from, and becomesmeaningless to the applications such as source tracking. Inaddition, location information can be used to design effi-cient networking and management protocols, for example,geographic based sensor network routing, network and sensornodes coordinations for target tracking. It is expected thatlocation information will be important as well for UWSNs.Due to the importance of location information, numerouslocalization algorithms have been proposed for wireless sensornetworks [12] [13] [14] [15]. However, as mentioned above,the acoustic channel is severely impaired and the GPS signalcan’t propagate far through water, the existing positioningschemes for terrestrial WSNs can not be used directly forUWSNs.In this paper, we are motivated to propose a hierarchical

localization scheme for large scale UWSNs. The new schememainly consists of four types of nodes, which are surfacebuoys, DETs, anchor nodes and ordinary nodes. Surface buoyis assumed to be equipped with GPS on the water surface.A DET is attached to a surface buoy and can rise and downto broadcast its position. The anchor nodes can compute theirpositions based on the position information from the DETs andthe measurements of distance to the DETs. Through the hier-

2009 11th IEEE International Conference on High Performance Computing and Communications

978-0-7695-3738-2/09 $25.00 © 2009 IEEE

DOI 10.1109/HPCC.2009.103

470

archical design, we can achieve scalability, and make balanceson the cost and localization accuracy. Initial simulation resultsshow the advantages of our proposed scheme. In the rest of thepaper, we will introduce the related work in Section II. Theproposed localization scheme will be presented in Section III.Simulation setting and result analysis are given in Section IV.Finally Section V concludes the paper.

II. RELATED WORK

Localization has been widely studied for terrestrial wire-less sensor networks. The primary purpose of localizationalgorithm is to estimate the locations (absolute or relative)of sensor nodes in a sensor networks, with a fraction ofthe anchor nodes (sensor nodes with known locations, eitherabsolute or relative). With regard to the mechanisms used forlocation estimation, localization algorithms can be divided intotwo major categories, range-based and range-free. In the range-based location algorithms, distance or angle estimates withneighbors will be used for calculating node locations. Typi-cal range-based algorithms include the sum-distances basedalgorithm [14] [15]. In the range-free location algorithms,the neighbor distance/angle information is assumed to beunavailable for positioning due to the cost and hardwarelimitation. DV-hop algorithm is a typical range-free locationalgorithm [12] [13]. In this paper, we will use a range-basedalgorithm, where the range between nodes will be measuredwith acoustic transceivers.The location algorithms can also be implemented in ei-

ther centralized or distributed way to tradeoff on accuracy,communication and computation overhead. Here centralizedalgorithms mean that a central entity (e.g., the sink) exists,which collects all the necessary information (e.g., measureddistances between the sensors in communications ranges andestimated distances to the anchor nodes), and determines thelocations for the sensors (excluding the anchor nodes) ina centralized way. After the central entity determines thelocations, it sends the location information back to the sensors.On the other hand, in the distributed localization algorithm,the function of determining the locations is distributed to thesensors themselves, instead of implemented in one centralentity.As more information is gathered and utilized to determine

the locations in the centralized localization algorithms, central-ized localization algorithms can have higher location estima-tion accuracy compared to distributed localization algorithms.However, it is obvious, centralized localization algorithmwill have big communication and computation challenges. Inaddition, as the sensor nodes normally have no identification,it can be very difficult for the central entity to send the lo-cation information back to the sensors. Therefore, centralizedlocalization algorithms is normally limited for use in smallscale sensor networks or used for understanding the upperperformance bounds of distributed localization algorithms.The localization is important and challenging for UWSNs.

There are several schemes proposed for UWSNs [7] [8][11]. In [7], a hierarchical localization scheme for large scale

UWSNs was proposed. The system mainly consists of threetypes of nodes: surface buoys, anchor nodes and ordinarysensor nodes. The buoys are equipped with GPSs. The firstlocalization step is anchor node localization, for which it isassumed all the anchor nodes can estimate their positionsby contacting directly with surface buoys. The second stepis ordinary node localization through anchor node’s positioninformation. The key of the scheme is that anchor nodes arelocalized through buoys. However, it can be very difficult forbuoys to communicate directly with anchor nodes under deepwater. On the other hand, a very large number of anchor nodesare used in [7], which results in very high cost and the local-ization performance is not satisfying. In [11], an interestingidea of Dive and Rise (DNR) positioning is presented. MobileDNR beacons are used to replace static anchor nodes. EachDNR beacon is equipped with GPS. When DNR beacon moveto water surface, they acquire x-y coordinate through GPS, andmove down to broadcast their position to help localize ordinarysensor nodes. The major drawback of the DNR scheme isthe high expense of the DNR beacons. There are 25 DNRbeacons for only 1km x 1km x 1km underwater area, so 25GPS and 25 moving equipments will be needed, which isvery expensive. And the ordinary sensor nodes can only usethe position information of DNR beacons to calculate theirpositions, which will degrade the localization performances.

