Optimal Sleep Scheduling Scheme for Wireless Sensor Networks Based on Balanced Energy Consumption

Transcription

1 6 JOURAL OF COMPUTER, VOL. 8, O. 6, JUE 3 Optmal leep chedulng cheme for Wreless ensor etworks Based on Balanced Energy Consumpton han-shan Ma College of Computer cence and Technology, Chna Unversty of Mnng and Technology, Xuzhou, 6, Chna Emal: ssma@cumt.edu.cn Jan-sheng Qan, Yan-ng un College of Informaton and Electrcal Engneer, Chna Unversty of Mnng and Technology, Xuzhou, 6, Chna Abstract ode schedulng scheme of sensor nodes s one of the most mportant method to solve the energy-constraned wreless sensor networks. Because there are the defects that hgh computatonal complexty of exact locaton nformaton and the energy consumpton unbalance of locaton-unaware n tradtonal schemes. Amng at these problems, an optmal sleep schedulng scheme based on balanced energy consumpton (ECB) was proposed n ths paper. Accountng the resdual energy, the precson for node redundancy evaluatng was mproved by usng the dstance nformaton between the sensor and ts neghbors. The numercal experments results llustrate that our schedulng scheme may mprove the energy effcency and extends the network lfetme whle ensure the coverage requrement. Index Terms wreless sensor networks; node schedulng algorthm; energy balance; Locaton-Unaware I. ITRODUCTIO Rapd advances n mcro-electro-mechancal systems and wreless communcaton have led to the deployment of large scale wreless sensor networks (Ws). The potental applcatons of sensor networks are hghly vared, such as envronmental montorng lke temperature, humdty, sesmc events, vbratons, and so on. But the energy source of Ws often conssts of a battery wth a lmted energy budget; and t s dffcult or mpossble to replace the power supples for sensor nodes after deployed.o lfetme s the key performance measure for Ws []. ensors are usually deployed densely to prolong the network lfetme. But a hgh-densty network wll waste a lot of energy and cause severe problems such as redundancy, rado channel contenton. A broadly-used method s to place nodes n sleep mode by schedulng sensor nodes to work alternatvely. But selectng the optmal sensng ranges for all the sensors s a well-known P-hard problem []. Random puttng nodes to sleep mode for fxed tme nterval [3 and 4] would cause the network to synchronze and may generate some blnd ponts that cannot be montored by any sensors [5,6]. Based on the locaton of sensor nodes, some schedule schemes are known as GAF [7], PEA [8], C [9], etc. Usng the geography (locaton, drecton, or dstance) wth global poston system (GP) or the drectonal antenna technology may ensure the coverage and connectvty effectvely. But the costs of GP or other complcated hardware devces are too hgh for tny sensors. Due to the lmted processng and memory capabltes, t s not realstc to take the sensor nodes equpped wth specalzed hardware components such as GP nto mass producton []. Furthermore, most applcatons may not sut equp wth GP, such as underground, etc. odes schedulng schemes wthout locaton nformaton are more valuable n practcal. Wthout accurate geography nformaton, however, t s very hard to check whether a sensor s sensng area can be completely covered by other sensors. Fortunately, most applcatons may not requre complete coverage of the montored area. Fewer researchers have proposed the node schedulng schemes wthout the accurate locaton nformaton. Gao et al [] propose a mathematcal model to descrbe the redundancy n randomly deployed sensor networks. The results ndcate that: a sensor requres about neghbors to get a 9% probablty of beng a complete redundant sensor. If we only requre a sensor s 9% sensng area to be covered by ts neghbors, 5 neghbors are necessary. Based on ths theoretcal analyss, a Lghtweght Deployment-Aware chedulng (LDA) scheme to turn off redundant sensors has been proposed []. LDA uses a weghted random votng method to decde who wll be elgble to fall asleep. But LDA only consder a sensor s -hop neghbors whch can cause larger redundancy coverage. Youns proposed two dstrbuted protocols (LUC-I and LUC-P) rely on dstance between one-hop neghbors along wth advertsed tow-hop neghborhood nformaton [3]. In [4], L-Hsng et al presented range-based sleep schedulng (RB) protocol, an optmal sensor selecton patter to ensure the coverage qualty. These methods can effectvely reduce network energy consumpton wthout any locaton or drectonal nformaton. But none of them take the balance of energy consumpton nto account. The do:.434/cp

