CESILA: Communication Circle External Square Intersection-Based WSN Localization Algorithm

Transcription

1 Sensors & Transducers 2013 by IFSA ttp:// CESILA: Communication Circle External Square Intersection-Based WSN Localization Algoritm Sun Hongyu, Fang Ziyi, Qu Guannan College of Computer Science and Tecnology Jilin University Cangcun, , P. R. Cina Received: 13 June 2013 /Accepted: 12 August 2013 /Publised: 20 August 2013 Abstract: Localization algoritm is one of te key tecnologies of WSN (wireless sensor networks), and tere are two kinds of WSN localization algoritm, distance-based and distance-free algoritms. APIT (approximate point in time) is a distance-free algoritm wic is not depends on ardware and as a relative low localization error rate. But tere are two obviously sortages of APIT, firstly, te average cover rate is not 100 %, and it will decrease wit te communication radius dramatically; secondly, te average cover rate will be lower and te localization error rate will be iger wen tere are obstacles in distribution environment. In order to solve te problem above, tis paper proposed a new localization algoritm CESILA (Communication Circle External Square Intersection-Based Localization Algoritm), tis algoritm can decrease te localization error by experience parameters, and can guarantee cover rate by external square intersection. Experiment sows tat te average localization error rate is 5 % lower tan APIT algoritm in regular distribution environment, and 7 % lower tan APIT in environment wit obstacles; te cover rate is 100 % watever te oter conditions are.. Copyrigt 2013 IFSA. Keywords: WSN, Localization algoritm, APIT, CESILA, Localization error rate, Localization cover rate, Different distribution. 1. Introduction WSN (Wireless sensor network) is organized by nodes wit caracters of small size, low price and low computing power, wic can organized automatically troug wireless communications. Tere are mainly four key tecnology of WSN, tey are routing, security, energy consumption and localization. Localization can provide support to oter aspect; terefore, more and more researcers pay teir attention to localization algoritm [1]. Tere are two kinds of nodes in WSN, one kind called ancor nodes wic can locate teir location wit ardware suc as GPS and etc., te oter kind called unknown nodes wic cannot locate temselves wit ardware. So, localization algoritm s goal is to locate te unknown nodes. Currently, te localization algoritms ave been separated into two kinds [2-4]: distance-based localization algoritm and distance-free localization algoritm, distance-based algoritm evaluate unknown node s location wit actual distances and angles of nodes wic can be get by ardware, distance-free algoritm, te distance-free localization algoritm as been used widely since te distance-free algoritm is lower in price and simple to distribute. Classical distance-free localization algoritms including Centroid algoritm, DV-Hop algoritm [5], Amorpous algoritm [6] and APIT [7, 8] algoritm, 128 Article number 1296

