An Improved Localization Scheme Based on DV-Hop for Large-Scale Wireless Sensor Networks

Transcription

1 Journal of Communcatons Vol., o., December 06 n Improved Localzaton Scheme Based on DV-Hop for Large-Scale Wreless Sensor etworks Janpng Zhu, Chunfeng Lv, and Zhengsu Tao SOU College of Engneerng Scence and Technology, Shangha Ocean Unversty, Shangha, 0306, Chna Department of Electronc, Informaton and Electrcal Engneerng, Shangha Jaotong Unversty, Shangha, 0040, Chna Emal: {p-zhu, cflv}@shou.edu.cn; zstao@stu.edu.cn bstract Wreless Sensor etworks (WSs) have revolutonzed the world of dstrbuted systems and enabled many new applcatons. nd, measurement data or nformaton echanges happened n WSs wthout locaton nformaton are meanngless. It s etremely urgent to establsh and mantan low cost and hgh effcent localzaton schemes for real-tme large-scale survellance systems. In ths work, an mproved DVhop (Dstance Vector-hop) based localzaton scheme IDV-hop (mproved DV-hop) embedded n WLS (weghted least square) method s proposed for the purpose of surroundng survellance, obect localzaton for early warnng, rescue operatons and restructurng plan et al. Two crtcal parameters, correcton coeffcent k c and weghted coeffcent w s,, are ntroduced nto IDV-hop scheme to mprove locaton performance. nd then, S- smulatons demonstrate that analyss results match well wth smulaton results. Besdes, performance comparsons of IDV-hop scheme wth other DV-hop based schemes are also proposed. nalyss and comparson results show that localzaton delay and accuracy of IDV-hop s mproved largely relatve to other schemes, especally for hgher node densty. Inde Terms Localzaton, range-free, mproved DV-hop scheme, WLS, WSs I. ITRODUCTIO WSs play more and more decsve roles n varous aspects such as envronmental and habtat montorng, precson agrculture, anmal trackng, dsaster rescue and almost touch upon all aspects of our lfe. Besdes energy effcency requrements n WSs that wthdraw energy from batteres, much other network nformaton such as sensor or event locaton, eecutng tme etc. needs to be obtaned to satsfy actual requrements of many real-tme applcatons. In many applcatons, measurement data or nformaton echanges happened n WSs wthout locaton nformaton are meanngless. For eample, locatons of echanged messages must be known n envronmental montorng applcatons for bush fre survellance, water qualty montorng or precson agrculture. Moreover, sensor postons can also help to facltate the network wth an overall pont of vew, such Manuscrpt receved June 9, 06; revsed December 0, 06. Ths work was supported by atonal ature Scence Foundaton of Chna under Grant o Correspondng author emal: cflv@shou.edu.cn. do:0.70/cm as routng or connectvty for WSs. Therefore, localzaton has become a fundamental element n WSs. Dependng on whether absolute range measurements are used or not, localzaton schemes can be roughly classfed nto two categores []: range based and rangefree based. Range-based algorthms measure the eact dstance or angle of pendng localzaton nodes adoptng technques ncludng TO (tme of arrval) [], TDO (tme dfference of arrval) [3], RSSI (receved sgnal strength ndcator) [4] or O (angle of arrval) [5]. TO technology s commonly used as a means of obtanng range nformaton va sgnal propagaton tme, and wth two types. One s one-way propagaton tme measurements and the other s roundtrp propagaton tme measurements. Smplcty, hgh precson are the outstandng features for sensor localzaton usng TO scheme. However, there are some nferortes for TO scheme, such as tme and hardware consumng for synchronzaton, and easly affected by noses etc. The most basc localzaton system to use TO technques s GPS (Global Postonng System). TDO scheme can obtan range nformaton va arrval tme dfferences of messages, whch can be proposed wthout strct tme synchronzng. Tme dfferences of localzaton messages may be tme dfferences of arrvng at two dfferent recevers for one message, and also be tme dfferences of arrvng at one recever for two dfferent types of messages, rado propagaton and ultrasound propagaton for eample. Lke TO technology, TDO also reles on etensve hardware that s epensve and energy consumng for low-power sensor networks, and needs more anchors. To augment and complement TDO and TO schemes, O technque has been proposed that allows nodes to estmate and map relatve angles between neghbors through nstallng the array antennas on sensor nodes. O scheme can get relatve satsfed localzaton precse, but plenty of energy consumptons for powerwastng rado communcatons enables O to be nfeasble n localzaton for WSs. Less etra hardware, no tme synchronzaton, RSSI scheme can translate receved sgnal strength nto dstance estmates adoptng ether theoretcal or emprcal models, and ths make t the most popular range-free localzaton schemes. But, adoptng RF (Rado Frequency) system, problems suchlke mult-path fadng, background nterference, and 06 Journal of Communcatons 057