2 JOURAL OF COMPUTER, VOL. 8, O. 6, JUE 3 6 unbalanced energy consumpton means that the nodes nequalty sleeps. It leads to the number of nodes premature death, and then speed up those nodes ded n ths regon, called as funnelng effect. Thus the energy hole are formed and the network lfetme s reduced [5~8]. Ideally, all of the nodes deployed n the regon should be consumed ther energy at the same tme as possble. The resdual energy of the entre network s almost zero when the network s death. In ths paper, we propose an optmal sleep schedulng scheme (ECB) whch reles on approxmate neghbor dstances and two-hop neghbors nformaton but no locaton nformaton. mulaton results ndcate that our scheme not only prolongs the network lfetme, but also mproves energy effcency. The reset of the paper s organzed as follows. ecton II ntroduces the system model and problem statement. ecton III presents and analyzes the algorthm. In secton IV, we present our expermental results for performance evaluaton. Fnally, secton V gves a summary and concluson. the wreless communcaton module to send the data s on the transmttng crcut and the power amplfyng crcut. And the manly energy consumpton to receve the data focus on the recevng crcut. Under the reasonable R condton, the transmsson energy consumpton to send k bt data s: ET ( k, d) = Eelec k+ ε fs k d d < dcross over 4 Eelec k+ ε mp k d d dcross over and the recepton energy consumpton s E = E k. R elec Among the formulas, E elec s the energy consumpton coeffcent for the rado electroncs, ε fs and ε mp are the energy consumpton coeffcents for a power amplfer under dfferent condton. Rado parameters are set as tableⅠ. We only consder the data aggregaton, whle gnore other processng energy consumpton. The energy for performng data aggregaton s 5nJ/bt/sgnal. II. YTEM MODEL AD PROBLEM TATEMET A. ystem Model We consder sensor nodes for whch r t s the transmsson range and r s s the sensng range. And our analyss s based on the followng assumes: () sensors are statonary and are deployed randomly wthn an area; () A sensor s sensng range s a crcle area; (3) all sensors are supposed to have the same sensng range and no two sensors can be deployed exactly at a same locaton; (4) no geography nformaton s avalable; (5) a node can estmate the approxmately dstance between tself and a neghbor based on the receved sgnal strength[9],and fuson, conflct and retransmsson are not taken nto account when data transmttng; (6) r t r s, under ths condton, coverage mples connectvty[]. Defnton (eghbor nodes): the neghbor set of sensor s defned as () = { ℵ d (, ) rs, ℵ, }. Where ℵ represents the sensor set n the deployment regon. d(,) denotes the dstance between sensor and. Defnton -hop neghbor of sensor : () = { () d(, ) rs, ℵ }. Defnton 3 Half-hop neghbor of sensor : D( ) = { ( ) d(, ).5 rs, ℵ}. Defnton 4 etwork lfetme: the runnng tme of the network meetng the requred coverage. B. Energy Dsspaton In our smulatons, we use the same energy parameters and rado model as dscussed n [] whch are used wdely. In the model, the manly energy consumpton of TABLE I. RADIO PARAMETER Parameter Value Threshold dstance(dcrossover)(m) 87 E elec (nj/bt) 5 ε fs (pj/bt/m ) ε mp (pj/bt/m 4 ).3 Intal energy(j).5 Data packet sze(bts) 4 C. Problem tatement Assume that nodes are dstrbuted n a feld, and the number of the actve nodes s A. Then the sleep rato of the network s defned as: A Q = () The sleep rato s one of the standards for measurng the effcency of energy consumpton. When the total number of nodes n the network s fxed, the hgher the sleep rato, the better the energy can be saved. If θ s the desred coverage rate of the network, the obectve of sleep schedulng scheme s to maxmze the lfetme and the sleep rato of the network whle ensue the coverage rate of actve nodes meet the θ requrement. III. OPTIMAL LEEP CHEDULIG CHEME A. Coverage Redundancy Determnes

