Coverage Optimization based on Redundant Sense Area Ratio in Wireless Multimedia Sensor Networks

Transcription

1 T.Srdev et al, / (IJCSIT) Internatonal Journal of Computer Scence and Informaton Technologes, Vol. 5 (), 04, Coverage Optmzaton based on Redundant Sense Area Rato n Wreless Multmeda Sensor Networks T.Srdev, R.Vnothn, N.Ramya, T.N.Prabhu Department Of Informaton Technology, Sr Ramakrshna Engneerng College, Combatore. Abstract-In wreless multmeda sensor networks coverage area and redundant sensor are a major problem thus throughput s mnmum and more delay.in order to optmze the network coverage, use a coverage-enhancng algorthm based on overlap-sense rato. By adjustng the sensng drectons of the nodes, the coverage area s ncreased wth the reducton of complexty. In addton modfed strategy algorthm s used to ncrease the network by reducng redundant. We proposes a load balancng algorthm whch ensures a far traffc load dstrbuton per lnk durng the network operaton and matches the values returned by the mathematcal plannng model for the set lfetme and throughput. Expermental results show that the proposed method acheves hgh throughput and less energy consumpton when compared to the exstng method. Keywords Coverage-enhancng, throughput, traffc load dstrbuton. I.INTRODUCTION Wreless sensor networks have drawn a more attenton n the last few years ncludng tradtonal WSNs and wreless multmeda sensor networks (WMSNs). The WSNs s the networks composed of low-cost, low-power, and small-sze sensors that have crcle sense area and communcate nformaton by multple hop and only provde smple sensng data, such as temperature, humdty, and so not as to meet the requrement of more complcated and precse data applcatons. But WMSNs s the dstrbuted sensng networks composed of vdeo cameras that have sector sense area and can process, send, and receve more ntensve and complcated vdeo nformaton data by packagng wth wreless transcever and dffer from the WSNs due to ther characterstc of drectvty and turnablty. In WMSNs, t can be determned by coordnate, radus of sensors, the drecton of sense sector, and the sze of separaton angle. After randomly deployed, Coverage-enhancng algorthm can maxmze coverage of a regon wth mnmum actvated sensors and maxmze networks lfetme wth maxmum redundant sensors. A coverage-enhancng algorthm s used based on overlap-sense rato (OSRCEA) for a gven regon. Assumng that parameters of neghborng sensors are known, the parameter of overlapsense rato (OSR) s ntroduced to represent the whole mpact of neghborng sensors []. In addton to that, a modfed strategy of shuttng off redundant sensor s proposed to prolong the network lfetme. But n ths method, the throughput s whch the traffc load that must be transmtted to each network lnk as a functon of the avalable battery level of the nodes. II. RELATED WORK The problem addressed n exstng model s two topcs namely.throughput,.more delay. Most of the proposed technques have gven only mplct assumptons about coverage area and network lfetme and dd not deal wth the throughput and delay [].Wthout addressng the problem of throughput and more delay, only the coverage area and network lfetme of wreless multmeda sensor networks s addressed. cluster the sensors nto several sets, such that sensors n each of the sets can completely perform the montorng task. Then, these sets are actvated successvely. We defne the actve set as a set of sensors whch perform the montorng task completely. Therefore, at any moment, only the sensors n one actve set go nto an actve state and perform the montorng task. On the contrary, all the other sensors, whch belong to the non actve sets, are n the low energy sleep state. Therefore, to maxmze the lfetme of the network, t s crtcal to rotate the roles of the actve set among the sensors n the network. It helps effcent usage of another scarce resource such as bandwdth. But they have hgh energy consumpton and delay [8].The problem of coverage by drectonal sensors wth tunable orentatons under the random deployment strategy. To develop solutons that maxmzes the number of targets to be covered whle mnmzng the number of sensors to be actvated at any nstant [3]. To provde hgher degree of coverage n whch multple sensors montor the same locaton at the same tme n order to obtan hgh confdence n detecton [8]. Consder the above defensves to ntroduce LOAM (optmal load balancng algorthm) n run tme, the traffc load that must be transmtted to each network lnk as a functon of the avalable battery level of the nodes to reduce the complexty of throughput and delay. III.SENSE THE DIRECTION OF REGION The overall technque s used to ncrease the throughput and less delay s shown n Fg. 79