2 Journal of Communcatons Vol., o., December 06 rregular sgnal propagaton characterstcs make range estmates naccurate. cknowledgng that the cost of hardware requred by range-based solutons may be napproprate n relaton to the requred locaton precson, researchers have sought alternate range-free solutons to localzaton problems n sensor networks. The typcal range-free algorthms [6] nclude Centrod [7], CPE (Conve Poston Estmaton) [8], PIT (ppromate Pont n Trangle) [9] and DVhop [0]. Centrod scheme was frstly proposed by Bulusu et al. n [], whch s a localzaton scheme based on network connectvty. nchors broadcast a beacon contanng ther postons and ID nformaton to neghbors. Unknown node can obtan ts coordnates by presentng the centrod of these anchors after receved beacons eceedng the preset threshold, or recevng tme eceedng some tme. Lots of anchors are requred n centrod scheme to provde locaton nformaton for obtanng accurate centrod pont, and hgher localzaton precson. Pont-to-pont communcaton connectons between nodes can be attrbuted to geometrcal constrant of locatons of nodes, whch s the key concept of CPE algorthm. Whole network s modeled as a conve set, and node localzaton can be come down to the optmzaton desgn. ode locatons are then obtaned by a global optmzaton soluton adoptng the sem-defnte programmng and l programmng. PIT s an areabased range-free localzaton scheme, whch employs a novel area-based approach to perform locaton estmaton by solatng the envronment area nto trangular regons between anchors, recevng beacons contaned locaton nformaton. By utlzng dfferent combnatons of anchors, node reduces the sze of the estmated area n whch the node resdes to obtan a good locaton estmate. Instead of broadcastng n sngle hop, anchors n DV-hop algorthm flood ther locatons throughout network mantanng a runnng hop-count at each node along floodng paths. Unknown nodes calculate ther postons adoptng trlateraton or other methods. mong range-free localzaton methods, DV-hop algorthm handles the case where a normal node has less than three neghbor anchors, and computaton complety s relatvely low whch saves lots of energy. Consderng these attractve advantages of DV-hop algorthm, we prefer to DV-hop based algorthm n our sensor networks. In our DV-hop based localzaton scheme, several means are proposed to mprove performance of DV-hop: Frstly, correcton coeffcent k c, evaluatng accuracy of dstances between unknown node and anchors through dstance dfferentatons among anchors n our unform sensor networks, s proposed to mprove precson between dstances of unknown node and anchors. Secondly, weghted coeffcent w S, s proposed, takng three nfluences nto account, whch are transmsson range, mnmum number of hops between unknown node and anchor and dstances among anchors. Thrdly, unknown node s localzed tself wthn -hop network for hgh node densty. odes, ncludng anchors, can transmt localzaton messages n -hop n order to avod plenty of collsons for adoptng CSM/C scheme. Consequently, delay s mproved largely. Lastly, dstance per hop for each anchor s partcpated n calculatng dstances between unknown nodes and anchors other than only dstance per hop for the est anchor. nd then, WLS method s adopted to mprove dstance calculaton accuracy. The rest of ths paper s structured as follows: Secton II gves a summary of related works and analyss premse of our model. n mproved DV-hop IDV-hop s proposed n Secton III after presentng nferors of the orgnal DV-hop algorthm, and several means are also presented to mprove performance of DV-hop. In Secton IV, valdatons of localzaton error, localzaton delay are presented usng S- smulator, and comparsons of our scheme wth other DV-hop based schemes are also proposed. Fnally, concludng remarks are presented n Secton V. II. RELTED WORKS DV-hop s an attractve and low energy consumpton localzaton scheme, whch s the most general range-free localzaton scheme based on connectvty nformaton between nodes. Many algorthms from DV-hop have been proposed these