3 6 JOURAL OF COMPUTER, VOL. 8, O. 6, JUE 3 Fgure. upposed that sensor has a neghbor sensor. and denote the crcle sensng area covered by node and respectvely. d s the dstance between node and. And denotes the sensng area that s covered by node and, as shown n Fgure. Refer to [], we can get that: d d rs arccos drs d rs = () rs 4rs otherwse o from formula (), we can get that when the dstance between node and s less than or equal to.5r, the redundant coverage area s more than about 68.5% of. When the dstance of node and node s more than.75r, the area s very small, about.5. These results can be used n our nodes schedulng. If d.75r, the effects that node to node wll be gnored n ths paper. If θ s the percentage of the redundant area covered by all the neghbors of node. Refer to paper [, 3], θ can be expressed as θ = = = m ( ) () () = (3) () s the area that covered by sensor but not covered by ts neghbors. Then, f node has a neghbor node k and d k.5r. Based on formula () and (3), the θ of node can be expressed as m θ.3 ( ) (4) = k uppose node s a neghbor node of node. Based on the above defnton, the dstance between node and satsfy the condton: <d <r s. By mathematcal statstcs knowledge we can get the probablty dstrbuton functon of node s f( ) =. Then π ( r) the probablty that node s an -hop neghbor of node s: p= f( ) d = = 4π r 4 Usng the same method, we can calculate the probablty that node deployed n dfferent area around node. In ths paper, the regon around sensor s dvded nto three parts: < d.5r s,.5r s < d r s, r s < d.75r s. Accordng to probablty dstrbuton functon, the number of sensors that deployed n dfferent parts can be calculated. Combned wth formula (3) and (4), we can get table Ⅱ. Where K, M, L s the neghbor numbers n dfferent regons. When the redundancy coverage area of a sensor meets the requrement θ, ths sensor can be off-duty. K (d.5r s ) B. Optmal leep chedulng cheme In ths paper sensors nclude three states: actve, sleep, and pre-sleep. Pre-sleep state s set to avod a blnd hole s Fgure. TABLE II. REDUDACY WITH DIFFERET EIGHBOR M (.5r s <d r s ) W Three states exchange machne. when several neghbor nodes to sleep smultaneously [, 4]. As shown n Fgure, when an actve sensor meets L (r s <d.75r s ) The redundancy coverage rate θ( ) 4 8.% % % % % % % %

4 JOURAL OF COMPUTER, VOL. 8, O. 6, JUE 3 63 the condton to sleep, t enters the pre-sleep state wth a random short tme T w. If the node receved other sensor s sleep-message at the pre-sleep state, t wll return the actve state. Otherwse, t broadcasts tself sleep-message after watng T w tme and then goes to sleep state; fall asleep for a perod of tme Ts. Based on the classc LEACH cluster protocol, tme s dvded nto fxed-length tme perods called rounds. Each round begns wth a competton phase, n whch every node determnes whether t can be actve or sleep. Then those actve sensors enter nto clusterng and sensng. We detal the steps as follows. tep: etworks ntalzaton. We assumed that all sensors are actve ntally. Each sensor broadcasts messages to estmate the dstance between tself and ts every neghbor and then record these nformaton. Accordng to the Qo demand (the coverage rate θ) of network, snk broadcasts the system message ncludng the two parameters HT and AT. Where HT s the mnmum number of actve neghbors wth one half-hop neghbor and AT s the mnmum number of neghbor nodes that have no half-hop neghbor. For example, the network coverage (θ) s requred to 85%. Accordng to tableⅡ, we can set HT = 5 and AT = 8. Whle the coverage rateθs more than 9%, we can set HT = 6 and AT = 9. tart cluster heads broadcast hello messages and other actve nodes select the closest head to on. tep 4: ensng. tep 5: The current round end and return step. IV. IMULATIO REULT We focus on the constructon of one cover and assume that nodes are deployed randomly n a meter meter square. Each sensor has a sensng range of 5 meters. The transmttng, recevng (dlng), and sleepng power consumpton rato s :4:.