A VORONOI-BASED DEPTH-ADJUSTMENT SCHEME FOR UNDERWATER WIRELESS SENSOR NETWORKS

Transcription

1 INTENATIONAL JOUNAL ON MAT ENING AND INTELLIGENT YTEM VOL. 6, NO. 1, FEBUAY 013 A VOONOI-BAED DEPTH-ADJUTMENT CHEME FO UNDEWATE WIELE ENO NETWOK Jagao Wu, Ynan Wang, Lnfeng Lu College of Computer Nanjng Unversty of Posts and Telecommuncatons Nanjng, Jangsu, Chna Jangsu Hgh Technology esearch Key Laboratory for Wreless ensor Networks Nanjng, Jangsu, Chna Emals: jgwu@njupt.edu.cn; @njupt.edu.cn; lulf@njupt.edu.cn ubmtted: ep. 18, 01 Accepted: Dec., 01 Publshed: Feb. 0, 013 Abstract- Underwater wreless sensor network (UWN) s a specal knd of wreless sensor network whch s composed of a large quantty number of wreless sensor nodes deployed n the water. Whle there are extensve studes on deploy-ssue of terrestral wreless sensor networks (WN), UWN has not been pad enough attenton due to the challenges of UWN, such as low avalable bandwdth, hghly varyng multpath, and large propagaton delays. In ths paper, we propose a depth-adjustment scheme to maxmze the coverage n 3D space. After deployng nodes n the water surface, we use Vorono dagram to compute redundant nodes whose dsappearance wll not decrease the coverage n D space, and then we determne the depth that redundant nodes should be moved towards. After all the redundant nodes have moved to the lower layer, the algorthm contnues to schedule redundant nodes of the lower layer untl 3D space coverage s fulflled. Index terms: UWN, Vorono, 3D space, ensor Deployng. 44

2 Jagao Wu, Ynan Wang, Lnfeng Lu, A Vorono-Based Depth-Adjustment cheme For Underwater Wreless ensor Networks I. INTODUCTION ecently, UWN has receved consderable attenton. These knd of sensng networks consst of a large number of underwater sensng nodes whch can communcate wth each other usng acoustc sgnals or rado, and lmted number of surface snks whch wll collect data from underwater nodes. Dfferent from terrestral WN sensor deployng problems, sensor deployng of UWN has 3D requrement characterstcs whch ntroduce new challenges n terms wth coverage, connectvty and moblty [1]. Coverage guarantees every spot of the regon can be montored n the UWN, and connectvty guarantees the data can be transmtted so that nodes can relay ther montored data to on-shore staton. ef. [] has nvestgated the problem of achevng maxmal 3D coverage wth the least number of sensors, and suggested the sensor deployment pattern that creates the Vorono tessellaton of truncated octahedral cells n 3D space. Whle a lot of research has been done for node deployment and self-organzaton n terrestral WN [3], there s stll much to do for node deployment n UWN [4]. Depth-adjustment system effectvely resolves the problem of human nterventon n node deployng [5][6][7]. However, factually the underwater envronment s unknown or dangerous such as dsaster area, toxc regon or deep sea. In addton, most of the node deployment scheme requres a large number of underwater sensor nodes be placed n the pre-determned locaton, so human nterventon brngs about unnecessary spendng-tme and addtonal cost. Wth the advantage of depth-adjustment, coverage of the network n underwater crcumstance wll be enhanced autonomous [8]. But deployng a mass of nodes n order to acheve hgh coverage of gven regon wthout human nterventon would make nodes gather n a hgh level densty whch would arouse problems such as coverage overlaps, redundant nodes, communcaton nterference and energy waste. Thus, a purely random deployment sensng network wthout human nterventon s not practcal. Currently, Vorono dagram s frequently used n WN coverage optmzaton [9][10][11]. ef. [11] ntroduced an approach of schedulng nodes based on the threshold value of Vorono polygon area whch every node s responsble for. In our proposed scheme, we use a smlar dea to ef. [11]. However, we do not set up a gven area value as threshold. 45