2 APIT algoritm as lowest average localization error rate among tem wic is appropriate to te applications require ig location accuracy suc as military target tracking, mine personnel positioning and etc.. But tere is a common caracter of tese kinds of environment, tat is full of obstacles in distribution environment, is it affect to te localization error rate and cover rate? Actually, because of te point in test, te average localization cover rate is low especially wen te communication radius is sort, and APIT algoritm is an obstacle-aware algoritm wic means te localization error rate is ig and cover rate is very ig in distribution environment wit obstacles. So, focusing on tose disadvantages of APIT, tis paper proposed a new localization algoritm wic can decrease te average localization error rate and increase te cover rate especially in environment wit obstacles. Tis paper is organized as follows: Sec 2, related work, performance of localization and principle of APIT are introduced, te simulations of APIT as been done in Sec 3, and te advantage and disadvantage of APIT is analysis troug te simulations, in Sec 4, te principles of CESILA (communication circle external square intersection-based localization algoritm) is introduced, and te comparison and experimental result as been given in Sec 4, and included in Sec Related Work 2.1. Performance of Localization Algoritm Performances are used to evaluate an algoritm, according to te usage and caracters of localization algoritm, te mainly performance is localization error rate, localization accuracy localization cover rate, grid density and ratio of localization error and cover rate, teir definition are as follows: 1) Localization error rate. Localization error rate presents te degree of te difference between te estimated coordinates and actual coordinates, te localization error rate is calculate as formula 1. error _ rate 2 2 (xie x i N it ) (yie y it ) (1) L( N ) R In formula (1), N presents assemble of node tat can be located by te localization algoritm, L presents te lengt of assemble of N, (x ie ) presents te estimated coordinates calculated by localization algoritm, (x it ), is te actual coordinates wen te node as been deployed. R presents te communication radius of te node. 2) Localization accuracy. Localization accuracy is opposite to localization error rate, te metod of calculating localization accuracy is sown as formula (2). localization _ accuracy 1 localization _ error (2) 3) Localization cover rate. Localization cover rate presents te ratio of located nodes to all unknown node, tus, if a localization algoritm as a iger means it as a better performance, te monitoring region will be larger simultaneously, te cover rate is calculated by formula (3): NResoved cov er _ rate (3) N Unknown In formula (3), N Re soved presents te number of located nodes, N Unknown presents te number of all unknown node, so cover rate is larger wit N Re soved. 4) Grid density. Grid density presents te relationsip of distribution side lengt and grid lengt; it will be calculate by formula 4: square _ L grid _ density (4) grid _ lengt 5) Ratio of localization error rate and cover rate. Ratio of localization error rate and cover rate presents bot te performance of localization error rate and cover rate, te Ratio will be calculated by formula (5). localization _ error ratio _ error _ cov er (5) cov er _ rate Form te formula (5), we can conclude tat te performance of te localization algoritm is proportional to ratio _ error _ cov er Principles of APIT Algoritm APIT (approximate point in test) algoritm is a distance-free WSN localization algoritm wic is proposed by Tian He and etc., te principles of APIT is triangle approac, te principle is sown as Fig. 1, assume tat tere are n unknown node wic can be communicate wit te unknown node, te algoritm will traverse C_n^3 different triangles, and calculate te overlap of te triangles. As sown in Fig. 1, te centroid of te overlap polygon is te estimated coordinates of unknown nodes. Te main prase of APIT is as followings: 1) Ancor node information collecting prase. Ancor node will broadcast message to its neigbor nodes, te message composed of ancor node ID, ancor node coordinates and te ancor node s signal strengt and etc. Te format of te message is sown as Fig

3 Fig. 3(a). Principles of PIT (unknown node in te triangle). Fig. 1. Principles of APIT algoritm. Fig. 2. Format of te message broadcasted by ancor nodes. Fig. 3(b). Principles of PIT (unknown node out of te triangle). Te node (includes oter ancor nodes and all unknown nodes) wo receive te broadcast message will work as te rules, first, it will analysis te ead of te message, and get te ID of te ancor node wo broadcasted te message; ten, it will judge if te ancor node already in its forwarding table, if it is, broadcast te message, but do not update te forwarding table; if it is not, also broadcast te message, but update te forwarding table wit te content of te message. 2) Point in test (PIT) executing pase. After te first pase, every unknown node mastered muc information of ancors nodes, so, te unknown node traverse every tree of tem and test if te unknown node in te triangle composed by te tree ancor nodes. Te principles of PIT is sown as Fig. 3(a) and Fig. 3(b), we can deduce form Fig. 3(a) tat, if unknown node in te triangle, wen te unknown node is close to one of te ancor nodes, it must be far away from te oter two nodes; we also can deduce tat if te unknown is out of te triangle sown as Fig. 3(b), wen te unknown node is far away from one of te ancor nodes, it must be far away from te oter two nodes simultaneously. However, te WSN node is static wen it as been deployed in te most conditions, so we cannot locate te unknown node by moving te node, instead, we can judge if te unknown in te triangle by calculate its neigbors and teir signal strengt. 3) Grid scanning pase. If te unknown node in te triangle composed by some tree ancor nodes, ten te grids count (initialized by 0) in te triangle will be crease by 1. 4) Centroid calculate pase. After te above tree steps, we can find te entire grid wo as te max count, ten calculate te centroid of te polygon tat composed by te grids. Te formulas are as sown in (6) and (7). L L x y x1 x 2... xn (6) n y1 y 2... yn (7) n In te formula, presents te orizontal coordinate of te unknown node, presents te vertical coordinate of te unknown node, ( x,y i i ) and presents te coordinates of te grids wit te max count. 3. Performance Analysis of APIT Localization Algoritm Te greatest advantage of APIT is te low localization error rate, but te inerent caracters of APIT decide tat tere are some obvious disadvantages as follows. 1) Low localization cover rate. Tere must be some unknown node not in any triangles, especially wen te communication radius is sort. 2) Localization error rate cange wit te distribution and environment dramatically, especially wen te distribution environment wit obstacles, because some nodes cannot communicate even teir pysical distance is very sort. For example, tis paper as analyzed te affection of localization error rate and cover rate of C-sape distribution environment. 3) localization error rate is related wit grid density, in a certain region, localization error rate will be lower wen te density is larger, but te localization error rate will go steady wen te grid density is large enoug, wat s te value of it? 130