[]. We conducted smulatons wth matlab smulator for comparng among two sleep schedulng methods: LDA and our proposed scheme (ECB). A. Coverage Effectveness et θ 9%. Run LDA and ECB at the same condton to compare. We sampled on the o. round respectvely as shown n Fgure 4 (only actve sensors are marked to see clearly). Only 58 nodes are actve n our algorthm, but 5 nodes are on-duty by LDA algorthm. And Fgure 5 shows the coverage condton wth the actve nodes on o. round by dfferent algorthm. It can be easy to see that the fewer numbers of actve nodes are needed n our algorthm to meet the same coverage requred and the sensors dstrbute more unform n Fgure 4(b). However, there s more redundancy coverage n Fgure 5(a). >HT Keep actve Y d> Y >AT The maxmal resdual energy of sensor n D > E Y ensor sleep Y ensor wth the mnmum resdual energy n D sleep 9 (a) LDA 8 End 7 6 Fgure 3 The schedulng process of an actve sensor tep : odes-schedulng. At the begnnng of each round, each actve node determnes whether t s a redundancy sensor or not. The schedulng scheme s detaled n Fgure 3. Where s the number of sensor s actve neghbors, and d s the number of sensor s half-hop neghbors, D s the set of sensor s half-hop neghbors, E s the resdual energy of sensor. tep 3: Clusterng. Actve nodes randomly select nodes as cluster heads based on LEACH algorthm. Then the (b) ECB Fgure 4 The dstrbuton of actve nodes on o. round

5 64 JOURAL OF COMPUTER, VOL. 8, O. 6, JUE 3 covered by those actve nodes at some tme. As shown n Fgure 6, the network coverage s reducng wth the network runnng usng both the two algorthms. The hgher the network coverage requred, the shorter survval tme of the network. Durng the ntal operaton, the two algorthms have mantaned a hgher coverage rate. But wth the operaton of network, more and more nodes exhausted ther energy, the network coverage also decreased. Furthermore, the coverage rate of ECB s always hgher than LDA at the same round durng the whole runnng tme. 8 (a) LDA the number of actve nodes θ=9%,lda θ=85%,lda θ=9%,ecb θ=85%,ecb 6 network coverage (b) ECB Fgure 5 The coverage condton on o. round.6 θ=9%,lda θ=85%,lda.55 θ=9%,ecb θ=85%,ecb runnng rounds Fgure 6 Comparsons of network coverage rato The network coverage (η) s the rato of the area covered by those actve nodes to the whole montorng area durng the nodes schedulng scheme runnng process. Aactve A η () t = (5) A A s the whole montorng area, and A actve s the area runnng round Fgure 7 Comparsons of actve nodes Fgure 7 shows the number of actve nodes durng the network runnng. As can be seen from Fgure 7 and Fgure 6, the number of actve nodes by ECB s always less than the number that used by LDA when the coverage rato meetng the requrement. Because there are more actve nodes n the early operaton by LDA, too much energy were consumed. The actve nodes decreased wth more and more nodes run out of ther energy. And the coverage percentage dropped from 98% to 5% quckly. But the number of actve nodes used by ECB algorthm s kept stablty n the whole runnng process. Usng the less actve nodes to meet a hgh coverage, thus the energy has been saved and the lfetme has been prolonged. B. etwork Lfetme Accordng to the defnton 4 n ths paper, network lfetme s the runnng tme of the network meetng the requred coverage. As llustrated n Fgure 8, the network lfetme s only 7 rounds wth no schedulng scheme. et θ 9%, usng LDA schedulng scheme the lfetme s 85 rounds and the frst dead node occurred on o.4 round. But by ECB schedulng scheme, the lfetme extends to 5 rounds and the frst dead node occurred on o.38 round. et θ 85%, the lfetme s rounds and the frst dead node occurred on o.7 round by LDA. But by ECB algorthm, the lfetme extends to 75 rounds and the frst dead node occurred on o.4 round. ECB algorthm can prolong the network lfetme effcently. And the lower requred coverage, the longer the network lfetme.

