A New Type of Weighted DV-Hop Algorithm Based on Correction Factor in WSNs

Transcription

1 Journal of Communcatons Vol. 9, No. 9, September 2014 A New Type of Weghted DV-Hop Algorthm Based on Correcton Factor n WSNs Yng Wang, Zhy Fang, and Ln Chen Department of Computer scence and technology, Jln Unversty, Changchun, , Chna Emal: yngw0517@hotmal.com; fangzy@lu.edu.cn; cldreamng@163.com Abstract In wreless sensor networks, the node localzaton algorthm s very mportant. In ths paper, we have further studed the tradtonal DV-Hop algorthm and proposed a new type of weghted DV-Hop algorthm. Based on the selectng of referenced anchor nodes and the calculaton method of the average hop dstance, we propose a new type of localzaton algorthm called wdv-cf (weghted DV-Hop algorthm based on correcton factor). Frstly, the unknown nodes record the average hop dstance from all anchor nodes n one hop dstance. Secondly, the percentage error s ntroduced as the correcton factor to calculate the average hop dstance for the unknown node. The smulaton results show that, compared wth the orgnal DV-Hop, the wdv-cf algorthm has a remarkable mprovement n the localzaton accuracy for the sensor nodes n WSNs. Index Terms 1 DV-Hop, weghted, correcton factor, WSNs I. INTRODUCTION Node localzaton algorthm n WSNs s one of the research hotspots n recent years. Accordng to the changes n the network topology, node self-organzed selects localzaton algorthm to locate. Techncans requred obtanng crtcal data nformaton by detectng the target node, and then they can analyze the envronment condton of current node, whch requres the target node feedback the poston nformaton to the techncan [1]. Such as forest fres, sensor nodes can not only send out fre alarm sgnals, but also need to transmt the approxmate poston of the fre to montors. So for the localzaton algorthm, the prevous researchers and scholars have made a great contrbuton, gven a varety of localzaton algorthms. When usng or mprovng the algorthm, people also need to pay attenton to the lmtatons of sensor nodes, such as the lmtaton of node energy[2], the random dstrbuton of node and the fckle node envronment. However, each localzaton algorthm has ts own advantages and dsadvantages, not only try to mprove the localzaton accuracy, to reduce the tme complexty and space complexty, at the same tme to extend the servce lfe of the sensor nodes and solve the energy savng problem and so on [3]. Therefore, how to maxmze use of advantages and reduce dsadvantages s the ultmate goal for studyng the localzaton algorthms. In wreless sensor networks, the localzaton algorthms of the sensor nodes are commonly dvded nto two categores range-based algorthm and range-free algorthm [4]. Ths classfcaton s based on the hardware of sensor node whether need to estmate and measure the dstance of neghbor nodes. Range-based algorthm has hgher localzaton accuracy compared wth the rangefree algorthm, and t usually ncludes RSSI algorthm (based on the ntensty of the arrvng sgnal), AOA algorthm (Angle of Arrval), TDOA algorthm (Tme Dfference of Arrval), TOA algorthm (Tme of Arrval) and so on. Although the range-based algorthm can get more accurate postonng results, and can use the hardware equpment to remove nodes whch have a larger error, t has ncreased the captal requrements, and not sutable for applcaton n low-cost proects. In ths paper, we wll focus on the DV-Hop algorthm, whch belongs to the range-free algorthm [5]. In recent years, the researchers n the feld of Node localzaton algorthm n wreless sensor networks have made a lot of mprovements for DV-Hop localzaton algorthm. In [6], Wang et al. dvded the measurement area to mprove the accuracy and took advantage of the partal Hope-Sze to estmate the dstance nstead of the global average hop dstance whch s employed by n standard DV-Hop algorthm and ts varants. In [7], Yu et al. ntroduced threshold M and used the weghted average hop dstances of anchor nodes wthn M hops to calculate the average hop dstance of unknown nodes, by changng the method to calculate the average hop dstance. In [8], Tomc ntroduced the weghts to weght the average hop dstance by changng the selecton of crtera beacon nodes. Although the