2 T.Srdev et al, / (IJCSIT) Internatonal Journal of Computer Scence and Informaton Technologes, Vol. 5 (), 04, Fg 3 Sensng model wth negborng sensors Fg Overall Archtecure In ths paper, sensng feld wth nodes beng deployed randomly square regon s used to descrbe the square regon Fg. (a).the b- dmensonal sector area s utlzed to model the sensng regon of drectonal sensor, whch s llustrated n Fg. (b). (a) (b) Fg Sensng feld and drectonal sensng feld The drectonal sensng model can be represented by (Pᵢ, R,,α) where Pᵢ(x, y) denotes the locaton of nodes Sᵢ, R s the radus of a sensng regon that ndcates the maxmum sensng range of nodes, s a unt vector called sensng drecton that dvdes a sensng regon nto two parts. Sensng angle s denoted by α, where α descrbes the sensor s feld of vew (FOV). Intersecton angle between the sensng drecton of nodes Sᵢ and X axs s called the drecton angle denoted by φ wth the range of[0,п). In the sensng feld, neghborng sensors are the nodes whose Eucldean dstance to the current node s less than R. It s enough to consder the effect of neghborng sensors and adjust node s sensng drecton. It can be seen that S and S are the neghborng sensors of Sᵢ, S3 s not the neghborng sensor because the dstance between Sᵢ and S3 s larger than R. The grd part represents the overlappng regon of S. The centrod of the overlappng regon s denoted by Cen The centrod angle s the angle s the angle between X axs and the lne whch pass through Cen and the node, denoted by ᵝᵢ wth the range of[0,п)shown n Fg.3.Based on that dagram fnd the drecton angle for rotatng sensors to cover the maxmum area. IV. NOVEL COVERAGE METHOD FOR COVERAGE AREA Ths method can be used to ncrease coverage area A Overlappng Regon In ths Novel Coverage Method for Coverage Area use OSR parameter, denoted by η s used to reduce the overlappng regon of current node. The OSR s represented by rato of the overlappng area and sensng area, M η = () M Where M=αR² s the sensng area and M s the overlappng area. It can be seen that ηϵ[0, ].Dfferent from tradtonal vrtual force methods, OSR s a scalar operaton. There are two specal cases n the relatonshp between OSR and the rotatonal angle. When η based on that condton the node should rotate ts drecton wth the greatest rotaton angle maxmally reduce the overlappng regon,..e, the rotaton angle s equal to sensng angle α. Ths case s shown n Fg.4(a).where S s almost covered by S and S should rotate counterclockwse drecton wth the angle of α Smlarly S should rotate clockwse drecton wth the same angle of α. Therefore the result s no overlappng between the regon of S and S, whch s shown n Fg.4 (b) the other case η=0, n ths case no overlappng regon between S and S so t s unnecessary to adjust ts sensng drecton of nodes. Fg 4: Adjust of sensng drecton when η The rotaton angle of node Sᵢ s denoted as θ = α g( η) () Where ɡ(ηᵢ) s the OSR functon should satsfy the condton

3 T.Srdev et al, / (IJCSIT) Internatonal Journal of Computer Scence and Informaton Technologes, Vol. 5 (), 04, , η = g( η ) = 0, η < ηthreshold (0,) ηthreshold η < When ηᵢ s very small the correspondng rotaton angle also very small.it s dffcult for sensor to mplement n an practcal applcaton. So therefore η threshold s ntroduced to meet physcal realzaton n rotaton. When ηᵢ>=η threshold the node wth greatest rotaton angle turns a larger rotaton angle. B Evaluaton of Overlappng Area To obtan the optmal rotaton angle, overlappng area should be calculated accurately. Due to rregularty of the overlappng regon the grds are adopted to compute the overlappng area. Frst sensng radus and FOV are dvded nto M and N peces, so the sensng angle s dvded nto M x N peces, whch s shown n Fg.5 The Center of each pece s represented by pece s locaton coordnates. The area of pece s: Area (3) R α = ( p (p ) ) M N (4) Based on that equaton each peces of the varable s only related to p varable and other s constant. Fg 5 Computaton of overlappng Area Accordng to the overlappng area M = Area (5) p= q= Where covered mark n p-rows and q-columns of Sᵢ. It s one f t s p-rows and q-columns are covered by ts neghborng sensors otherwse t s zero. Substtute the equaton (5) nto () the OSR equaton s wrtten as: η M N p q Area = = = (6) α R C Determnaton Rule for Rotaton Drecton Node s rotatng drecton s equal to the locaton of Cen. Cen s denoted the dstrbuton of overlappng regon and the overlappng regon s represented by the sum of overlappng regon. When Cen s located n the clockwse drecton that seems more overlappng regon n the sensng