3 INTENATIONAL JOUNAL ON MAT ENING AND INTELLIGENT YTEM VOL. 6, NO. 1, FEBUAY 013 In ths paper, we propose a mechansm to maxmze the coverage of the total montored regon wth lmted number of sensor nodes. We use Vorono dagram to determne whch sensor nodes n the same depth of the water are redundant node. To acheve maxmum coverage of the 3D montored regon, we let redundant nodes snk to a certan depth, and then contnue to use Vorono dagram to determne the new group of redundant nodes n the new layer of the nodes. Algorthm wll be stop untl there s no room for nodes to be decent. The organzaton of the paper s as follows. After a general ntroducton of the depth-adjustment scheme for UWN, the descrpton of our system model and assumpton s presented to the secton II. The scheme of depth-adjustment for underwater 3D space has been proposed n secton III. In secton IV, performance of our scheme s evaluated. The paper has been concluded n secton V. II. YTEM ILLUTATION AND DEFINITION a. Problem Defntons We propose the scheme attempt to maxmze the total coverage of 3D underwater space whle strvng to mnmze the total number of sensor nodes. Computatonal geometry s frequently used n WN coverage optmzaton, the most commonly used approach are Vorono dagram. In our proposed scheme, we also use Vorono dagram. To determnes f a node should adjust ts depth, we employ the average area of all the Vorono polygons that nodes n the same depth of the water are responsble for as the threshold. If the area of a Vorono polygon s smaller than the average area of the total Vorono polygons of the same depth, the node responsble for the Vorono polygon should be lst nto the schedulng lst. b. ystem Model and Assumpton Our proposed archtecture s depcted as Fgure 1. Intally, sensor nodes are random unformly spread on the surface of the ocean. In addton, a snk staton should be deployed n the center of the surface as a management node of all sensor nodes. Then, The snk construct Vorono polygons of sensor nodes based on receved ther locaton nformaton. After computng the 46

4 Jagao Wu, Ynan Wang, Lnfeng Lu, A Vorono-Based Depth-Adjustment cheme For Underwater Wreless ensor Networks average area of Vorono polygons, our algorthm compares each Vorono polygons and lsts the nodes whose responsble Vorono polygon has a smaller area than the average value nto the schedulng lst. Our algorthm runs teratvely untl sensor nodes reach the bottom of the ocean. To acheve coverage n underwater crcumstance, we assume that all these sensors have the ablty to adjust ther postons n vertcal drecton. In addton, each node knows ts local poston n the montored regon. In every round of the algorthm, nodes exchange ther locaton nformaton to ther neghbours. Further, we use ensng ange ( ) and Transmttng ange ( T ) represents the range of sensng ablty and communcaton ablty, respectvely. ur f ace s nk st at on Under wat er sensor EA Fl oor Fgure 1. Underwater ensor Deployment Archtecture c. Formal Defnton of our work 1) Vorono Dagram Vorono Dagram s an mportant structure n computatonal geometry. It represents the proxmty nformaton about a set of geometrc nodes. The Vorono dagram of a number of nodes dvdes the space nto polygons. Every pont n a gven polygon s closer to the node n ths polygon than to any other node. In a D regon V ( s ) Fgure. { x s x s j x,, we defne Vorono polygon as j} where s s the set of sensor nodes, whch s llustrated n 47

5 INTENATIONAL JOUNAL ON MAT ENING AND INTELLIGENT YTEM VOL. 6, NO. 1, FEBUAY 013 标标. ensor nodes Vor ono edge Vor ono Pol ygon- - V( s ) Fgure. Vorono Dagram ) chedulng Nodes Current Layer urface :chedulng Node 1 s Δh s Depth Fgure 3. Presentaton of h chedulng nodes are those nodes should adjust depth to a deeper layer. The determnaton of schedulng nodes can be llustrated as follows: um up all the areas of Vorono polygons n the current layer of the water and calculate the Average area ( Area ) of the total Vorono polygons wth Equaton (1). Area N A 1 N (1) 48

6 Jagao Wu, Ynan Wang, Lnfeng Lu, A Vorono-Based Depth-Adjustment cheme For Underwater Wreless ensor Networks Where A represents the area of Vorono polygon V s ), and N s the total number of nodes n the current layer. We call s as schedulng node f ts responsble Vorono polygon has a smaller ( area than the average value, e.t. A. Area 3) Depth Adjustng Depth that schedulng nodes should be adjust to s gven by Where Depth h () s h represents the depth dfference between the ntersecton of nodes coverage overlaps and the surface of the current layer of water. Accordng to Fgure 3, h (3) After the determnaton of all the schedulng nodes of the same layer, the algorthm compute the average dstance ( ) between every two nodes wth Equaton (4). N j d(, j) k d (, j) n Equaton (4) s the dstance between any two nodes s and s j, and k s the number of nodes pars. In a montored feld wth N nodes, nodes pars k could be decded by Equaton (5). (4) N *( N 1) k (5) From Fgure 3 we could see that the coverage overlaps among nodes are assocated wth the dstance between nodes. When, the gap between nodes would be much more bgger than what t s when.as the dstance contnues gettng larger,δh wll reach to 0 when. In the determnaton ofδh, we defer to the MINIMIZE prncple: Chose a smallerδh would make the blnd area that brought about by the nodes of upper layer be covered by nodes of lower layer. WhenΔh = 0, the blnd area may be well-fxed by nodes n lower layer. In the other hand, when reach a smaller value than s, dense nodes would brng about 49

