Equalization of Energy Consumption at Cluster Head for Prolonging Lifetime in Cluster-based Wireless Sensor Networks

Transcription

1 WSS RNSIONS on OMMUNIIONS Yung-a Huang, Neng-hung Wang, hin-ling hen, Jong-Shin hen, Zong-Min Guo qualization o nergy onsumtion at luster Head or Prolonging Lietime in luster-based Wireless Sensor Networks YUNG- HUNG, NNG-HUNG WNG, 3 HIN-LING HN, 4 JONG-SHIN HN 3 ZONG-MIN GUO Deartment o Inormation ommunication ngineering haoyang University o echnology, aichung ounty, IWN yahuang@cyut.edu.tw, htt:// Deartment o omuter Science Inormation ngineering National United University, IWN ncwang@nuu.edu.tw 3 Deartment o omuter Science Inormation ngineering haoyang University o echnology, aichung ounty,iwn clc@cyut.edu.tw 4 Deartment o Inormation ommunication ngineering haoyang University o echnology, aichung ounty, IWN jschen6@cyut.edu.tw bstract: - In cluster-based wireless sensor networks (WSN), the most o energy is consumed in cluster head which erorms data aggregations relaying. he lietime in cluster-based WSN is subjected by the lietime o cluster heads. hereore, in this aer, we equalize the energy consumtion o the cluster heads to rolong the lietime o WSN. ter equalization, we obtain the length o sides o cluster area. Numerical results show that the length o ar site cluster is shorter than the nearest cluster to erorm equalization or all three routing schemes. Moreover, simulation results show that the three-ho scheme can obtain the most energy eiciency than other two schemes. Key-Words: - Wireless sensor network, lustering, energy eiciency, luster head, nergy consumtion equalization, Network lietime. Introduction Recently, the raidly develoed technologies o microelectro-mechanical systems telecommunication battery make the small sensors comrise the caabilities o wireless communication data rocessing []. hese small sensors could be used as the surveillance the control caability under a certain environment. Secially, the location o wireless sensor network (WSN) could be a region where eole could not easily reach there is a diiculty to recharge the device energy. hereore, the energy eiciency o the sensor networks is an imortant research toic the lietime o WSNs could be considered as the most signiicant erormance in the WSN []. Moreover, there are two main issues in the lietime rolong roblems. One is to minimize the energy dissiation or all energy constrained nodes. he other one is to balance the energy dissiation o all nodes []. he energy in WSN is mainly consuming on the direct data transmission []. irstly, each sensor collects data delivers the data to the base station directly, called as sink. lying this mode, the sensor will have quick energy exhaustion i it is aart rom the base station. hus, this kind transmission scheme is not suitable in a large area []. hen, secondly, to enable communication between sensors not within each other s communication range, the common multi-ho routing rotocol is alied in the ad hoc wireless sensors communication networks [3]-[5]. In this scheme, several multi-ho aths exist to erorm the network connectivity. ach ath in the coniguration will have one link head to collects data rom sensors. very sensor node in the WSN sends both the sensing data o itsel the receiving data rom revious nodes to its closer node. hen, the destination node delivers the data collection in the ath to the base station. he nodes closer to the base station need more energy to send data because the scheme uses hierarchy transmission. However, due the highly comlexity in routing rotocols the ISSN: Issue 5, Volume 8, May 009