lterature mentoned above consders the average hop dstance n studyng nodes localzaton [9], they ust focus on makng a lttle change n computng the accuracy of average hop dstance. In ths paper, we focus on the DV-Hop algorthm, and put forward a new type of weghted DV-Hop algorthm based on correcton factor. Frst, we descrbed the DV- Hop algorthm n detal; secondly, we propose an mproved algorthm and descrbe the prncples of the mproved algorthm; fnally, we desgn a multple smulatons to verfy the effectveness of the new algorthm n WSNs. 1 Manuscrpt receved June 11, 2014; revsed September 16, Correspondng author emal: yngw0517@hotmal.com. do: /cm II. RELATED WORK 2014 Engneerng and Technology Publshng 699

2 Journal of Communcatons Vol. 9, No. 9, September 2014 A. DV-Hop Algorthm DV-Hop algorthm s one of the wdely used localzaton algorthms n wreless sensor networks, whch belongs to the range-free algorthm. The algorthm s one of the APS dstrbuted localzaton, proposed by Dragos Nculescu n Rutgers Unversty, based on the dstance vector routng algorthm and GPS postonng deas [10]. The man advantages of the DV-Hop algorthm are the low demand for hardware devces, convenent operaton, hgh effcency and low energy consumpton. Therefore, t can be wdely appled to practcal applcatons. In DV-Hop algorthm, the sensor nodes do not need to confgure addtonal hardware devces, and t mproves the relablty of node localzaton by recevng a great deal of redundant nformaton durng transmsson. The man dea of DV-Hop algorthm s that unknown nodes record average hop dstance form the frst receved anchor node as ther average hop dstance, and then calculate the length of path between unknown nodes and the anchor nodes by usng the average hop dstance and the mnmum hop counts between the unknown nodes and anchor nodes. Fnally, after gettng three or more anchor nodes locaton nformaton, use trlateraton method or maxmum lkelhood estmaton method to calculate ther coordnates. Process of node localzaton n DV-hop algorthm s dvded nto the followng three steps: Step 1: Calculate the mnmum hop count between the unknown node and each anchor node. Anchor nodes n wreless sensor networks broadcast the group of ts own locaton nformaton to the neghbor nodes by floodng algorthm, and the group ncludes ump dgtal secton whch s ntalzed to 0. The recevng nodes record the mnmum hop count to each anchor node, and add 1 to the hop count. The new hop nformaton wll be forwarded to the neghbor nodes. Step 2: Calculate the estmated dstance between the unknown nodes and the anchor nodes. Accordng to the poston nformaton and hop count recorded n the frst stage, each anchor node computes the average hop dstance by usng the equaton (1): HopSze ( x x )2 ( y y )2 h (1) where HopSze s the average hop dstance of anchor node ; (x, y) and (x, y) are the coordnates of the anchor nodes and ; h s the hop count between and. Anchor nodes broadcast calculated average hop dstance group wth a lfetme feld to the network. The unknown node only records the frst receved average hop dstance, and forwards t to the neghbor nodes. After obtanng average hop dstance, unknown nodes compute the dstance to each anchor node accordng to the hop count obtaned n the frst stage Engneerng and Technology Publshng Step 3: Calculate ther own coordnates by usng the trlateraton method or maxmum lkelhood estmaton method. The unknown nodes use the trlateraton method or maxmum lkelhood estmaton method to calculate ther own coordnates, accordng the dstance to each anchor node recorded n Step 2. The flow chart of the DV-Hop algorthm s shown n Fg. 1. Begn Intalzaton Anchor nodes broadcast ther locaton nformaton packet locaton,hops Unknown nodes record the packet wth mnmum hop count form anchor nodes Anchor nodes calculate average hop dstance and broadcast to network Unknown nodes record the frst receved average hop dstance, and forward t to neghbor nodes Unknown nodes compute the dstance to each beacon node Unknown nodes calculate ther poston by usng the trlateraton method or maxmum lkelhood estmaton method End Fg.1. The flow chart of DV-Hop algorthm. B. The Error Analyss of DV-Hop Algorthm DV-Hop algorthm s very smple, and t has convenent operaton, hgh effcency