III. THE PROPOSED HIERARCHY LOCALIZATION SCHEME

A. Network Architecture

So far, there is not a very good solution for large scaleUWSN localization in deep water. We are motivated to proposea new DET based hierarchical localization scheme. The newscheme will inherit the merits of DNR scheme, such assimplicity and high localization ratio, but can significantlydecrease cost of the system, and increase scalability andlocalization performances thanks to the hierarchical design.The hierarchical network architecture is shown in Fig.1. It iscompose of four types of nodes: the surface buoys, the DETs,anchor nodes and ordinary sensor nodes. We assume thatall the underwater nodes are equipped with pressure sensors,which can provide the depth (z coordinate) information for thenodes. We also assume the network is static. But it is notedthat the proposed scheme can be extended easily for non-staticnetworks.Next we will describe the functions of the nodes in details.• Surface buoys: A number of Surface buoys are placed onthe water surface. The surface buoys are equipped withGPS to obtain exact positions. The buoy will keep fixedand can play some important roles, such as navigationguidance. The buoys will be expense, but can be reusedto provide position information in our proposed schemefor the DET attached to it, when the DET rises to thesurface. It is not necessary for the DET and the surfacebuoy to communicate by acoustic transceiver.

• DETs: A DET is mainly composed of an elevator andan acoustic transceiver. The elevator helps the DET

471

rise or dive in vertical underwater, and the transceivercommunicates with the anchor nodes at different depths.We use a small number of surface buoys [10] with GPSs.Each buoy is equiped with a DET, which will be used tolocalize static anchor nodes. We assume that the DETcan move in vertical underwater, for example, operate asthe DNR beacon [10]. The DET get x-y coordinate fromits buoy when it moves to water surface and links to itsbuoy, then moves down to broadcast position informationat some pre-configured depths. For example, if the sensornetwork monitoring task is at the depth from 3km to4km, the DETs can start broadcasting at 3km depth andcontinue broadcasting every 200m until it dives to 4km.

• Anchor nodes: The main task of anchor nodes is tohelp locate the ordinary sensor nodes. They will havemore energy and use acoustic transceiver to communicatewith the DETs. It will listen to broadcast messagesfrom DETs and has larger communications range, ifcompared to the ordinary sensor nodes. Once a anchornode receives position messages from more than 3 DETs,it can calculate its coordinates. A localization confidencethreshold will be used to check if the position estimationis satisfying. If the estimation is accepted, the anchornode will broadcast its position to help locate the ordinarysensor nodes sometime later.

• Ordinary sensor nodes: The ordinary sensor nodes willmainly be used for the sensing task. There can be alarge number of sensor nodes, if compared to the an-chor nodes. The ordinary sensor nodes are assumed tohave less battery energy, and their acoustic transceiverwill be switched off for the concern of saving energy.The ordinary sensor nodes with unknown positions willlisten to broadcast messages periodically from the anchornodes. If more than 3 broadcast messages are receivedfrom different anchor nodes, the ordinary sensor nodeswill start to calculate its own position.

It is expected that the proposed scheme will have goodscalability and localization performances, and can be usedin a wide range of network scenarios. We will investigatelater how the numbers of buoys, anchor nodes, and ordinarysensor nodes, and the communication ranges will affect thelocalization performances under different network settings.

B. Computation of Node PositionsOnce the anchor nodes or the ordinary sensor nodes received

more than 3 broadcast messages from different nodes, they canstart to calculate their positions. As we have assumed that thez-coordinate of the nodes can be obtained by pressure sensors,the computation of node position is a 2-dimension positioningproblem. Generally nodes can determine their positions basedon the distance estimates by lateration, as used in [12] [13][15]. A much simpler method, call Min-max, is used in [14].In both cases no additional communications will be need forthe determination of the node positions.