2 WSS RNSIONS on OMMUNIIONS Yung-a Huang, Neng-hung Wang, hin-ling hen, Jong-Shin hen, Zong-Min Guo most likely heavy load on the relaying nodes, this scheme is not suitable or the highly densely WSNs. he third scheme is the cluster-based one that those closer sensors belong to their own clusters. One o sensors, called cluster head (H), in each cluster is resonsible or delivering data back to the base station. In this scheme, the H erorms data comressing sending back to the base station. hus, the lietime o H may be shorter than that o other sensors [6]-[7]. hereore, or WSNs with a large number o energy-constrained sensors, it is very imortant to design an algorithm to organize sensors in clusters to minimize the energy used to communicate inormation rom all nodes to the base station. Moreover, the energy eiciency the lietime rolonging are the most imortant toics in WSNs [8]-[0]. s we know that both energy eiciency energy balancing are two main issues or rolonging the lietime WSNs. hen the structure o low-energy adative clustering hierarchy (LH) roosed by []-[3] is one o the initiate algorithm to balance the energy issues or cluster based WSNs. In LH, the H can be elected by uniorm rom number to uniormly take turn being the H. hen the energy consumtion or all nodes will be nearly balanced in a round []. However, due to the romly clustering the cluster area may be too large to consume large energy or the H. hen the energy dissiation becomes ineiciently. hereore, in this aer a ixed area clustering method is roosed to imrove the energy eiciency. Moreover, to equalize the energy consumtion o Hs can rolong the network lietime eectively. hereore, an analysis on the equalization o the energy consumtion o Hs is investigated in this aer. In the ollowing section, the network models o the WSN are described. hen, the analysis o energy consumtion o Hs is described. In order to investigate the energy eiciency on the clustering algorithms, numerical results are erormed to comare energy eiciency or dierent channel environments. inally, conclusions are given in inal section. Network Models he sensing area is assumed to be a ree sace area where there are no any obstacles all the nodes are not moveable. hen the wireless channel model can be exressed by ath loss model. hus, the received ower at receiver can be exressed by Pt Pr c () d where d is the distance between the receiver transmitter, P t is the transmission ower o the transmitter, c is the coeicient o the antennas. Moreover, to investigate a large sensing area we set the ath loss exonent between 6 [4]. o be simliication, the sensing nodes are assumed to be uniormly distributed by d s (/ m ). hen the data gathering in a cluster is roortional to its cluster area. o rolong the lietime o Hs, the energy consumtion o Hs in one round is exressed by H d x, () where d x is the distance between the Hs S, is the length o one side o the square cluster area, (J/m /m ) is the energy consumtion er area unit or transmission one meter. Due to the sensor nodes are uniormly deloyed in the sensing area, the data or H relaying is deendent on the area o cluster. esides, with the requirement o received ower by P r c c or eective communication link, where c is a constant related to channel noise, rom () () we obtain P, (3) where r is the total time duration o ackets transmission or one round 3 nalysis on nergy qualization at luster Heads 3. Direct ransmission In ractical, the geometry o the WSN is nonregular. However, the square is a basic area to be consisted o non-regular area. hus, or simliication, in this aer we adot square area aroximation or WSNs. he sensor area is with uniormly distributed Hs is shown in ig.. In ig., the symbol is reresented as a location o H whereas the symbol o is reresented as a location o the sensing node. When the cluster area is o rom distributed, the energy eiciency o sensor nodes on data transmitting is terrible [3]. ased on the coniguration o square area, the sensors are suosed to be sread out uniormly to the whole area. he data rom each cluster will be r t ISSN: Issue 5, Volume 8, May 009

3 WSS RNSIONS on OMMUNIIONS Yung-a Huang, Neng-hung Wang, hin-ling hen, Jong-Shin hen, Zong-Min Guo collected by the H these data will be sent back to the base station located at the oint (0, -). ig.. he sensing area in desired wireless sensor network. the square area o cluster. Due to the uniormly deloyed sensors in the area, the amount o data o Hs gathering relaying deends on the clustering area. urthermore, with direction transmission o the Hs to base station, the arer Hs consumes more energy. hereore, the arer area o clusters should be smaller than those nearer clusters, as the dividing area in ig.. In the areas o ig., we denote the six clusters by,,, D,, resectively. he length o one side the H or clusters,,, D,, are denoted by,,, D,, H, H, H, H D, H, H, resectively. he distance between Hs base station are denoted as d, d, d, d, d d, resectively. x x x xd x It is obvious that the distance between dierent clusters to S is dierent as shown in ig.. hus to simliy the analysis, the area o the cluster is assumed to be square. Moreover, to equalize to energy consumtion o Hs, it is obvious that the data transmitted by the arer cluster should be less than that o near clusters. We can calculate the energy consumtion o the Hs in ig. by distribute the length variable into (). hus, we obtain the energy consumtion o six Hs in one round by x, (4),(5) ig.. he clustering toology in a quarter o the WSN. In this aer, we deine the sensing area as shown in ig. where base station is deloyed at center o the area. However, to be simliication, we investigate the quarter area as shown in shadow area in ig. ig.. o develo the investigation or the energy consumtion equalization o the Hs, we assume that all Hs are deloyed in the centroid o, (6) D D D, (7) ISSN: Issue 5, Volume 8, May 009