7 INTENATIONAL JOUNAL ON MAT ENING AND INTELLIGENT YTEM VOL. 6, NO. 1, FEBUAY 013 extensve coverage overlaps, as the dstance nfntely close to 0, Δh nfntely close to : lm 0 h. nce 0 s the ultmate value that nfntely close to, the coverage blnd area would stll exst between layers fδh =, thus t cannot guarantee the coverage blnd area be covered by lower layer. nce lm h 3, from the consderaton of the MINIMIZE prncple, h 3 We summarze the range of would effectvely mnmze the blnd area n the stuaton of h spans correspondngly as follow: 0. h 3 0, Therefore, accordng to Equaton () and (6), the depth for adjustng wll be, 0 (6) Depth, 3, 0 (7) III. DEPTH-ADJUTMENT FO UNDEWATE 3D PACE uppose nodes are spread randomly and unformly n the surface of the montored regon wth the snk staton, the algorthm starts wth 5 steps: 1) coverng; ) Area Calculatng; 3) chedulng nodes determnaton; 4) Depth adjustng; 5) e-examnaton; 1) coverng Ths s the ntal phase to construct Vorono dagram of the network at the same depth. Frstly, nodes communcate wth the snk to broadcast ther poston nformaton wth a hello packet, whch contans node ID and planar coordnates. Then, the snk constructs the Vorono dagram of the network based on nodes poston nformaton delvered by these hello packets. ) Area Calculatng The snk calculates the area of each node's responsble Vorono polygon and the average area n ths phase. The area nformaton s stored n the area nformaton table. 3) chedulng nodes determnaton 50

8 Jagao Wu, Ynan Wang, Lnfeng Lu, A Vorono-Based Depth-Adjustment cheme For Underwater Wreless ensor Networks The snk searches n the table to determne whch nodes are schedulng nodes. 4) Depth Adjustng chedulng nodes move to the certan depth based on the average area as we talked about last ecton. It s worth notce that the snk wll empty ts area nformaton table and savng memory space for contnuous storng before the termnaton of ths phase. 5) e-examnaton After nodes descendng, the repetton of above phases wll extend to a deeper level, the newly generated layer by nodes descendng called "deeper layer". In ths phase algorthm checks f there s stll room for deeper expandng, f yes, the algorthm repeats on the deeper layer. TAT cover phase Generate Vorono dagram Calculate Vorono polygon Area A and Average area Area A Area Y Descendng N tayng each the sea bottom? Y N Done Fgure 4. The flow of the algorthm 51

9 INTENATIONAL JOUNAL ON MAT ENING AND INTELLIGENT YTEM VOL. 6, NO. 1, FEBUAY Input Nodes; Intalze Nodes ID and Nodes Poston; whle MaxDepth 0 Intalze Vorono Dagram; for all nodes do calculate Vorono_Area(ID); end for compute Average_tance ( ); f then Depth ; else Depth ( 1 3 ) ; end f Area calculate Average_Area(); for all nodes do f Voron_Area(ID)<Average_Area() then descendng_lst Node(ID); end f end for for all nodes n descendng_lst do Descendto(Depth); end for MaxDepth MaxDepth Depth; end whle Fgure 5. Pseudo-code of the algorthm Accordng to the steps above, the flow of the algorthm s depcted n Fgure 4. In addton, we show the pseudo-code of the algorthm n Fgure 5. At the begnnng of the algorthm, nodes transfer ther ID and locatons to the snk. Base on ths nformaton, the snk generates the 5