6 JOURAL OF COMPUTER, VOL. 8, O. 6, JUE 3 65 E () t = me () t = (6) The energy varance functon s: = D () t = E E [ E () t m () t ] (7) C. Energy Effcency Fgure 8 Comparson of network lfetme The average energy θ=9%,lda θ=85%,lda θ=9%,ecb θ=85%,ecb runnng round Fgure comparson of the average resdual energy θ=9%,lda θ=85%,lda θ=9%,ecb θ=85%,ecb Fgure 9 Comparson of sleep rato As mentoned above, the sleep rato s an mportant parameter to descrbe the stuaton of savng energy durng the operaton. When meetng the coverage requrement, the hgher the sleep rato, the better the energy can be saved. Fgure 9 shows that the sleep ratos of ECB are always hgher than that of LDA algorthm and mantan stablty n the whole runnng tme. Moreover wth dfferent coverage requrement, the sleep ratos of LDA are also much dfferent. The hgher the network coverage requres the lower sleep rato. But the sleep ratos of our algorthm have a lttle change. Fgure shows the average resdual energy of network durng operaton. It confrms that the resdual energy of ECB s always hgher than that of LDA on the same round. leep rato can only demonstrate the total condton of energy consumed, but not measure the balance of energy consumed. In ths paper, the average resdual energy and the energy varance functon are used to measure that the energy consumed s balanced or not at some tme [5]. Consderng the two values, the larger the average resdual energy and the smaller the energy varance, the better balance of the energy consumed n the network. The average resdual energy functon s: Energy varance runnng round Fgure comparson of the energy Varance From Fgure and Fgure, t can be seen that the ECB algorthm has a better balance of energy consumed. By LDA algorthm, the m E (t) decreased more rapdly and the D E (t) were larger. The experment data shows that usng LDA algorthm some nodes stll remaned more than 9% energy even when the network ded. But usng ECB algorthm, the maxmal rato of the resdual energy to the ntal energy was about 4% when the network ded. It also ndcates that LDA algorthm exts the problem that energy consumes uneven. Thus t wll lead to some nodes run out ther energy earler. And then energy hole are formed so as to make the network dyng prematurely. Ideally each node n a network runnng out ts energy at the same tme wll obtan the optmal energy effcency.