4 (8) (9),resectively. o equalize the energy consumtion o Hs in one round, let equal the energy consumtion o H D, D. hen, we obtain the relationshi o D by D D. (0) hus, based on the length o cluster,, we can adjust the length o cluster D, D. urthermore, to arrange the length o cluster, we let the energy consumtion o H H be equalized, then obtain the relations o as () rom ig., it is observed that. () rom () (), we obtain the relations o, by ( ) ( ), resectively. Moreover, to equalize both with, we adjust the other side length o cluster cluster by, resectively. hen we obtain the relations o,,, by (3) (4),resectively. hereore, ater distributing ( ) ( ) in (3) (4) resectively, we obtain other side lengths o cluster cluster by ( ) 3 ( ) 4, resectively. Similarly, because cluster cluster are symmetrical to cluster cluster, resectively, the length o the sides o cluster cluster, the relations o,,, can be obtained as (5), (6), (7) resectively. hen the length o,,, can be obtained as the unction o by ( ), (8a) WSS RNSIONS on OMMUNIIONS Yung-a Huang, Neng-hung Wang, hin-ling hen, Jong-Shin hen, Zong-Min Guo ISSN: Issue 5, Volume 8, May 009

5 WSS RNSIONS on OMMUNIIONS Yung-a Huang, Neng-hung Wang, hin-ling hen, Jong-Shin hen, Zong-Min Guo ( ), (8b) 3 ( ), (8c) 4 ( ). (8d) o illustrate the relations o all length o the clusters with rom (0) (8), ig. 3 shows the relations or equalization o Hs in WSNs with the ath loss exonent 4. rom ig. 3, it is observed that with both 4 the length o other ive sides linearly deend on. ig. 4 shows that the length o the clusters m or equalization o Hs with 5. hus, we can obtain the ive lengths by the known ath loss exonent or equalization on the energy consumtion o the Hs. rom ig. 4, it is observed that other lengths o clusters decrease as increases with m. hus, we can obtain the ive lengths o D by the known ath loss exonent or equalization on the energy consumtion o the Hs. (a) ig. 4. he omarisons o the length o clusters m or equalization o Hs with wo-ho ransmission Direct transmission is the simlest scheme. However, the energy eiciency is the worst than other multi-ho schemes. In this section, we will discuss the equalization o Hs or two-ho scheme. irstly, the network area is redeined as shown in ig. 5. (b) ig. 3. he length o the clusters vs. or equalization o Hs with (a) (b) 4. ISSN: Issue 5, Volume 8, May 009

6 WSS RNSIONS on OMMUNIIONS Yung-a Huang, Neng-hung Wang, hin-ling hen, Jong-Shin hen, Zong-Min Guo hus, the length o cluster, y, can be obtained as the unction o the length o cluster, x, as y ( x ). (3) 5 However, the energy consumtion o H,,,,, should be also be equalized to,,, D,. onsequently, the data usion ratio η is ut into (9) to erorm ossible equalization o,,. We then obtain the relations o η, x y by η ( x y ) x y ( x y ).(4) ig. 5. he sensing area or two-ho cluster-based WSNs. rom (4), the data usion ratio η can be obtained as the unction o x, as η ( x ). (5) 6 We assume that the two-ho routing is deicted in ig. 6. hereore, the energy consumtion o our Hs,,,,,, D, in one round can be obtained by η ( ), x y x, (9) ( ), y x y, (0) ( ) ( ) x y x y, D, (),resectively. In (9)-(), the length, x y are the length o one side o cluster cluster, resectively as shown in ig. 6. In (9), the actor η is assumed to be the data usion ratio erormed by H,. o equalize the,,,,, D, is mean to adjust the length o sides, x y. irstly, we equalized,,, D,, then obtain the relations o x y by y ( ) ( ) ( ) x y x y x y. () ig. 6. he two-ho routing in WSNs. 3.3 hree-ho ransmission o erorm more energy eiciently, the three-ho transmission scheme is roosed as shown in ig. 7. hus, we urther analyze the energy equalization o Hs or the three-ho scheme. In ig. 7, we select two Hs, H,, in cluster to urther imrove the energy eiciency. he exact osition o H, is illustrated in ig. 8. ISSN: Issue 5, Volume 8, May 009