10 Jagao Wu, Ynan Wang, Lnfeng Lu, A Vorono-Based Depth-Adjustment cheme For Underwater Wreless ensor Networks Vorono Dagram, and calculate each node s relevant Vorono polygon area n Lne 4-7. Lne 8-13 determnes the depth that schedulng nodes should descend to. After comparng the average area and node s Vorono polygon area n Lne14-19, schedulng nodes are determned. Then, the schedulng nodes are descended to deeper layer n Lne 0-. Algorthm crculates untl nodes reach the bottom of the sea. Fgure 6. The fnal topology after the algorthm Fgure 6 shows the fnal topology of the algorthm wth 500 nodes are deployed randomly n a 500m 500m 00m 3D regon. We can see that after the executon of the algorthm, nodes are dstrbuted not only n the surface of the water but the entre regon of the water as we expected. IV. PEFOMANCE EVALUATION a. ettngs In ths secton we present the results obtaned from the smulaton and dscuss ther mplcatons. The smulatons were realzed by Matlab. Here, we consder the coverage of the whole regon and the connectvty between nodes as the man crterons of our network. 53

11 Coverage Percentage Coverage Percentage INTENATIONAL JOUNAL ON MAT ENING AND INTELLIGENT YTEM VOL. 6, NO. 1, FEBUAY 013 We start our smulaton wth varyng numbers of nodes unformly deployed n a gven space of 500m 500m 00m. We use the average value as our fnal results. Besdes, we assume that there were no errors occur n transmttng message phases and n nodes-descendng process. b. Coverage Number of ensors Fgure 7(a). ensor Number vs Coverage ensng ange(m) Fgure 7(b). vs Coverage 54

12 Connectvty Percentage Jagao Wu, Ynan Wang, Lnfeng Lu, A Vorono-Based Depth-Adjustment cheme For Underwater Wreless ensor Networks Obtanng an optmstc coverage n a gven 3D regon s the fundamental purpose of our scheme. Fgure 7(a) shows that coverage of the target space varyng wth dfferent number of nodes. The coverage acheves 98.6% wth 1400 sensor nodes whle the sensng range of each node s 50m. Wth the ncrease of sensors, t s obvously that there exst a sharp ncrease of coverage when number of nodes varyng between , meanwhle the whole coverage stays nearly stll wth nodes over 800. Ths suggests that 800 sensors are suffcent to cover the desre regon under the sensng range of 50m. In Fgure 7(b), the coverage of the network wth varyng sensng range s depcted. We access the coverage usng dfferent sensng range, and we set the number of nodes as 500. As expect, the coverage has a dstngushed mprovement when the sensng range ncreases. Under the sensng range of 70m, the network coverage reaches 99.15%. c. Connectvty Connectvty of nodes s crucal for transmttng messages as well. To guarantee the message can be transfer effectvely to the snk, we defne the connectvty of the network wth the percentage of nodes that can reach the surface snk. We used the followng metrcs to observe the connectvty performance: Total number of nodes: total number of sensors that deployed n the 3D space; snk : Transmttng ange of the surface snk; T / (r): rato of Transmttng ange to ensng ange of the sensor; r=1.0 r=1.1 r= Transmttng ange of nk(m) 55

13 Connectvty Percentage INTENATIONAL JOUNAL ON MAT ENING AND INTELLIGENT YTEM VOL. 6, NO. 1, FEBUAY 013 Fgure 8.(a) snk vs Connectvty r=1.1;snk=80m r=1.1;snk=55m r=1.0;snk=80m r=1.0;snk=55m Number of ensor Fgure 8.(b) ensor Number vs Connectvty Fgure 8(a) represents the relaton between snk and r. In ths experment we set total number of nodes=500. It s obvously that connectvty of the network has a wde dfference between r=1.0 and r=1.1. At the meantme, as snk gettng larger, connectvty has a gentle ncrease. nce nodes can communcate wth each other through mult-hops, t s easer for sensor to transmt data to the surface snk as ther Transmttng ange gettng larger. The results shown n Fgure 8(a) ndcate that the capablty of the snk has lttle nfluence on network connectvty though the connectvty of our approach s defned by the percentage of nodes that can reach the snk. After nodes descend to a certan depth beyond snk transmsson ablty, data transmsson manly depends on mult-hops between nodes. Fgure 8(b) shows connectvty wth varous node number, r and snk. As what we have learned from Fgure 8(a) that snk has lttle nfluence on connectvty, we can see from Fgure 8(b) that there are lttle changes n connectvty when snk change under the same number of nodes and the same r. Besdes, when the sensor number grows, relevant connectvty grows slowly. On the other sde, however, connectvty has a sharp ncrease when r>1.0. Note that nodes are randomly deployed n the water, the dstance between nodes can not be assure. As r grows, the transmttng capablty strengthens accordngly. Therefore, nodes can transmt data to longer neghbors under a certan sensng range. That leads to connectvty mprovement. 56