and low energy consumpton. It uses the average hop dstance to calculate the actual dstance, whch has a low demand for hardware devces. The dsadvantage s that usng hop dstance nstead of straght lne dstance causes some errors. What s more, consderng the factors such as network latency, the average hop dstance n DV-Hop algorthm s dffcult to guarantee that t s obtaned from the nearest anchor node. Therefore, the localzaton accuracy n DVHop algorthm needs further mproved. III. THE WDV-CF ALGORITHM AND THE FLOW CHART OF WDV-CF ALGORITHM A. Onehop-DV algorthm Onehop-DV algorthm s a weghted sensor node localzaton algorthm based on DV-Hop algorthm. Onehop-DV algorthm also has three stages n the process of node localzaton. In the Step 2 of Onehop-DV algorthm, the unknown node records all HopSze from anchor nodes n one hop dstance, and forwards t to the 700

3 Journal of Communcatons Vol. 9, No. 9, September 2014 neghbor nodes. After obtanng all the average hop dstance from the anchor nodes whch are n one hop dstance, unknown nodes take the average of all recorded average hop dstance as ther own average hop dstance. The Step 1 and Step 2 n Onehop-DV algorthm are same wth the correspondng stage n DV-Hop algorthm. Onehop-DV algorthm ust plays a connectng role. In the smulaton experments, we wll use the algorthm agan, so we do not do too much ntroducton here. B. The wdv-cf Algorthm By analyzng the steps n process of node localzaton n the DV-Hop algorthm, the man mpact factors of error are n Step 2 and Step 3, and localzaton accuracy of average hop dstance form unknown node n Step 2 s one of the man factors. In ths paper, we make some mprovements n the calculaton of average hop dstance n Step 2. The unknown node records all average hop dstance from anchor nodes n one hop dstance, and takes the average of all the recorded average hop dstance as ts own ntal average hop dstance. Then we take the recprocal of percentage error form anchor node plus one as the correcton factor for weghtng the ntal average hop dstance, and the weghted result s the fnal average hop dstance. The man steps of wdv-cf algorthm are as follows: 1) As the Step 1 n DV-Hop algorthm, anchor nodes n wreless sensor networks broadcast the group of ts own locaton nformaton to the neghbor nodes by usng the floodng algorthm, and the recevng nodes record the mnmum hop count to each anchor node. 2) Calculate the ntal average hop dstance. Anchor node compute the Hopsze by usng the equaton (1) n DV-Hop algorthm, and then broadcasts the calculated average hop dstance group wth a lfetme feld to the network. The unknown node records all HopSze from anchor nodes n one hop dstance, and forwards t to the neghbor nodes. After obtanng all the average hop dstance from the anchor nodes whch are n one hop dstance, unknown nodes compute the dstance to each anchor node. Each unknown node computes the ntal average hop dstance by usng the equaton (2). OHopSze HopSze AOneHop (2) AN Among them, OHopSze refers to the ntal average hop dstance; refers to the unknown node; refers to the anchor node; AOneHop refers to the collecton of adacent nodes of node n one hop dstance; AN refers to the number of elements n the collecton AOneHop. 3) Calculate the percentage error of anchor nodes. Multplyng the HopSze n Step 2 and mnmum hop count between anchor nodes and unknown nodes n Step 1, the result s the estmated dstance between anchor nodes. Then usng the dfferences between the actual dstance and estmated dstance of the anchor nodes dvded by the actual dstance, the result s the localzaton error of anchor nodes(as shown n equaton (3)). Errpercent D HopSze h (3) N where and are the anchor nodes; D s the actual dstance between and ; N denotes the number of anchor nodes adacent to node ; Errpercent s the localzaton error of node. Ths step can guarantee the unknown nodes receve the HopSze from the nearest anchor node, whch reduces the error of average hop dstance and mproves the localzaton accuracy. 