years. ot smlar to other range-free localzaton schemes, DV-hop algorthm handles the case where a node has less three neghbor anchors. Three steps, localzaton nformaton echange phase for obtanng hop counters, average hop dstance computaton phase for every anchor and estmated poston phase usng trlateraton or mamum lkelhood estmaton method, are frstly proposed by culescu et al. n [0]. n mproved DV-Hop algorthm s proposed n [] to reduce locaton error accumulated over multple hops by usng a dfferental error correcton scheme. The dfference or error between estmated dstance of two anchors and actual dstance of these anchors s calculated, and ths error can be generalzed to calculate estmated error of dstance between unknown node and ts est anchor. DV-Loc algorthm n [3], uses Vorono dagrams to lmt the scope of the floodng and the error of computed postons to mprove localzaton accuracy through mprovng accuracy of hop count. n dvance DV-hop localzaton scheme [4] adopts the hop-sze of anchors, from whch unknown node measures dstance between anchors or between unknown node and anchors. Inherent error n estmated dstance between anchor and unknown node s reduced n the thrd step of dvance DV-hop, and WLS method s used, n whch weght factor s set as the nverse of the mnmum number of hops between unknown node and anchor. nd, locatons are refned by usng etraneous nformaton obtaned by solvng mathematcal equatons. But, weght factor n [4] has busness wth mnmum number of hops between 06 Journal of Communcatons 058

3 Journal of Communcatons Vol., o., December 06 unknown nodes and anchors only, not takng the hop dstances among anchors and transmsson range nto account. nd these three factors are all taken nto account n our IDV-hop. threshold s ntroduced n [5], whch uses weghted average hop dstance of anchors wthn hops, not all anchors n networks of orgnal DV-hop scheme to calculate average hop dstance of unknown nodes, and locaton results are corrected of ths DV-hop based scheme. HDV-hop (Hybrd DV-hop) n [6] obtans hgh localzaton accuracy and mnmzes floodng and then reduces energy consumpton, n whch anchors are deployed only on the permeter of sensor network and not nsde t. Because of anchors are located on the permeter of network, unknown node can contact wth anchors through two or more hops, whch consumes lots of energy to establsh or mantan DV-hop calculaton chans. We can constran our locatons n - hop, no matter anchor or unknown node n IDV-hop. Two refned localzaton algorthms, hyperbolc-dvhop localzaton algorthm and mproved weghted centrod DV-hop localzaton scheme (IWC-DV-hop) are proposed n [7]. Instead of takng average hop-sze of the est anchor to unknown node, hyperbolc-dv-hop scheme chooses average hop-szes among all anchors as average hop-sze of unknown node. Quad DV-hop and other two DV-hop based schemes, DV-hop and DVhop, are proposed n [8] to mprove the localzaton accuracy. Quad DV-hop formulates localzaton problems as bounded least squares problems, to be solved by quadratc programmng. Checkout and Selected 3-nchor DV-hop schemes are proposed n [9]. Former one adusts poston of a normal node based on ts dstance to the est neghbor anchor, n whch a checkout step s added to change estmated poston to a new one, a relatve accuracy value for computng the dstance between unknown node and each anchor. The other one chooses the best 3 anchors based on connectvty parameters. Mostly, three anchors can suffcently localze the normal sensor, rather than nvolvng n all avalable anchors n network. But, how to choose approprate 3-anchor to mprove localzaton accuracy requres takng network topology nto account. nd, teratve selectng best 3-abchor group can consume lots of energy. Moreover, there are many measurement methods n DV-hop based localzaton schemes for WSs. In the thrd step of orgnal DV-hop scheme, calculaton of estmated locaton s acheved usng trlateraton method or mamum lkelhood estmaton method. Most of the localzaton scheme based on DV-hop algorthm adopt these two calculaton methods. But n [0], normal sensor computes hop-sze based on all hop-sze values t receves from anchors, nstead of ust takng the frst receved hop-sze value, that s, hop-sze value of the est anchor. So, postons of normal nodes can be calculated by usng weghted least square method. s related above, dvanced DV-hop n [4] also uses WLS algorthm to calculate postons of unknown nodes. nd also, hyperbolc locaton algorthm s used to obtan locatons of normal nodes related