4 In order to analyze te APIT algoritm, simulation and performance analysis as been done in tis paper. Te conditions of simulations are sown as Table 1. Te analyzed performance of APIT include localization error rate, cover rate, relationsip of grid density and localization error rate, C-sape affection of localization error rate and cover rate. Table 1. Initialized Condition of Apit Simulator. Rectangle-sape C-sape Experimental area m 2 300, 300, 1000, 700 Total number of nodes (n) Proportion of ancor nodes GPS error rate (GPS-R) , Communication radius 100, 200,, 200,, (R) 1000 (m) 1000 (m) certain limitation, but it will be coming steady wen te grid density is large enoug; as sown in Fig. 4 (c), localization error rate is also proportional to communication radius, wen te radius of te nodes equals wit te side of te distribution areas, te localization error rate is approximated to 60 % wic can be took as a limitation. Te reason of tis penomenon is tat more and more unknown nodes as been located as te communication radius is larger, but te localization error rate of tese node are larger and larger, since te average localization error rate increased. Tree groups of experiments ave been done in te conditions sow as Table 1. 1) Test te localization error rate. First, fixing te communication radius sort for R, let R=100 m, ten experiments in different random distribution as been done, and te localization error rate is sown as Fig. 4 (a). Secondly, canging te grid density, te grid density canged from L/10,L/20,, L/20 (L presents te side of te rectangle), te localization error rate is sown as Fig. 4 (b); Tirdly, canging te communication radius, let R equals 100, 200, 300,, 1000 separately, te localization error rate is sown as Fig. 4 (c). Fig. 4 (b). Average localization error wen grid density is cange (R=100 m). Fig. 4 (c). Average localization error wen communication radius canging. Fig. 4 (a). Average localization error rate in different distributions (R=100m). As sown in Fig. 4 (a), te average localization error rate is around 40 % wen te communication radius R=100 m; as sown in Fig. 4 (b), localization error rate is proportional to te grid density in a 2) Test te localization cover rate. First, fixing te communication radius sort for R, let R=100 m, ten experiments in different random distribution as been done, and te localization cover rate is sown as Fig. 5 (a); Secondly, canging te communication radius, let R equals 100, 200, 300,, 1000 separately, te localization cover rate is sown as Fig. 5 (b). As sown in Fig. 5 (a), te average localization cover rate is about 76 % - 90 %, te cover rate is increasing wit communication radius, wen te communications is about m, te cover rate can be reac 100 %. 131

5 Fig. 5 (a). Average localization cover rate in different distributions (R=100 m). Fig. 6 (b). Comparison of localization cover rate in different distribution (rectangle-sape distribution and C-sape distribution). Fig. 5 (b). Average localization error wen communication radius canging. 3) Test te affection of te different distributions of APIT s localization error rate and cover rate, te condition of te experiment are sown as table 1, te average localization error rate is sown as Fig. 6 (a), and te average rate is sown as Fig. 6 (b), te orizontal coordinate of bot Fig. 6 (a) and Fig. 6 (b) presents te communication radius. Fig. 6 (a). Comparison of localization error rate in different distribution (rectangle-sape distribution and C-sape distribution). As sown in Fig. 6(a), te localization error rate in C-sape distribution is iger tan te one in Rectangle-sape, and as sown in Fig. 6(b), localization cover rate of C-sape distribution is lower tan te one of Rectangle-sape. Concluded from te above, te main disadvantages of APIT are: 1) Average localization error rate is proportional to communication radius; te value of te error rate is range from 40 % to 60 %. 2) Te cover rate of APIT cannot reac 100 % wen te communication radius is less tan alf of te side of distribution area. 3) Not adaptive wen tere are obstacles in distributions environment, suc as C-sape distribution, te localization error rate is iger and te cover rate is lower compared wit te rectangle distributions. 4. Communication Circle External Square Intersection-Based Localization Algoritm Focusing on te tree disadvantage of APIT algoritm, tis paper proposed anoter geometry-intersection localization algoritm - CESILA (Communication Circle External Square Intersection-Based Localization Algoritm), te principles of CESILA is sown as Fig. 7 (a) and Fig. 7 (b). As sown in Fig. 7 (a), te red circle presents unknown nodes, and te blue circle presents te ancor nodes, te black circle presents te forwarding nodes, R is sort for communication radius, n presents te ops from unknown node to ancor node. As sown in te figure, te unknown node j must be in a circle of wic te ancor node i as te center of te circle and wit 2nR for te radius of te circle if unknown node j can communicate wit ancor node i troug n ops, so it must be in te external square of te circle. Tus we estimated 132