• Lateration: Lateration is a form of triangulation and isthe most common method for deriving a position. For

Surface buoy DET Anchor node

Up Down

Depth area

Up Down

Up Down

Water surface

Fig. 1. System architecture.

example, assume there are n anchors. For an ordinarysensor node with estimated distances (di) to the anchori and it position (xi, yi), i ∈ [1, n], the following systemof equations can be derived as shown in [15] for the nodeposition (x, y):

⎧⎪⎪⎪⎪⎪⎨⎪⎪⎪⎪⎪⎩

(x1 − x)2 + (y1 − y)2 = d21,

(x2 − x)2 + (y2 − y)2 = d22,

...(xn−1 − x)2 + (yn−1 − y)2 = d2

n−1,(xn − x)2 + (yn − y)2 = d2

n.

(1)

By subtracting the last equation from the first n − 1equations, the system of equations can be linearized andshown in the form Ax = b, where

A =

⎡⎢⎢⎢⎣

2(x1 − xn) 2(y1 − yn)2(x2 − xn) 2(y2 − yn)...2(xn−1 − xn) (yn−1 − yn)

⎤⎥⎥⎥⎦ , (2)

and

B =

⎡⎢⎢⎢⎣

x21 − x2

n + y21 − y2

n + d2n − d2

1

x22 − x2

n + y22 − y2

n + d2n − d2

2...x2

n−1 − x2n + y2

n−1 − y2n + d2

n − d2n−1

⎤⎥⎥⎥⎦ , (3)

The equations can be solved by a standard least-squaresapproach: x̂ = (AT A)−1AT b [15].

• Min-max: Lateration is quite expensive in the numberof required floating point for computation limited wire-less sensor networks. A much simpler method has beenproposed by Savvides et al. in [14]. The main ideais to construct a bounding box for each anchor usingthe anchor’s position and estimated distances, and thento determine the intersection of these boxes [15]. Forexample, let sensor node k be the one to determine itsposition. The bound box of an anchor i for the node

472

k is created by adding and subtracting the estimateddistance (dki) from the anchor position (xi, yi). We canget the coordinates of the lower left and upper rightcoordinates of the box for the anchor i and node k as(xi− dki, yi− dki) and (xi + dki, yi + dki), respectively.The intersection of the bounding boxes is a smaller box,with the lower left and upper right points computed bytaking the maximum of all coordinate minimums and theminimum of all maximums as (max(xi−dki), max(yi−dki)) and (min(xi + dki), min(yi + dki)), respectively.The final position is obtained by averaging both cornercoordinates.

Both of the above methods can be used in our proposedhierarchical localization scheme. But in our experiments, weonly tested the lateration based method.

IV. SIMULATION RESULTSA. Simulation settingsWe have implemented our proposed scheme in Matlab

to evaluate its localization performances. In our simulationexperiments, the anchor nodes and the ordinary sensor nodesare assumed to be static. The water currents and the driftsare not considered. The anchor nodes and the ordinary sensornodes are distributed randomly in 1km x 1km plane area andthe underwater depth is from 3km to 4km. Therefore the wholeinvestigated underwater area is is 1km x 1km x 1km. Thenumber of the underwater sensor nodes (including the anchornodes) vary from 100 to 500. The anchor nodes have thecommunication ranges from 250m to 400m. The percentageof anchor nodes varies from 15%, 20%, 25%, to 30% in oursimulation. The measured distance between nodes is assumedto follow normal distributions with real distances as meanvalues and standard deviations to be one percent of realdistances. The normal distributions distance error is used inmany solutions [7] [9] [10].We use only 3 buoys equipped with GPS, which are placed

in fixed positions on the water surface. Each buoy is attachedby a DET. The DET uses the acoustic transceiver to commu-nicate with anchor nodes, with a configured communicationrange of 1 km in order that the 3 DETs can locate all theanchors in the 1km x 1km plane area. After these DETsacquire their x-y positions from buoys on the surface, theymove in vertical and broadcast their positions information atevery configured depth interval less than 1km starting from thedepth of 3 km. All the nodes underwater have pressure sensorwhich provides depth (z coordinate), so the nodes can start toestimate their positions with three different position broadcastinformation. The anchor nodes and the ordinary sensor nodescalculate their x-y coordinates by lateration based method.