n [7]. In ths paper, an mproved DV-hop (IDV-hop) scheme s proposed for real-tme large-scale montorng and localzaton system. t frst, nferortes of orgnal DVhop scheme are denoted elaborately after brought brefly about orgnal DV-hop. Then, several means are proposed to mprove performance of DV-hop accordng to these nferortes. nd then, localzaton behavors such as error and delay are valdated adoptng S- smulatons, takng parameters descrbng about network nto account, such as rato of anchors ( p ), node densty ( ) and transmsson range ( R ) et al. Moreover, performance comparsons between IDV-hop algorthm and other DVhop based schemes, such as DV-hop [4], Selectve-3- anchor [9], HDV-hop [6], are proposed n ths paper. III. PROPOSED SCHEME IDV-HOP. Orgnal DV-Hop Scheme Frst, we brefly eplan the orgnal DV-hop scheme as well as some pendng mprovements n ths scheme. Fg.. The orgnal DV-hop scheme Step Frst, each anchor broadcasts throughout the network a message contanng ts poston and hop count feld hop set to 0. Hop count value hop ncreases wth hop ncreasng durng the message broadcastng, whch means hop count value n the message wll be ncremented as soon as a node receves ths message. Every node (ether anchor or normal node) records the poston of and ntalzes the value of hop as hop count value n the message. nd, hop s the mnmum hop count between and. If the same message s receved agan, node mantans hop, and f the receved message contans a lower hop count value than t, wll update t wth that lower value and relay the message. Otherwse, wll gnore the message. Through ths mechansm, each node can obtan t s mnmum hop count hop between each anchor and t separately. Step Second, when an anchor receves postons of other anchors as well as the mnmum hop counts to other anchors, anchor can calculate ts average dstance per 06 Journal of Communcatons 059

4 Journal of Communcatons Vol., o., December 06 hop, whch s denoted as shown n followng (). dph = M, dph. Calculaton of ( ) (y y ) M hop,, dph s In (), M s the number of anchors n the network, and node refers to other anchor dfferent from anchor. hop s the mnmum dstance between anchor and measured by hops. (, y) and (, y ) refers to coordnates of anchor and, respectvely. Once dph s calculated, t wll be broadcasted by. nd then, all anchors can obtan all other anchors dph, but unknown nodes can only mantan dph broadcasted by ther est anchor (ether -hop or hgher hop neghbor anchor). When recevng hop sze of, unknown node can obtan dstances away each anchor (ncludng ), whch multples hop, (ts dstance to by hop count) by dph. Ths dstance s denoted as d,. Then, M dstances are obtaned by node, whch refers to d,, d, tll d M,.,, ote that all multple factor dph () d dph hop () s the same value n (), whch s average dstance one hop for from. For eample, anchor s 40 meters and hops away from, 00 meters and 6 hops away from 3, and 40 meters and 3 hops away from 4 n Fg.. can calculate dph usng (). Hence, dph s equal to ( ) / ( 6+3) 6.36m. In smlar way, values of dph, dph3, dph4 as 7.5 m,6.88 m,7.73m, respectvely. Then, each anchor broadcasts t s dph, so other nodes ncludng anchors receve t. wll mantan dph, the est neghbor anchor of, and calculate dstances d,, d,, d 3, and d 4, as m, m, m and m, respectvely. Step3 Thrd, when recevng dstances between and, can calculate ts estmated poston by trlateraton or other arthmetc methods as follows. (, y) n (3) and (4) s the coordnate of anchor. In whch, ( ) ( y y) d ( ) ( y y) d M M M ( T T X ) B (3) (4) ( M) ( y ym) ( M M ) ( ym ym ) M y ym d M d B y y d d M M M M M M X y B. Motvatons for Improved lgorthm DV-hop algorthm can obtan relatve satsfed localzaton errors wth less complety and less addtonal hardware, to derve locatons usng less than 3 neghbor anchors. But, there are some defects for orgnal DV-hop scheme, and localzaton performance do not come up to our epectatons. Frst, hop can be ncremented wth message spreadng f t s receved hop count s less than former mantaned one n step. Hop count s ncremented so long as the message s broadcasted once, no matter f net node s wthn transmsson range or not, no matter f node densty s hgh or not. Ths leads to over-estmate hop count value f node densty s relatvely hgh, and subsequently average hop dstance s underestmated. For eample, hop count between 3 and s 6, and geometrcal dstance between them s 00 meters, whle hop count between 3 and 4 s 5, and geometrcal dstance between them s 95 meters. If hop count s ncreased