6 te coordinates of te unknown node by te intersection of te rectangles, sown as Fig. 7 (b), wic is easy to calculate and convenient to unified computing because te intersections is also rectangle. 2nR R In te formula, FT presents te forwarding table; te format of FT is te same as message format wic is compose of ID of ancor node, op from one ancor node to tis node, and te coordinates of ancor node. Te triple (i,op 1,( P ix,p iy )) presents information of ancor node i in forwarding tables. Te value of FT(NodeID) returns te ID of ancor node, FT(I,Hop) returns ops from ancor node i to tis node, nodes update teir forwarding table according to rules defined by formula (8). 2) Grid partition and scanning pase. Assume tat te distribution area is rectangle, tus te coordinates of eac grid is calculated by formula (9). Fig. 7(a). Principles of CESILA. L L f(i,j) x i n 2 n L (i 0,1,... ) L L n f y(i,j) j n 2 n (9) As sown in formula (9), L presents te side of distribution; n presents te grid density, (f x(i,j),f y(i,j))presents te coordinates of te grid, and present te location of te grid (row and column), te count of eac is initialized by 0, tat is g k (i,j), k presents te ID of tis unknown node, so, for eac unknown node s count g k (i,j) is calculated by formula (10). Fig. 7(b). Scematic diagram of rectangular intersecting. According to te principles of CESILA, te procedures of it are as follows: 1) Ancor node information collecting prase, ancor nodes broadcast message wit NodeID wic is used to identify nodes from eac oter, Hop, and coordinates of itself. Te format of te message is sown as Fig. 8. NodeID Hop Localization(P x, P y ) Fig. 8. Message formats broadcasting by ancor node. Wen te forwarding nodes receive te message, it will update forwarding table as formula (8) (i,op1,(p ix,p iy )) if ( i FT(NodeID)) (i,op 1,( P ix,p iy )) FT FT if ( i FT(NodeID)and(op1) FT(I,Hop)) (i,op 1,( P ix,p iy )) if ( i FT(NodeID)and(op1) FT(I,Hop)) (8) m g k (i,j)(location(k) ( 1, 1)) k 1 (10) g(i,j) 0 (location(k) ( 1, 1)) As sown in formula (10), g k (i,j) presents ancor node wo can communicate wit unknown node, m is te number of ancor node tat can communicate wit, if means te node is not located yet, te calculate metod of g k (i,j) is sown as formula (11). g k (i,j) 1 (xmin f x( i, j ) x max, ymin f y( i, j ) ymax g k (i,j) g k (i,j) (oters) (11) (3) Calculate te centroid of te intersection rectangle wic as te max count of te grids, te calculate formula is sown as formula (12). L L (x x ) k max min x 2 (y y ) k max min y 2 (12) 133

7 In te formula, x max, x min, y max, y min present te max orizontal coordinate, te min orizontal coordinate, te max vertical coordinate, and te min vertical coordinate among te grids wit te max count separately. 5. Simulation Result In order to prove te correctness and effective of CESI localization algoritm, simulations ave been done in tis paper. CESILA is simulated in MATLAB 2012, and te comparison of CESILA and APIT algoritm is given, te result analysis mainly about localization error rate, localization cover rate, grid density s affection to localization error rate and cover rate, different distribution s affection to localization error rate and cover rate, te condition of te simulation is also sown as Table 1. In order to prove te performance of CESILA, five experiments ave been done in tis paper, and te comparison as been done between APIT and CESILA. 1) Fixing te communication radius sort for R, let R=200 m, te results of localization error rate of te two algoritms in ten different Rectangle distributions are sown as Fig. 9. Fig. 10. Localization covers rate comparisons between APIT and CESILA. 3) Canging te grid density, let te grid lengt form 100 meters, 90 meters, 0.5 meters, te values of te gird lengt is sown as te orizontal axis, grid density is in inverse proportion to grid lengt. Te localization error rate in different grid density is sown as Fig. 11, we can get tat te localization error rate is in inverse proportion to grid density in a certain range, and te localization error rate will tend to a constant wen te gird density is large enoug. Fig. 9. Localization error rate comparisons between APIT and CESILA. 2) Canging te communication radius, let R equals 100 meters, 200 meters, 1000 meters separately, te results of localization cover rate of te two algoritm in ten is sown as Fig. 10, te orizontal coordinate of te figure presents te communication radius of te node, and vertical coordinate presents te localization cove rate of te algoritm. Form te Fig. 9, we can get tat te average localization error rate of CESILA is lower tan APIT by about 5.9%; and we can get form Fig. 10 tat te cover rate of CESILA is 100% watever te experimental conditions are, wile te cover rate of APIT is canging wit te communication radius. Fig. 11. Localization error rates in different grid density of CESILA. 4) Fixing te communication radius let R=200 meters, te localization error rate as been concluded in Rectangle distribution and C-Sape distribution, te result is sown as Fig. 12. Form Fig. 12, APIT is a distribution aware localization algoritm, and its localization error rate is iger wen te node is distributed in C-sape distributions, wile CESILA is not a distribution aware localization algoritm, and its localization error rate in C-sape distributions is te same as te one in Rectangle distributions. 5) Canging te communication radius R, let R equals 100 meters, 200 meters, 300 meters,, 1000 meters separately, te cover rate of te two algoritms are sown as Fig