drecton. So the nodes need to rotate counterclockwse drecton to reduce overlap and vce versa. Cen s located n the sensng drecton that ndcates both clockwse and counterclockwse are n overlappng area so the nodes can choose rotatng drecton randomly. The rotatng drecton s decde for the relatonshp between drecton angle φᵢ and centrod angle ᵝᵢ.It s not necessary to calculate the centrod angle there s no overlappng otherwse calculate the centrod angle ᵝᵢ. The Horzontal Cen of coordnates xcen M N p= q= = p= q= x The Vertcal Cen of Coordnates ycen M N p= q= = p= q= y Where x and y are the horzontal and vertcal coordnates of p-rows and q-columns of Sᵢ. When s obtaned through (7) and (8) equaton.the Eucldean dstance between Cen and node s: d = ( xcen xp) + ( ycen y p) Where x p and y p are the horzontal and vertcal coordnates of Sᵢ.Consderng the range of ᵝᵢ centrod angle nto two cases based on the vertcal coordnates when y cen <y p the centrod angle s: xcen xp β = π arccos (9) d When y cen >y p the centrod angle s: xcen xp β = arccos (0) d Once centrod angle s obtaned to determne the rotatng drecton by the relatonshp between drecton angle and centrod angle. Specfc determnaton rule s dvded nto the followng two stuatons: (7) (8) Clockwserotaton, ϕ< β 0 ϕ π α, Randomrotaton, ϕ= β Counterclockwserotaton, othercases Fg 5(a) Determnaton rule for rotatng 0 φ π -α () 73

4 T.Srdev et al, / (IJCSIT) Internatonal Journal of Computer Scence and Informaton Technologes, Vol. 5 (), 04, Counterclockwserotaton, ϕ > β π α < ϕ < π, Randomrotaton, ϕ = β Clockwserotaton, othercases Fg 5(b) Rule for rotatng π α < ϕ < π Algorthm OSR. Fnd neghborng sensors;. Set parameter state= 3. Whle(turn==) 4. Calculate OSR; 5. Node turns optmal angle accordng to the rotaton angle functon; 6. f(network s equlbrum) 7. turn=0; 8. end 9. end 0. calculate OSR;. whle(osr>=predefned threshold). calculate prorty; 3. f(prorty s hghest) 4. state=0; 5. send state nformaton to ts neghborng sensors 6. else 7. calculate OSR; 8. end 9. end V.MODIFIED STRATEGY The optmzaton of network coverage s fnshed but there s some redundant sensors n the network due to some sensors are overlappng wth the neghborng sensors. The node whose OSR s larger than predefned threshold there s some redundant sensors. It leads to unnecessary energy consumpton to keep the redundant sensors actve. In order to ncrease the network lfetme after optmze network coverage. There are several steps for shuttng redundant sensors. Frst compute the OSR for each node.then determne each node have redundant or not. There s redundant n the node set prorty for each node. If the prorty s hgh shut off the node otherwse lowest prorty wll be changed because of the closed sensor.it means some of redundant may be no longer redundant. It s necessary to decde whether unclosed nodes are stll redundant or not. The above step should be repeated there s no redundant node n the network. VI.LOAD OPTIMAL BALANCING ALGORITHM Optmal Load Balancng Algorthm s amed at approachng the best redstrbuton of the traffc load accordng to the connectvty results offered by the former optmzaton. Ths algorthm determnes, n runtme, the traffc load must be transmtted to each network lnk. In every node runnng LOAM dstrbutes the data flow among all ts neghbors n coverage from source to destnaton. In ths method the throughput per lnk has the restrcton mposed by the traffc generated (v) as lower lmt and the nomnal bandwdth of the network (Ψ) as upper bound. These bounds must be fulflled regardless of the type of traffc generated (sensng/montorng data or vdeo traffc). Therefore, once the value of the throughput per lnk s bonded. Fnally, the aggregate throughput at expresson s upper bounded by the sum of nomnal bandwdths of all avalable lnks and t s lower bounded by zero, whch s the lack of data transmssons. Ths fact assures the convergence of our calculatons because the outcomes must be comprsed between these bounds N N V N N D Maxmze^ = th.... j Kj S j th j Kj S = j + = = j= j V D thj > 0, thj > 0, TNetwork > 0, (, j) N, j V D thj = thj + thj thj ψ Where th j s acheved throughput of the lnk. th v js acheved throughput of source data of the lnk. denotes the aggregated throughput. K j s the parameter that ndcates the exstence of a lnk between nodes I and j.s ->j when a message s transmtted to a node, parameter that ndcates f ths node s nearer to the snk than the transmtter one. If S - >j = the nodes selected s nearer to the snk; otherwse S - >j=0.ψ denotes the maxmum transmsson bt rate. VII.PERFORMANCE EVALUATION In ths secton, the exstng and the proposed methods are compared. In ths exstng system, n order to enhance the coverage area a Coverage-Enhancng Algorthm s used based on overlap sense rato. By adjustng the sensng drecton of the nodes, the coverage area s ncreased wth the reducton of computatonal complexty. In the proposed system, a load balancng algorthm s used whch ensures a far traffc load dstrbuton per lnk durng the network operaton and matches the values returned by the mathematcal plannng model for the set lfetme and throughput Fg 6 Sensng radus 73