wth broadcastng, real dstance between and s underestmated. Secondly, t cannot decde to select whch anchor to calculate dph f there are several anchors wth the same hop away from unknown node. For eample, node has two -hop neghbor anchors and 4 n Fg., and t cannot decde whch one t mantans dph or dph 4. nd, mantans only dph to calculate dstances between anchors and t, whch s not so proper. Thrdly, network connectvty also plays an mportant role n DV-hop localzaton scheme. Dstance between two nodes (ether anchor or normal node) s represented by hops, rather than by geometrcal dstance. etwork connectvty eerts a tremendous nfluence on hop dstance, whch means hop dstance can brng about greater naccuracy f network connectvty s relatvely low. For eample, hop dstance between anchor and anchor s 3 hops through broadcastng n Fg., but the geometrcal dstance s only 40 meters n fact, almost one hop away from each other. Fnally, unknown node can obtan ts locaton through trlateraton, whch need three known anchors at least, no matter where anchors located. Hops can be enlarged f anchor densty s relatvely low, whch brngs about dph beng lower than actual one. nd, unknown node 06 Journal of Communcatons 060

5 Journal of Communcatons Vol., o., December 06 tres to seek farther anchors through two or more hop relay transmsson, whch can brng about a plenty of unnecessary energy consumpton. C. IDV-Hop Scheme Several means are proposed to mprove nferors above related n Secton I. Frstly, correcton coeffcent kc s proposed to evaluate accuracy of dstances between unknown node and anchors through the dstance dfferentatons among anchors n our unform sensor networks. Secondly, weghted coeffcent w S,, a comprehensve and precse llustratng parameter, s proposed to present the nfluence of transmsson range, mnmum number of hops between unknown node and anchor, dstances among anchors, other than presentng only one parameter of dstances between unknown node and anchors more. Thrdly, unknown node s localzed tself n -hop network for hgh node densty. odes, ncludng anchors, can transmt localzaton messages n -hop n order to avod plenty of collsons for adoptng CSM/C scheme. Consequently, delay s mproved largely. Lastly, dstance per hop for each node s partcpated n calculatng dstances between unknown node and anchors. nd then, WLS method s adopted to mprove dstance calculaton accuracy. ow, IDV-hop s presented n followng. Step Localzaton request Frst, hop s ntalzed to 0. fter network ntalzaton completed, each node can be aware whch hop t s belong to. lso, all nodes n the network can establsh ther one-hop neghbor lst and update ths lst at every localzaton slot. s localzaton request phase starts up, unknown node n ts -hop crcle can confrm ts -hop anchor neghbors (anchor nodes n -hop) ust n -hop crcle. nd then, sends a RL (Request Localzaton) frame to anchors, and RL frame contans all -hop neghbors. If the number of s greater than or equal to 3, node can drectly go to eecute DV-hop scheme as ntal DV-hop scheme. If the number of s less than 3, t need spread ts RL frame to -hop neghbors. nd also, -hop neghbors ncludng anchor neghbors (anchor nodes n -hop) are added to ths frame. If the sum of and s greater than or equal to 3, localzaton scheme can be promoted. Otherwse, ths localzaton can be falure. Step Hop nformaton echange Second, localzaton request phase fnshed, anchor broadcasts a HIE (Hop nformaton echange) frame contanng locaton of, ID of and also hop. There are three cases for obtanng hop. The frst one s that the number of anchor nodes n the ntersecton area for -hop neghbors of and -hop neghbors of s greater than or equal to 3, each hop s set, and sum hop ( ) s set to the number of other anchor nodes (case ). If anchor number n ths ntersecton area s less than 3, hop s set to f other anchors n -hop of, and set to f other anchors n -hop of. Sum s ( (case ). Of course, f anchor number n -hop of s less than 3, IDV-hop can epand to anchors n -hop of, and ths case s the same as second one (case 3). where case hop ( ) case,3 (5) R 4p 4 R d d / R p R 4p 4R d d / R p( R d (3 Rd )/d R 4 R d) d (3 Rd )/d and refers to the number of neghbor anchors n -hop crcle and the number of neghbor anchors n -hop rng, respectvely. Step 3 Dstance calculatons Dstance per hop dph can be obtaned as follows. Once dph s calculated, t wll be broadcasted by then, all anchors can obtan dph as (6). dph = or ( ) (y y ) hop, and Unknown node mantans each dph broadcasted by each anchor, not smlar to n orgnal DV-hop, dph calculates dstance away from each anchor dph, n (7). In whch, dph and dph refers to average dstance per hop of and, respectvely. dph, dph case * dph case,3 Step 4 Locaton calculatons Correcton coeffcent kc s ntroduced n IDV-hop to mprove accuracy of dstance dph, n (8).,, est true c, (6) (7) dph dph k (8), In (8), actual dstance of each anchor s calculated usng the geometrc method nvolvng actual coordnates, estmaton dstance dph smlar way, dstance dph ( ) ( y y ) and, true, est s calculated through Eq.6. In, of calculated usng (9). ctual dstance and anchors s dph and, true 06 Journal of Communcatons 06