B. Simulation results and analysisIn this paper we considered two performance metrics in our

simulations: the localization ratio of located ordinary sensornodes to the total nodes and the needed number of anchornodes for good performance. The localization ratio is definedas the number sensor nodes that can be localized (with at

least 3 broadcast messages from different anchor nodes). Thelocalization ratio is considered at the different scenarios, withthe number of sensor nodes varying from 100 to 500, thepercent of anchor nodes varying from 15% to 30%, and thetransmission range of anchor nodes varying from 250m to400m. We also considerate the needed number of anchornodes for different localization ratio at different anchor nodetransmission range.We compare the localization ratio at different anchor node

transmission range and anchor node ratio. The localization ra-tio performance is shown in Fig.2-Fig.5 for the communicationranges of 250m, 300m, 350m, and 400m, respectively. It canbe observed that the communication range has a big impact onthe localization ratio. With a transmission range of 400m, it iseasy to achieve a more than 90% localization ratio. Note thatonly one hop broadcast message is used in the localizationin our simulation, it is expected that the localization ratiocan be significantly improved if multi-hop broadcast messagesare to be used, which is left for our future work. It is alsoobserved when the number of anchor nodes is less than 70,the localization ratio can be improved very quickly with theincrease of the transmission range.

100 150 200 250 300 350 400 450 5000

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Number of sensor nodes

Loca

lizat

ion

ratio

s

Anchors Transmission range 250 m

Anchor rate=0.15Anchor rate=0.20Anchor rate=0.25Anchor rate=0.30

Fig. 2. Localization ratio versus the number of sensor nodes

100 150 200 250 300 350 400 450 5000.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Number of sensor nodes

Loca

lizat

ion

ratio

s

Anchors Transmission range 300 m

Anchor rate=0.15Anchor rate=0.20Anchor rate=0.25Anchor rate=0.30

Fig. 3. Localization ratio versus the number of sensor nodes

The corresponding average localization error performancesare presented in Fig.6-Fig.9 for the communication ranges of

473

100 150 200 250 300 350 400 450 5000.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Number of sensor nodes

Loca

lizat

ion

ratio

s

Anchors Transmission range 350 m

Anchor rate=0.15Anchor rate=0.20Anchor rate=0.25Anchor rate=0.30

Fig. 4. Localization ratio versus the number of sensor nodes

100 150 200 250 300 350 400 450 5000.4

0.5

0.6

0.7

0.8

0.9

1

Number of sensor nodes

Loca

lizat

ion

ratio

s

Anchors Transmission range 400 m

Anchor rate=0.15Anchor rate=0.20Anchor rate=0.25Anchor rate=0.30

Fig. 5. Localization ratio versus the number of sensor nodes

250m, 300m, 350m, and 400m, respectively. It can be observedthat the average localization error are quite small and isacceptable. With increased communications range, the averagelocalization error decreases. The increased number of anchornodes will also help reduce the average localization error. Butif we compare to the localization ratio and localization errorperformances, we can find that the impact of communicationsrange and number of anchor nodes on localization error ismuch smaller that that on localization ratio. Therefore, multi-hop based localization approach should be used to solve thelocalization ratio problem, which will be our future work.To help on the network planning, we also studied the num-

ber of anchor nodes required to achieve a certain localizationratio. Typical results on the number of required anchor nodesversus the anchor node communication range are presentedin Fig.10. The expected localization ratio is configured to0.7, 0.8 and 0.9, respectively. The number of required anchornodes can achieve the expected localization ratio for all theinvestigated scenarios with 100 to 500 overall sensor nodes. Itis observed that with 350m and 400m communication ranges,it is easy to achieve good ratio with less than 60 anchor nodesin 1km x 1km x 1km underwater area. Hence we can use only3 buoys with DETs, 60 anchor nodes with 350m transmissionrange in 1km x 1km x 1km volume that can acquire the samelocalization ratio as the DNRs scheme [11]. Compared to 25

100 150 200 250 300 350 400 450 5003.5

4

4.5

5

5.5

6

6.5

7

Number of sensor nodes

Dis

tanc

e er

rors

(m)

Anchors Transmission range 250 m

Anchor rate=0.15Anchor rate=0.20Anchor rate=0.25Anchor rate=0.30

Fig. 6. Distance errors versus the number of sensor nodes

100 150 200 250 300 350 400 450 5003.5

4

4.5

5

5.5

6

6.5

Number of sensor nodes

Dis

tanc

e er

rors

(m)

Anchors Transmission range 300 m

Anchor rate=0.15Anchor rate=0.20Anchor rate=0.25Anchor rate=0.30

Fig. 7. Distance errors versus the number of sensor nodes

100 150 200 250 300 350 400 450 5003

3.5

4

4.5

5

5.5

6

6.5

Number of sensor nodes

Dis

tanc

e er

rors

(m)

Anchors Transmission range 350 m

Anchor rate=0.15Anchor rate=0.20Anchor rate=0.25Anchor rate=0.30

Fig. 8. Distance errors versus the number of sensor nodes

buoys with GPS and similar DETs, our proposed scheme hasmuch lower cost.