4) Calculate the fnal average hop dstance. For each unknown node, uses all average hop dstance from anchor nodes n one hop dstance recorded n Step 2 and localzaton error of anchor nodes n Step 3 to calculate the fnal average hop dstance by usng the equaton (4) and equaton (5). rehopsze 1 (4) 1 Errpercent AOneHop HopSze AOneHop λ where s the unknown node and s the anchor nodes; s the weghtng factor; AOneHop represents the collecton of adacent nodes of node n one hop dstance; reohopsze represents the weghted revsed average hop dstance of node. 5) The unknown nodes calculate ther own coordnates by usng the trlateraton method or maxmum lkelhood estmaton method. The unknown nodes use the trlateraton method or maxmum lkelhood estmaton method to calculate ther own poston, accordng the dstance to each anchor node recorded n Step 4. In DV-Hop algorthm, the unknown nodes record the frst receved average hop dstance from anchor node as ther average hop dstance. Takng the envronmental condtons nto account, the frst receved average hop dstance s not necessarly comng from the nearest anchor node. Whle, n wdv-cf algorthm, the unknown nodes take the average hop dstance of anchor node n one hop dstance as the reference nformaton, whch can ensure that t s from the nearest node. Moreover, ntroducng the percentage error as the correcton factor to calculate the average hop dstance for the unknown node, reduces the error of average hop dstance and mproves the localzaton accuracy n WSNs. C. The Flow Chart of wdv-cf Algorthm The flow chart of the wdv-cf algorthm s shown n Fg. 2. IV. SIMULATION AND ANALYSIS (5) 2014 Engneerng and Technology Publshng 701

4 Journal of Communcatons Vol. 9, No. 9, September 2014 A. The Confguraton of Expermental Envronment Confguraton The algorthm s smulated by the software of MATLAB R2010a. MATLAB R2010a runs on the Wndows XP SP3 system wth Intel Core 2 Duo CPI and a 2GB of memory. The expermental detecton regon sze s 100 m 100 m. The number of sensor nodes, proporton of anchor modes and the communcaton radus are determned accordng to the actual stuaton of the experment. In order to verfy the accuracy of the smulaton, the smulaton runs repeatedly 100 rounds. One of nodes dstrbuton s shown n Fgure 3. In 100 m 100 m square area, 150 nodes are randomly scattered. 150 scattered randomly through nodes, of whch there are 30 anchor nodes, and the communcaton radus s 30. Wth the changes n expermental condtons, the expermental parameters wll change accordngly. In Fg. 3, the red stars on behalf of anchor nodes and the black crcles represent the unknown nodes. B. Compared wth other Algorthms In ths secton, we wll compare the localzaton error of DV-Hop algorthm, Onehop-DV algorthm and twdvcf algorthm. We conducted the contrast experment n three dfferent condtons. Frstly, n a fxed area wth the same proporton of anchor nodes, the proporton of sensor nodes s a varable; Secondly, n a fxed area wth the same number of sensor nodes, the communcaton radus s a varable; fnally, n a fxed area wth the same communcaton radus, the densty of anchor nodes s a varable. 1) DV-Hop algorthm and Onehop-DV algorthm The localzaton error of both DV-Hop algorthm and Onehop-DV algorthm s compared n condtons of dfferent densty of sensor nodes as follows. In the smulaton area of 100 m 100 m, the communcaton radus s 40 and the proporton of anchor nodes s 20%. The scope of network node densty s 0.5, 1, 1.5, 2, 2.5, 3, 3.5 and 4. The expermental results are shown n Fg. 4 (a). The localzaton error of both DV-Hop algorthm and Onehop-DV algorthm s compared n condtons of dfferent communcaton radus as follows. In the smulaton area of 100 m 100 m, the number of sensor nodes s 100 and the proporton of anchor nodes s 20%. The scope of communcaton radus s 20, 25, 30, 35, 40, 45 and 50. The expermental results are shown n Fg. 4 (b). Begn Intalzaton Anchor nodes broadcast locaton nformaton to neghbors,and the unknown nodes record packet wth the mnmum hop count. Anchor nodes calculate average hop dstance and broadcast to network Unknown nodes calculate OhopSze accordng to HopSze receved from anchor nodes n one hop dstance Anchor nodes calculate the percentage error (Errpercent) Unknown nodes calculate the fnal average hop dstance FHopSze based on weghted OhopSze Unknown nodes compute the dstance to each beacon node Unknown nodes calculate ther poston by usng the trlateraton method or maxmum lkelhood estmaton method End Fg. 2.The flow chart of wdv-cf algorthm Fg. 4. (a) Localzaton error wth dfferent densty of sensor nodes Fg. 3.The dstrbuton of sensor nodes 2014 Engneerng and Technology Publshng Fg. 4. (b) Localzaton error wth dfferent communcaton radus. 702