6 Journal of Communcatons Vol., o., December 06, estmaton dstance dphest s calculated as the same way,,, for the dfference between and dphest s consdered as the smlar to the dfference between,, and dphest for our unform networks. S, S, dphest kc (9) Coordnate of s denoted as (, y), locaton of s then calculated by usng followng (0-), n whch, s presented as a smple form d. nd n s the number of anchors,.e. the smple form for. ( ) ( y y ) d ( ) ( y y ) d ( n ) ( y yn ) d n WLS algorthm s adopted to solve the coordnates of IDV-hop scheme, to mprove locaton accuracy. In WLS method, unknown parameters n (4) s presented as: Z (Q 'W 'WQ) Q 'W 'WH (4) In whch, W s weghted matr whch s presented the nfluence of dstance between anchors and unknown node, transmsson range and dstances between an anchor and other anchors. w, w, 0 W 0 0 w, n (5) Weght w, n (5) s taken as the nverse of hop, (0) between each anchor and n [4, 0]. ot only as [4], [0], weght w, should also take R and h nto account n ths work, whch s demonstrated as (6). ( ) ( y y ) ( n ) ( y yn ) d d n () ( n ) ( y yn ) ( n ) ( y yn ) d n d n ( ) ( y y ) ( n ) ( y yn ) d d n Squarng both sdes and smplfyng (), we can obtan () as the same as [4]. d dn ( y y ) ( n ) y ( y yn ) ( y ) d d n ( y n yn ) () ( n n ) y ( yn yn ) ( y ) d n d n ( n yn n yn ) ws, hopk ( ) hops, k R (6) In (6), s presented as (5). Weght values nfluenced by these three parameters play mportant roles on the localzaton accuracy, whch wll be analyzed through smulatons n Secton 4 later. ( n ) y ( y yn ) ( y ) n ( n ) y ( y yn ) ( n ) y ( y yn ) Q ( n n ) y( yn yn ) d d n ( y n yn ) d d n ( y n yn ) H d n d n ( n yn n yn ) 06 Journal of Communcatons Etensve smulatons are presented n ths secton to valdate accuracy of evaluated metrcs for localzaton error and delay usng S- smulator, based on analyses of dfferent parameters such as anchors rato, node densty etc. S- s a popular dscrete-event smulator whch was orgnally desgned for wred networks and has been subsequently etended to support wreless smulatons. nd also, performance comparsons between IDV-hop scheme and other DV-hop based schemes, such as DV-hop [4], Selectve-3-anchor [9], HDV-hop [6], are proposed to valdate some accuracy and delay superorty of ths tme-crtcal scheme IDV-hop. s shown n Fg., hgh anchor rato means that anchors n -hop of are enough for localzng unknown node. nd also, wth anchor rato p ncreasng, weght w, of each anchor decreases, each dstance Z y k QZ H IV. PERFORMCE VLIDTIOS n (3) between anchor and unknown node can only play an nsgnfcant role on localzaton accuracy. Moreover, wth anchor number ncreasng, accuracy of kc ncreases, consequently locaton for unknown node has more accuracy shown n Eq. 8 and Eq. 6. Transmsson range R ( R denoted n all Fgs.) plays an mportant role on localzaton error. Wth R ncreasng, error decreases. nd R ncreases to a hgher value such as R 40m, error ncreases as shown Fg. 3 for nodes ncludng anchors 06

7 Journal of Communcatons Vol., o., December 06 need more tme to dentfy ther -hop or -hop neghbors. Seen from Fg. and Fg. 3, analyss results are consstent wth smulaton results, whch s abbrevates as ana and sm n all fgures, respectvely. Transmsson range R also plays an mportant role on localzaton delay shown n Fg. 5. When R ncreases to R 40m, delay ncreases sharply ana,lambda=0. sm,lambda=0. ana,lambda=0.4 sm,lambda=0.4 ana,lambda=0.6 sm,lambda=0.6 ana,lambda=0.8 sm,lambda= Delay(ms) verage Localzaton Error (meters) 9 ana,r=5 sm,r=5 ana,r=0 sm,r=0 ana,r=0 sm,r=0 ana,r=40 sm,r= rato of anchors(%) Lambda Fg. 5. Localzaton delay based on Fg.. Localzaton error IDV-hop shows better localzaton behavors wth the varety of node densty shown n Fgs.6-7, comparng to DV-hop scheme [4], HDV-hop scheme [6] and Selectve-3-anchor scheme [9]. s related above, DVhop scheme reduces