7 WSS RNSIONS on OMMUNIIONS Yung-a Huang, Neng-hung Wang, hin-ling hen, Jong-Shin hen, Zong-Min Guo where the distance between cluster head H, 5 S can be calculated by x. Moreover, the 9 distance between H, H D, is calculated by x. hen, we can exress the energy 3 consumtion o Hs, H, H, in one round by ( x y ), y, (7), D, ig. 7. he three-ho routing in WSNs. x y y x 3 y x 6, (8) resectively. o equalize the energy consumtion o H, H,, then we obtain the relations o x,h y by y ( x y ) x y x x y. (9) hus, we obtain the length o cluster, y, by the unction o the length o cluster, x, as y ( x ). (30) 7 ig. 8. he exact osition o H, in three-ho MSNs. Similarly, we can exress the energy consumtion o both Hs, H, in one round by, ( x y ) 5, η x 3, (6) However, in three-ho schemes, both the Hs o cluster must aggregate the data sensing rom whole area send to S. hen, likewise to the two-ho scheme, a data usion ratio η is alied or the H,. he energy consumtion o H, H, are urther equalized by η ( x y ) y 5 3 ( ) x y x. (3) ISSN: Issue 5, Volume 8, May 009

8 WSS RNSIONS on OMMUNIIONS Yung-a Huang, Neng-hung Wang, hin-ling hen, Jong-Shin hen, Zong-Min Guo hus, the data usion ratio η can be obtained by the unction o the length o cluster, x, as η ( x ). (3) 8 Due to the same osition or H, H, the symmetric routing shown in ig. 8, the relations o H, can also exressed by (8). rom ig. 0, it is seen that when the exonent o ath loss increases, both the data usion ratio η η decrease. he data usion ratio η o twoho scheme decrease more dee than η o three-ho scheme with 5, x x m or two-ho three-ho scheme resectively. 4 Numerical Results he duration that all the sensing data are transmitted to S is say one round. In one round excet the nodes transmitting receiving data, the nodes are assumed to be in slee mode in which the nodes do not consume any energy. In the numerical results or two-ho threeho scheme, ig. 9 shows the length comarisons o y x y vs. x with 5. rom ig. 9, it is obvious that when the ath loss exonent increases, both lengths y ( y ) decrease to shrink the area (data) or equalization o the,, (, ),. (, ). Moreover, orm ig. 0, it is observed that when the ath loss exonent increases, both data usion ratio η η decrease to shrink the transmission data o H, or equalization o,, (, ), (, ). ig. 9. he comarisons o lengths y y with 5, x x m or two-ho threeho scheme, resectively. ig. 0. he comarisons o data usion ratio η η with 5, x x m or two-ho three-ho scheme resectively. Moreover, we comare the sensing area or direct, two-ho three-ho schemes with 5 x x x 0m as shown in ig.. rom ig., it is observed that under equalization o energy consumtion o Hs, the sensing area o three-ho is about double o direct two-ho routing scheme with, 3. However, the sensing area o three-ho is slightly larger than both direct two-ho routing scheme as 4, 5. With the equalization requirements or the Hs, we urther investigate the total energy consumtion or the direct, two-ho three-ho based WSNs as shown in ig.. In ig., to airly comare the energy consumtion o three routing schemes in one round, we assume that 0000 nodes are uniormly distributed in the same sensing area normalized with x 0m. rom ig., it is aarent that multiho schemes can erorm more energy eiciency than direct transmission three-ho outerorm twoho scheme in energy eiciency or various 5. ISSN: Issue 5, Volume 8, May 009