5 Journal of Communcatons Vol. 9, No. 9, September 2014 The localzaton error of both DV-Hop algorthm and Onehop-DV algorthm s compared n condtons of dfferent proporton of anchor nodes as follows. In the smulaton area of 100 m 100 m, the number of sensor nodes s 200 and the communcaton radus s 30. The scope of proporton of anchor nodes s 0.05, 0.1, 0.15, 0.2, 0.25, 0.3, 0.35 and 0.4. The expermental results are shown n Fg. 4 (c). scope of proporton of anchor nodes s 0.05, 0.1, 0.15, 0.2, 0.25, 0.3, 0.35 and 0.4. The expermental results are shown n Fg. 5 (c). Fg. 5. (a) Localzaton error wth dfferent densty of sensor nodes Fg. 4.(c) Localzaton error wth dfferent proporton of anchor nodes Comparson chart of localzaton error for DV-Hop algorthm and Onehop-DV algorthm s shown n Fg. 4. Among them, the abscssa represents the vbratonal parameters, and the vertcal axs ndcates the localzaton error. As can be seen from Fg. 4, Onehop-DV algorthm s superor to the DV-Hop algorthm. The theoretcal analyss has proved that, obtanng the average hop dstance from anchor nodes n one hop dstance n Onehop-DV algorthm have ensured that unknown node can receve the average hop dstance from the nearest anchor node. In Onehop-DV algorthm, the unknown nodes have chosen a more approprate reference node and thus mprovng the localzaton accuracy. 2) Onehop-DV algorthm and wdv-cf algorthm The localzaton error of both Onehop-DV algorthm and wdv-cf algorthm s compared n condtons of dfferent densty of sensor nodes as follows. In the smulaton area of 100 m 100 m, the communcaton radus s 40 and the proporton of anchor nodes s 20%. The scope of network node densty s 0.5, 1, 1.5, 2, 2.5, 3, 3.5 and 4. The expermental results are shown n Fg. 5 (a). The localzaton error of both Onehop-DV algorthm and wdv-cf algorthm s compared n condtons of dfferent communcaton radus as follows. In the smulaton area of 100 m 100 m, the number of sensor nodes s 100 and the proporton of anchor nodes s 20%. The scope of communcaton radus s 20, 25, 30, 35, 40, 45 and 50. The expermental results are shown n Fg. 5 (b). The localzaton error of both Onehop-DV algorthm and wdv-cf algorthm s compared n condtons of dfferent proporton of anchor nodes as follows. In the smulaton area of 100 m 100 m, the number of sensor nodes s 200 and the communcaton radus s 30. The Fg. 5. (b) Localzaton error wth dfferent communcaton radus. Fg. 5. (c) Localzaton error wth dfferent proporton of anchor nodes Comparson chart of localzaton error of Onehop-DV algorthm and wdv-cf algorthm s shown n Fg. 5. Among them, the abscssa represents the vbratonal parameters, and the vertcal axs ndcates the localzaton error. As can be seen from Fg.5, wdv-cf algorthm has a low localzaton error than Onehop-DV algorthm, so wdv-cf algorthm s superor to the Onehop-DV algorthm. The theoretcal analyss has proved that, based on the Onehop-DV algorthm, wdv-cf algorthm ntroduces the 2014 Engneerng and Technology Publshng 703

6 Journal of Communcatons Vol. 9, No. 9, September 2014 percentage error as the correcton factor to calculate the average hop dstance from the unknown node, whch reduces the error of average hop dstance and mproves the localzaton accuracy n WSN. 3) Comparson of three algorthms The localzaton error of DV-Hop, Onehop-DV and wdv-cf algorthm s compared n the same condtons as follows. In the smulaton area of 100 m 100 m, the number of sensor node s 120, and the communcaton radus s 30 and the number of anchor nodes s 30. The expermental results are shown n Fg. 6. In ths paper, we have further studed the tradtonal DV-Hop algorthm. In the step of computng the average hop dstance of the unknown nodes, we made some mprovements based on the selectng of referenced anchor nodes and the calculaton method of the average hop dstance from the unknown node; we have proposed a new type of weghted DV-Hop algorthm based on correcton factor (wdv-cf) n WSNs. Smulaton results show that, compared wth the orgnal DV-Hop