localzaton errors usng WLS method and other mproved methods, takng the weght factor as the nverse of the mnmum number of hop count between unknown node and each anchor. In fact, weght factor w, has also busness wth the number of 8 ana,r=5 sm,r=5 ana,r=0 sm,r=0 ana,r=0 sm,r=0 ana,r=40 sm,r=40 verage Localzaton Error (meters) anchors, transmsson range and hop counts among anchors. These three factors are taken nto account n our IDV-hop scheme, whch brng out hgher accuracy Lambda Fg. 3. Localzaton error based on rato of anchor=0% 8.. ana,lambda=0.0 sm,lambda=0.0 ana,lambda=0.04 sm,lambda=0.04 ana,lambda=0.06 sm,lambda=0.06 ana,lambda=0.08 sm,lambda= Delay(ms) verage Localzaton Error (meters) ana,idv-hop sm,idv-hop ana,dv-hop sm,dv-hop ana,selectve-3-anchor sm,selectve-3-anchor ana,hdv-hop sm,hdv-hop Lambda Fg. 6 Localzaton error comparson based on rato of anchors(%) Fg. 4. Localzaton delay. If node densty (Lambda denoted n all Fgs.) ncreases, more nodes ncludng anchors wll access the channel smultaneously, more collsons are brought out for adoptng CSM/C transmsson scheme, leadng to retransmssons ncreasng. Consequently, localzaton delay behavor wll become nferor shown n Fg Journal of Communcatons Three estmated dstance values away from three dfferent anchors are suffcent for unknown node to calculate ts locaton related n selectve 3-anchor DVhop (selectve-3-anchor). Based on the frst two steps of orgnal DV-hop, unknown node can obtan a group of canddates to calculate ts locaton. Then, the best 3anchor group s chosen to establsh estmated poston usng teratve method, whch consumes lots of energy for the computatonal complety, and t s unsutable for 063

8 Journal of Communcatons Vol., o., December 06 wreless sensor networks and s also ncomparable to IDV-hop scheme. HDV-hop s sutable for localzng events n hostle envronments, whch anchors are deployed on the permeters of the networks rather than scatterng them nsde the hostle terran. Consequently, unknown node whch localzes tself wll transmt the localzed messages, ncludng hop count, hop sze and the dstance of ths unknown node and each anchor, traversng two or more hops. Ths consumes plenty of energy whch s ntractable to normal montorng sensor networks. nchors are located randomly n crcle plane, accordng to a twodmensonal Posson dstrbuton wth a densty of n IDV-hop. ode s localzed wthn -hop or -hop, whch can present hgher localzaton accuracy and save much transmsson energy. odes ncludng anchors ncreasng, unknown node can eecute IDV-hop n -hop sensor networks, whch ncreases accuracy of hop count hop sze and dstance of unknown node and each anchor shown n Fg. 6. nd also, localzaton delay comparsons show that localzaton delay presents pror behavors, especally for hgher node densty shown n Fg. 7. Wth nodes ncreasng, more anchors are located n the transmsson range of unknown node n IDV-hop. nd unknown node can request to be localzed n -hop, localzng tself n -hop, whch can save lots of tme. Dstance dfferences, among anchors or between anchors and unknown node, can be reduced to small values. The value of anchor rato s assgned to 0% for the performance comparsons n ths paper, takng the localzaton characterstcs of all four schemes nto account. rato of anchor=0% Delay(ms) ana,idv-hop sm,idv-hop ana,dv-hop sm,dv-hop ana,selectve-3-anchor sm,selectve-3-anchor ana,hdv-hop sm,hdv-hop Lambda Fg. 7. Localzaton delay comparson based on V. COCLUSIOS In ths paper, we have presented a range-free localzaton scheme IDV-hop n WSs embedded n WLS method. t frst, nferortes of orgnal DV-hop scheme are denoted elaborately after brought brefly about orgnal DV-hop. Then, an mproved localzaton scheme embedded n WLS method s proposed based on these nferortes, and two crtcal parameters, correcton coeffcent kc and weghted coeffcent ws,, are ntroduced nto ths scheme to mprove localzaton performance. nd then, performance of error and delay s valdated. 