V. CONCLUSIONIn this paper we proposed a new DET based hierarchical

localization scheme. The new scheme inherits the merits ofDNR scheme, such as simplicity and high localization ratio,but can significantly decrease cost of the system, and increasescalability and localization performances thanks to the hierar-chical design. The new scheme mainly consists of four types

474

100 150 200 250 300 350 400 450 5003

3.5

4

4.5

5

5.5

6

6.5

Number of sensor nodes

Dis

tanc

e er

rors

(m)

Anchors Transmission range 400 m

Anchor rate=0.15Anchor rate=0.20Anchor rate=0.25Anchor rate=0.30

Fig. 9. Distance errors versus the number of sensor nodes

250 300 350 40020

40

60

80

100

120

140

Anchor nodes transmision range

Num

ber o

f anc

hor n

odes

localiztion rate=0.7localiztion rate=0.8localiztion rate=0.9

Fig. 10. How the number of anchor nodes and the transmission rangeinfluence localization ratio.

of nodes: surface buoys, Detachable Elevator Transceivers(DETs), anchor nodes and ordinary nodes. Surface buoy isassumed to be equipped with GPS on the water surface. ADET is attached to a surface buoy and can rise and downto broadcast its position. The anchor nodes can compute theirpositions based on the position information from the DETs andthe measurements of distance to the DETs. The hierarchicallocalization approach is scalable, and can be used to makebalances on the cost and localization accuracy. Simulation re-sults show that our proposed scheme has successfully achievedthe design goals and outperform the existing schemes. In ourfuture work, we will test the proposed scheme under morenetwork scenarios and present more results.

ACKNOWLEDGMENT

The work is supported by the National Grand Fun-damental Research 973 Program of China under GrantNo.2006CB303007 and No.2007CB316506.

REFERENCES

[1] I.F. Akyildiz, “State of the art in protocol research for underwateracoustic sensor networks,” Proceedings of the 1st ACM internationalworkshop on Underwater networks, 2006.

[2] J. Kong, J. Cui, et al, “Building underwater ad-hoc networks and sensornetworks for large scale real-time aquatic applications,” IEEE MILCOM,2005.

[3] I.Akyildiz, et al, “Underwater acoustic sensor networks: research chal-lenges,” Ad Hoc Networks Journal,

[4] V. Chandrasekhar,et al, “Localization in underwater sensor networks:survey and challenges,” WUWNet, pp.33-40, 2006.

[5] J. Heidemann, et al, “Research challenges and applications for underwa-ter sensor networking,” IEEE Wireless Communications and NetworkingConference , 2006.

[6] J. Partan, J. Kurose, et al, ”A survey of practical issues in underwaternetworks,” WUWNet,

[7] Z. Zhou, J.-H. Cui, and S. Zhou, “Localization for large-scale underwatersensor networks,” in UCONN CSE Technical Report:UbiNet-TR06-04,2004.

[8] V. Chandrasekhar and W. K. G. Seah, “Area localization scheme forunderwater sensor networks,” Proceedings of the IEEE OCEANS AsiaPacific Conference, 2006.

[9] L. Doherty, K. S. J. Pister, et al, “Convex position estimation in wirelesssensor networks,” Proceedings of IEEE Infocom,Mar. 2001.

[10] C. Bechaz and H. Thomas, “GIB system: The underwater GPS solution,”Proceedings of 5th Europe Conference on Underwater Acoustics, May2000.

[11] M. Erol, L. Vieira, et al, “Localization with Dive’N’Rise (DNR)beacons for underwater acoustic sensor networks,” Proceedings of ACMInternational Workshop on Underwater Networks , Sep 2007.

[12] D. Niculescu, B. Nath, “Ad-hoc positioning system,” in proc: IEEEGlobecom, 2001.

[13] C. Savarese, K. Langendoen, J. Rabaey, “Robust positioning algorithmsfor distributed ad-hoc wireless sensor networks,” in USENIX TechnicalAnnual Conference, CA, pp. 317-328, 2002.

[14] A. Savvides, H. Park, M. Srivastava, “The bits and flops of the N-hop multilateration primitive for node localization problems,” in proc:First ACM International Workshop on Wireless Sensor Networks andApplications (WSNA), Atlanta, GA, pp. 112-121, 2002.

[15] K. Langendoen, N. Reijers, “Distributed localization in wireless sensornetworks: a quantitative comparison,” Computer Networks, vol.43, pp.499-518, 2003.

475