14 Jagao Wu, Ynan Wang, Lnfeng Lu, A Vorono-Based Depth-Adjustment cheme For Underwater Wreless ensor Networks V. CONCLUION In ths paper, we ntroduce a dstrbuted approach for under water 3D crcumstance. The proposed approach amed at coverng an underwater space wth less human nterventon. Although sensor nodes are randomly deploy n the surface of the water and can not move n horzontal drecton, wth the help of depth adjustment technque, sensors can be lower to any depth so that the coverage of underwater crcumstance could be nsure. The depth adjustment s done based on the densty of sensors, we use Vorono approach to calculate the densty to decde whch nodes should be descend down to a deeper layer. Experment turn out that our approach has a good performance n network coverage and connectvty. However, the experments also pont out that connectvty of the network s very senstve to T /. From Fg. 13 and Fg. 14, we learn that T / play a key role n network connectvty. Besdes, there stll exst some lmtatons to be consdered such as network lfetme and data transmsson effcency, whch should be done n our future work. ACKNOWLEDGMENT Ths research s supported by Open esearch Fund from Key Laboratory of Computer Network and Informaton Integraton n outheast Unversty, Mnstry of Educaton, Chna, under Grant No. K , K ; Natonal Natural cence Foundaton of Chna under Grants Nos , , , , ; Postdoctoral cence Foundaton of Chna under Grant No ; Talents tart esearch Foundaton of Nanjng Unversty of Posts and Telecommuncatons under Grant No. NY EFEENCE [1]. Meguuerdchan, F. Kaushanfar, M. Potkonjak and M. B. rvastava, Coverage Problems n Wreless Ad-hoc ensor Network, Proceedngs of IEEE INFOCOM 001, vol. 3, pp , Aprl,001, Alaska, UA. 57

15 INTENATIONAL JOUNAL ON MAT ENING AND INTELLIGENT YTEM VOL. 6, NO. 1, FEBUAY 013 []. N. Alam and Z. Haas, Coverage and connectvty n three-dmensonal networks, Proceedngs of the 1th Annual ACM Internatonal Conference on Moble Computng and Networkng (MobCom 06), pp , eptember, 006, Los Angeles, UA. [3] M. Youns and K. Akkaya, trateges and technques for node placement n wreless sensor networks: A survey, Elsever Ad Hoc Networks, vol. 6, no. 4, 008, pp [4] T. Clouqueur, V. hpatanasuphom, P. amanathan and K. k aluja, ensor Deployment trategy for Target Detecton, Proceedngs of the 1st ACM Internatonal Workshop on Wreless ensor Networks and Applcatons(WNA), pp. 4-48, eptember, 00, Atlanta, Georga, UA. [5] C. Detweler, M. Donec, I. Vaslescu, E. Basha, and D. us, Autonomous Depth Adjustment for Underwater ensor Networks, Proceedngs of the 5th ACM Internatonal Workshop on UnderWater Networks(WUWNet), pp.1-4, February, 010, Massachusetts, UA. [6] E. Cayrc, H. Tezcan, Wreless ensor Networks for Underwater urvellance ystems, Elsever Ad Hoc Networks, vol. 4, no.4, 006, pp [7] K. Akkaya and A. Newell, elf-deployment of ensors for Maxmzed Coverage n Underwater Acoustc ensor Networks, Computer Communcatons, vol. 3, no.7-10, 009, pp [8] N. A. A. Azz, K. A. Azz, and W. Z. W. Ismal, Coverage trateges for Wreless ensor Networks, World Academy of cence, Engneerng and Technology 6, 009, pp [9] G. Wang, G. Cao, and T. L. Porta, Movement-Asssted ensor Deployment, Proceedngs of IEEE INFOCOM 004, vol. 4, pp , March, 004, Hong Kong, Chna. [10] J. Jang, L. Fang, H. Zhang and W. Dou, An Algorthm for Mnmal Connected Cover et Problem n Wreless ensor Networks, Journal of oftware, vol.17, no., 006, pp [11] M. Vera, L. Vera, L. uz, A. Lourero, A. Fernandes and J. Noguera, chedulng Nodes n Wreless ensor Networks: a Vorono Approach, Proceedngs of the 8th Annual IEEE Internatonal Conference on Local Computer Networks. pp , October, 003, Bonn, Germany. 58