06 Journal of Communcatons Moreover, performance comparsons between IDV-hop other DV-hop based schemes are proposed. CKOWLEDGMET Ths work was supported by atonal ature Scence Foundaton of Chna under Grant o REFERECES [] G. Q. Mao, B. Fdan, and B. D. O. nderson, Wreless sensor network localzaton technques, Computer etworks, vol. 5, pp , 007. [] J. Cheon, H. Hwang, D. S. Km, and Y. Jung, IEEE ZgBee-Based tme-of-arrval estmaton for wreless sensor networks, Sensors, vol. 6, p. 03, 06. [3]. Boukerche, H.. B. Olvera, E. F. akamura,.. F. Lourero, Localzaton systems for wreless sensor networks, IEEE Wreless Communcatons, vol. 4, pp. 6, 007. [4] P. Pvato, L. Palopol, and D. Petr, IEEE accuracy of rssbased centrod localzaton algorthms n an ndoor envronment, IEEE Transactons on Instrumentaton and Measurement, vol. 60, pp , 0. [5] D. culescu and B. ath, d hoc postonng system (PS) usng O, n Proc. th nnual Jont Conference of the IEEE Computer and Communcatons, IEEE Socetes, 003, pp [6]. Stanoeva, S. Flposkac, V. In, and L. Kocarev, Cooperatve method for wreless sensor network localzaton, d Hoc etworks, vol. 40, pp. 6-7, 06. [7] J. Wang, P. Urrza, U. X. Han, and D. Cabrc, Weghted centrod localzaton algorthm: Theoretcal analyss and dstrbuted mplementaton, IEEE Transactons on Wreless Communcatons, vol. 0, pp , 0. [8] L. Doherty, K. S. Pster, and L. El Ghaou, Conve poston estmaton n wreless sensor networks, n Proc. IEEE Twenteth nnual Jont Conference of the IEEE Computer and Communcatons Socetes. 00, pp [9] Z. Lu, X. Feng, J. J. Zhang, T. L, and Y. L. Wang, n mproved GPSR algorthm based on energy gradent and PIT grd, Journal of Sensor, vol., pp. -7, 05. [0] D. cuescu and B. ath, DV based postonng n ad hoc networks, Telecommuncaton Systems, vol., pp. 6780, 003. []. Bulusu, J. Hedemann, and D. Estrn, GPS-less low cost outdoor localzaton for very small devces, IEEE Personal Communcatons Magazne, vol. 7, pp. 8-34, 000. [] H. Chen, K. ezak, P. Deng, and H. CheungSo, n mproved DV-hop localzaton algorthm wth reduced node locaton error for WSs, IEICE Transactons on Fundamentals of Electroncs, Communcatons and Computer Scences, vol. 8, pp. 3-36, 008. [3]. Boukerche,. B. F. Horaco, E. F. akamura, and. F. ntono, DV-Loc: scalable localzaton protocal usng vorono dagrams for wreless sensor networks, IEEE Wreless Communcatons, vol. 6, pp ,

9 Journal of Communcatons Vol., o., December 06 [4] S. Kumar and D. K. Lobyal, n advanced DV-hop localzaton algorthm for wreless sensor networks, Wreless Personal Communcatons, vol. 7, pp , 03. [5] Y. Hu and X. M. L, n mprovement of DV-hop localzaton algorthm for wreless sensor networks, Telecommuncaton Systems, vol. 53, pp. 3-8, 03. [6] H. Safa, noval localzaton algorthm for large scale wreless sensor networks, Computer Communcatons, vol. 45, pp. 3-46, 04. [7] G. Z. Song and D. Y. Tam, Two novel DV-hop localzaton algorthms for randomly deployed wreless sensor networks, Internatonal Journal of Dstrbuted Sensor etworks, pp. -9, 05. [8] S. Tomc and I. Meze, Improvements of DV-hop localzaton algorthm for wreless sensor networks, Telecommuncaton Systems, vol. 6, pp , 06. [9] L. Q. Gu, T. Val,. We, and R. Dalce, Improvement of range-free localzaton technology by a novel DV-hop protocol n wreless sensor networks, d hoc etworks, vol. 4, pp.55-73, 05. [0] J. L, J. M. Zhang, and X. D.. Lu, weghted DV-Hop localzaton scheme for wreless sensor networks, n Proc. IEEE Internatonal Conference on Scalable Computng and Communcatons, 009, pp [] X. Chen and B. L. Zhang, Improved DV-Hop node localzaton algorthm n wreless sensor networks, Internatonal Journal of Dstrbuted Sensor etworks, vol. 0, pp. -7, 0. Scence and Technology n Shangha Ocean Unversty. Zhu s man research nterests nclude wreless sensor networks, dgtal communcaton systems analyss and desgn, nformaton theory. Chunfeng Lv receved B.E. and M.E. degrees n automaton from the Unversty of Kunmng scence and technology, and Ph.D. degrees n department of Electronc, Informaton and Electrcal engneerng of Shangha Jaotong Unversty, Chna. Currently he s a lecture of SOU College of Engneerng Scence and Technology n Shangha Ocean Unversty. Lv s man research nterests nclude detecton technque usng WSs, dgtal communcaton systems analyss and desgn, nformaton theory. Zhengsu Tao receved M.S. degree n department of Electroncs and Informaton Engneerng from Huazhong Unversty of Scence and Technology, Chna, and a Ph.D. degree n department of Electroncs and Informaton Engneerng from Hongkong Unversty of Scence & Technology. Currently he s a Professor of department of Electronc, Informaton and Electrcal engneerng of Shangha Jaotong Unversty. Hs research nterests nclude detecton technque usng WSs. Janpng Zhu receved B.E. and M.E. degrees n automaton from the Unversty of Kunmng scence and technology n 000 and 005, respectvely, and Ph.D. degree n department of Electronc, Informaton and Electrcal engneerng of Shangha Jaotong Unversty, Chna. Currently she s an assocate professor of SOU College of Engneerng 06 Journal of Communcatons 065