and the Onehop-DV algorthm, the wdv-cf algorthm has mproved the postonng accuracy of the sensor nodes sgnfcantly. However, the wdv-cf algorthm stll has some defcences, such as the postonng problem of solated nodes, communcaton overhead, etc. Therefore, about the ssue of node localzaton n WSNs, we wll do further research. REFERENCES K. Shrawan and D. K. Lobyal, An advanced DV-Hop localzaton algorthm for wreless sensor networks, Wreless Personal Communcatons, vol. 71, pp , July [2] X. L. Huang, Target localzaton based on mproved DV-Hop algorthm n wreless sensor networks, Journal of Networks, vol. 9, pp , [3] K. Z. Lu, X. P. Yan, and F. P. Hu, A modfed DV-Hop localzaton algorthm for wreless sensor networks, n Proc. IEEE Internatonal Conference on Intellgent Computng and Intellgent Systems, November. 2009, pp [4] J. Chao, G. J. Han, C. Zhu, and L. Shu, Performance evaluaton of DV-hop localzaton algorthm wth moblty models for moble wreless sensor networks, n Proc. 9th Internatonal Wreless Communcatons and Moble Computng Conference, July. 2013, pp [5] A. A. Anl, A. A. Amrta, and R. S. Patl, Evaluaton of DV hop localzaton algorthm n wreless sensor networks, n Proc. Internatonal Conference on Advances n Moble Networks, Communcaton and Its Applcatons, August 2012, pp [6] R. J. Wang, B. Zhang, and Y. Shen, PHDV-Hop: A more accurate DV-Hop postonng algorthm n WSN, Internatonal Journal of Dgtal Content Technology and ts Applcatons, vol. 6, no. 13, pp , July [7] Y. Hu and X. M. L, An mprovement of DV-Hop localzaton algorthm for wreless sensor networks, Telecommuncaton Systems, vol. 53, pp , [8] S. Tomc and I. Meze, Improved DV-Hop localzaton algorthm for wreless sensor networks, n Proc. IEEE 10th Jublee Internatonal Symposum on Intellgent Systems and Informatcs, September 2012, pp [9] B. J. Zhang and M. N. J, A weghted centrod localzaton algorthm based on DV-Hop for wreless sensor network, n Proc. conference on Wreless Communcatons, Networkng and Moble Computng, Shangha, Chna, September 21-23, [10] L. M. Sun, Wreless Sensor Network, ed. Tsnghua Unversty Press, Beng, 2010, ch. 6, pp [1] Fg. 6(a) Dstrbuton of nodes Fg. 6(b) Comparson chart of localzaton error for three algorthms As can be seen from Fg. 6, wdv-cf algorthm s superor to the Onehop-DV algorthm, and DV-Hop algorthm has the maxmum localzaton error obvously. In theory, Onehop-DV algorthm guarantees unknown nodes can receve the average hop dstance from the nearest anchor node by changng the selectng of referenced anchor nodes; the wdv-cf algorthm takes the percentage error as the correcton factor to weghtng the average hop dstance obtaned n Onehop-DV algorthm, whch reduces the localzaton error and mprove the postonng accuracy of sensor nodes n WSNs. In summary, the smulaton results show that, among the above three algorthms, the wdv-cf algorthm s more superor to the others. V. Yng Wang was born n May She receved bachelor of engneerng n Shaanx Normal Unversty n Chna. Now she s pursung master's degree n Computer Archtecture from School of Computer Scence and Technology, Jln Unversty, Changchun, Chna. Her research nterests are manly on the wreless sensor networks. CONCLUSION 2014 Engneerng and Technology Publshng 704

7 Journal of Communcatons Vol. 9, No. 9, September 2014 Zhy Fang was born n February, n And he receved Computer Scence Ph.D. n Jln Unversty, Changchun, Chna. And hs research nterests are manly on parallel/ dstrbuted systems and wreless sensor networks. He s a professor of Computer Scence and Technology Insttute, Jln Unversty, doctoral tutor. Hs keynote courses are "dstrbuted systems", "parallel computng system" and "computer system archtecture." He has vsted the Unversty of Calforna, Santa Barbara. He served as a senor vstng scholar n Unversty of Queensland, Australa. And he has publshed SCI/ EI retreval Internatonal Conference/ ournal artcles more than 40 artcles. Ln Chen was born n February, n She receved bachelor of engneerng n Jln Unversty n Chna. Now she s pursung master's degree n Computer Archtecture from School of Computer Scence and Technology, Jln Unversty, Changchun, Chna. Her research nterests are manly on the wreless sensor networks Engneerng and Technology Publshng 705