5 T.Srdev et al, / (IJCSIT) Internatonal Journal of Computer Scence and Informaton Technologes, Vol. 5 (), 04, The graph shows comparson of sensng radus n OSR algorthm and Load Optmal Balancng Algorthm.In OSR the sensng radus for covered percentage s larger than 0% but n the Load Optmal Balancng Algorthm sensng radus for covered percentage s hgher than 0%.Ths experment shows Load Optmal balancng algorthm has larger covered percentage. Fg 7 Increased percentage wth number of sensors When the number of sensors s more than 00, Load Optmal balancng algorthm gans more ncrease of coverage percentage than that of OSR. When the number of sensors s 00, the ncreased percentage of Load Optmal balancng algorthm s 50% hgher than that of OSR. Therefore t can make a concluson that no matter the nodes are dense or sparse the proposed algorthm can acheve more coverage enhancement. Fg 8 Comparson of convergence rate VIII.CONCLUSION In the Wreless multmeda sensor networks, a novel coverage-enhancng algorthm based on b-dmensonal sensng model s proposed to enhance the coverage area. The OSR (Overlap sense area) s presented to quantfy the total effect of neghborng sensors. The rotaton angle of node s obtaned accordng to the OSR and the rotatng drecton s determned by the centrod of overlappng regon. Furthermore, n order to prolong network lfetme, a modfed strategy based on prorty s proposed to shut off redundant sensors. To enhance the throughput the load balancng method s proposed n whch that the traffc load that must be transmtted to each network lnk as a functon of the avalable battery level of the nodes. It obtans, as a result throughput and network lfetme values smlar to those calculated by the prevous analytcal plannng model. For future work, to maxmze the multmeda qualty rate control schemes s used that s based on analytcal and emprcal models of vdeo dstorton conssts of a new cross-layer control algorthm that jontly regulates the endto-end data rate, the vdeo qualty, and the strength of the channel codng at the physcal layer. REFERENCES [] Jan Chen, Lu Zhang, and YounghongKuo, Coverage Enhancng Algorthm Based on Overlap-Sense Rato n Wreless Multmeda Sensor Networks,IEEE SENSORS JOURNAL,vol. 3,no.6,june. 03. [] J. A and A. A. Abouzed, Coverage by drectonal sensors n randomly deployed wreless sensor networks, J. Combnat. Optm., vol., no.,pp. 4, Feb [3] T. Yan, T. He, and J. A. Stankovc, Dfferentated survellance for sensor networks, n Proc. st Int. Conf. Embedded Netw. Sensor Syst., Oct. 003, pp [4] D. Tan and N. D. Georganas, A coverage preservng node schedulng scheme for large wreless sensor networks, n Proc. st ACM Int. Workshop Wreless Sensor Netw. Appl., Sep. 00, pp [5] I. F. Akyldz, T. Meloda, and K. R. Chowdhury, Wreless multmeda sensor networks: Applcatons and testbeds, IEEE Proc., vol. 96, no. 0, pp , Oct [6] J. Wang, C. Nu, and R. Shen, Prorty-based target coverage n drectonal sensor networks usng a genetc algorthm, Compu. `Math. Appl., vol. 57, nos., pp. 95 9, Jun [7] A. Ghosh, Estmatng coverage holes and enhancng coverage n mxed sensor networks, n Proc. 9th Annu. IEEE Int. Conf. Local Comput. Netw. Nov. 004, pp [8] N. Ahn and S. Park, An optmzaton algorthm for the maxmum lfetme coverage problems n wreless sensor network, Int. J. Manage. Sc., vol. 7, no., pp. 4, Nov. 0. [9] H. Huang, L. Sun, R. Wang, and J. L, A novel coverage enhancement algorthm for mage sensor networks, Int. J. Dstrb. Sens. Netw. vol. 0, no , pp., Mar. 0. [0] N. Tezcan and W. Wang, Self-orentng wreless multmeda sensor networks for maxmzng multmeda coverage, n Proc. IEEE Int. Conf. Commun., May 008, pp [] E. Yldz, K. Akkya, E. Sskoglu, and M. Sr, An exact algorthm for provdng mult-perspectve event coverage n wreless multmeda sensor network, n Proc. Int. Wreless Commun. Moble Comput. Conf., Aug. 0, pp [] C. K. Lang, C. H. Tsan, and T. H. Chu, Coverage enhancng algorthms n drectonal sensor networks wth rotatable sensors, n Proc. IEEE Asa-Pacfc Servces Comput. Conf., Dec. 0, pp [3] B. Wang, Coverage problems n sensor networks: A survey, ACM Comput. Surv., vol. 43, no. 4, pp. 53, Oct