9 WSS RNSIONS on OMMUNIIONS Yung-a Huang, Neng-hung Wang, hin-ling hen, Jong-Shin hen, Zong-Min Guo or each cluster. hus, we obtain the length o sides o cluster area or the equalization. Numerical results show that the length o ar site cluster is shorter than the nearest cluster to erorm equalization or all three routing schemes. Moreover, simulation results show that the three-ho scheme can obtain the most energy eiciency than other two schemes. ig.. he comarisons o sensing area or direct, two-ho three-ho schemes with 5 x x x 0 m. ig.. he comarisons o energy consumtion in one round or direct, two-ho three-ho schemes with 5. 5 onclusions In this aer, to rolong the network lietime we equalize the energy consumtion o Hs or clusterbased WSNs. We analyze the energy consumtion o Hs or direct, two-ho three-ho routing scheme or the Hs. ter the equalization o the energy consumtion o Hs, we calculate the area Reerences: [] D. uller, D. strin, M. Srivastava, Guest ditors' Introduction: Overview o Sensor Networks, I omuter, Vol. 37, Issue 8,. 4-49, ug []. J. Duarte-Melo M. Liu, nalysis o nergy onsumtion Lietime o Heterogeneous Wireless Sensor Networks, Proc. o I Global elecommunication onerence,. 5, Nov. 00. [3] V. Raghunathan,. Schurgers, S. Park M.. Srivastava, nergy-aware Wireless Microsensor Networks, I Signal Processing Magazine, Vol. 9, No., , March 00. [4]. Schurgers M.. Srivastava, nergy icient Routing in Wireless Sensor Networks, Proc. o I Military ommunications onerence, Vol., , Oct. 00. [5] I.. kyildiz, W. Su, Y.Sankarasubramaniam,. ayirci, Wireless Sensor Network: a Survey, omuter Networks, Vol. 38, , 00. [6] O. Younis S. ahmy, HD: a Hybrid, nergy-icient, Distributed lustering roach or d Hoc Sensor Networks, I rans. Mobile omuting, , 004. [7] J. Zhu S. Paavassiliou, On the nergy- icient Organization the Lietime o Multiho Sensor Networks, I ommun. Letters, Vol. 7, No., , Nov [8] J. Levendovszky,. ojarszky,. Karlocai,. Olah, nergy alancing by ombinatorial Otimization or Wireless Sensor Networks, WSS rans. ommun., Issue, Vol. 7,. 7 3, eb [9].-J. han, M.-. hen, Y.-. Huang, J.-Y. Lin,.-R. hen, Otimal luster Number Selection in d-hoc Wireless Sensor Networks, WSS rans. omm., Issue. 8, Vol. 7, , ug [0] J.-Y. hoi, S.J. Yim, Y. J. Huh Y.-H. hoi, Distributed dative Scheme or Detecting aults in Wireless Sensor Networks, WSS rans. omm., Issue, Vol. 8, , eb ISSN: Issue 5, Volume 8, May 009

10 WSS RNSIONS on OMMUNIIONS Yung-a Huang, Neng-hung Wang, hin-ling hen, Jong-Shin hen, Zong-Min Guo []. Mainwaring, D. uller, J. Polastre, R. Szewczyk, J. nderson, Wireless Sensor Networks or Habitat Monitoring, Proc. o he irst M International Worksho on Wireless Sensor Networks lications (WSN '0), Georgia, Setember 00. [] Y.-. Huang, W.-H. Luo, J. Sum, L.-H. hang,.-w. hang, R.-. hen, Lietime Perormance o an nergy icient lustering lgorithm or luster-ased Wireless Sensor Networks, in rontiers o High Perormance omuting Networking, LNS 4743, Sringer-Verlag erlin Heidelberg, , ug [3] R. Szewczyk. erencz, nergy Imlications o Network Sensor Designs, ech. Re., erkeley Wireless Research enter, Santa lara, ali, US, 000. ISSN: Issue 5, Volume 8, May 009