8 CESILA is te communication external square intersection metod. In addition, te simulations and performance analysis as been done in tis paper, te simulation results sow tat te localization error rate of CESILA is lower tan APIT by about 5.9 %, and te cover rate of te CESILA is 100 % in any conditions. However, CESILA is limited by parameters, for example, te cover rate of CESILA migt reac 100 % and as a lower localization error rate wen te one eac side of te square is less tan 2nR, so wat is te optimized value of te one eac side of te square will be furter researc in future work. Fig. 12. Localization error rate in different distributions of CESILA and APIT. Fig. 13. Localization cover rate in different distributions of CESILA and APIT. Form Fig. 13, we can get tat APIT is a distribution aware localization algoritm, and its localization cover rate is lower wen te node is distributed in C-sape distributions, wile CESILA is not a distribution aware localization algoritm, and its localization cover rate in C-sape distributions is te same as te one in Rectangle distributions. 6. Conclusion Te researc of te tis paper is about one of te key tecnology of WSN, tat is localization algoritm, tis paper analyzed and summarized te disadvantages of APIT according to te problem of low cover rate of APIT. In order to solve te problem, a new localization algoritm CESILA as been proposed in tis paper, te principles of Reference [1]. M. B. Srivastava, R. Muntz and M. Potkonjak, Smart Kindergarten: Sensor-based Wireless Networks for Smart Developmental Problem-solving Environments, in Proceedings of te 7 t Annual International Conference on Mobile Computing and Networking, July 2001, Rome, Italy, pp [2]. J. Li, J. Jannotti, D. S. J. DeCouto, D. R. Karger and R. Morris, A Scalable Location Service for Geograpic Ad-Hoc Routing, in Proceedings of te Sixt Annual International Conference on Mobile Computing and Networking, August 2000, Boston, Massacusetts, USA, pp [3]. K. Amouris, S. Papavassiliou, M. Li, A Position-Based Multi-Zone Routing Protocol for Wide Area Mobile Ad-Hoc Networks, in Proceedings of te IEEE Veicular Tecnology Conference (VTC 99), May 1999, Houston, Texas, USA, Vol. 2, pp [4]. M. Mauve, J. Widmer and H. Hartenstein, A Survey on Position Based Routing in Mobile Ad-oc Networks, IEEE Network Magazine, Vol. 15, No. 6, November 2001, pp [5]. Hongbin Tan, Feng Liu, Researc and Implementation of APIT Positioning Algoritm in WSN, in Proceedings of te 9 t International Conference on Fuzzy Systems and Knowledge Discovery (FSKD 12), [6]. Amitangsu Pal, Localization Algoritms in Wireless Sensor Networks: Current Approaces and Future Callenges, Network Protocols and Algoritms, 2010, pp [7]. Yong Zou, Xin Ao, Sixiong Xia, An Improved APIT Node Self-localization Algoritm in WSN, in Proceedings of te 7 t World Congress on Intelligent Control and Automation, Congqing, Cina, June 2008, pp [8]. Ji Zeng Wang, Improvement on APIT Localization Algoritms for Wireless Sensor Networks, in Proceedings of te International Conference on Networks Security, Wireless Communications and Trusted Computing, 2009, pp Copyrigt, International Frequency Sensor Association (IFSA). All rigts reserved. (ttp:// 135