7 66 JOURAL OF COMPUTER, VOL. 8, O. 6, JUE 3 V. COCLUIO Energy savng n Ws has attracted a lot of attenton n the recent years. Extensve research has been conducted to address these lmtatons by developng schemes that can mprove resource effcency. In ths paper, we have ntroduced an optmal energy-effcent sleep schedulng scheme for Ws. Wthout accurate geography nformaton, the two-hop neghbors are consdered. mulaton results show that our schedulng scheme has mproved the sleep rato and extended the network lfetme. But n the smulaton experments, we dscovered that there s approxmately 7% resdual energy when the network ded. Consderng the death spread from the border of the montor regon to the central, we beleve that there s stll space to mprove. o, one of our future works s to fnd a soluton to allevate the nequalty sleep of the boundary nodes. ACKOWLEDGMET Ths work was supported under the atonal cence Foundaton of Chna (5947, 5457); The Chna Postdoctoral cence Foundaton (474). REFERECE [] Lu, X. Jang,. Horguch, T. T. Lee, "Analyss of random sleep scheme for wreless sensor networks", Internatonal Journal of ensor etworks, Vol. 7, o./, pp. 7-84,. [] Ossama You ns, rmvasan Ramasubramanan, and Marwan Krunz, Locaton-Unaware ensng Range Assgnment n ensor etwroks, etworkng 7, pp. -3, 7. [3] Lu C, Wu k, Xao Y, et al, Random coverage wth coverage wth guaranteed connectvty: ont schedulng for wreless sensor networks, IEEE Transactons on Parallel and Dstrbuted ystems, Vol. 7, o. 6, pp , 6 [4] Jang J, L F, et al, Random schedulng for wreless sensor networks, IPA 9, ydney: IEEE C Press, pp.34-33, 9 [5] Ln JW, Chen YT, Improvng the coverage of randomzed schedulng n wreless sensor networks, IEEE Transactons o Wreless Communcatons, Vol.7, o., pp , 8 [6] Qng L, Zh T, Mnmum node degree and k-connectvty of a wreless mult-hop network n bounded area, GLOBECOM 7, EW York: IEEE Press, pp. 96-3, 7 [7] Xu Y, Hedemann J, Estrn D, Geography-nformed energy conservaton for ad hoc routng, Proceedngs of ACM Conference on Moble Computng and etworkng, UA:ACM, pp.6-, [8] F.Ye,Zhong,. Lu, L. Zhang, PEA : A robust energy conservng protocol for long-lved sensor networks, n Proc. of the ACM MobCom Conf, pp.9-43,4 [9] GAO han, CHIH T U, LI Yng-shu, et al, ensor schedulng for k-coverage n wreless sensor networks, Moble Ad-hoc and ensor etworks, vol. 43, no.5, pp.68-8, 6 [] We We, Hu Yang, Hao Wang, etc, Queung chedule for Locaton Based on Wreless Ad-hoc etworks wth D-Cover Algorthm, Internatonal Journal of Dgtal Content Technology and ts Applcatons, vol.5, no., pp , [] Gao Y, Wu K, L F, Analyss on the Redundancy of Wreless ensor etworks. WA 3[C]. ew York: ACM Press, pp.8-4, 3 [] KUI Wu, et al, Lghtweght Deployment-Aware chedulng for Wreless ensor etworks, Moble etworks And Applcaton, vol., no.6, pp , 5 [3] Youns O, Krunz M, Ramasubramanan, Locaton-Unaware Coverage n Wreless ensor etworks, Ad Hoc etworks, vol.6, no.7, pp.78-97, 8 [4] L-Hsng Yen, Yang-Mn Cheng, Range-Based leep chedulng (RB) for Wreless ensor etworks, Wreless Pers Commun, Vol 48, o. 3, pp.4-43, 9 [5] Cheng Te Ee, Ruzena Bacsy, Congeston control and farness for many-to-one routng n sensor networks, Proc of the nd ACM Conf on Embedded etworked ensor ystems (enys).baltmore:acm Press,pp.48-6, 4 [6] Khaled Matrouk, Born Landfeldt, RETT-gen:a globally effcent routng protocol for wreless sensor networks by equalsng sensor energy and avodng energy holes, Ad Hoc etworks,vol.7, o.3, pp , 9 [7] Wu X B,Chen G,Das K, Avodng energy holes n wreless sensor netwrks wth non-unform node dstrbuton, IEEE Transactons on Parallel and Dstrbuted ystems,vol 9, o. 5, pp.7-7, 7 [8] L J,Mohapatra P, An analytcal model for the energy hole problem n many-to-one sensor networks, Proceedngs of the IEEE Vehcular Technology Conference. Dallas,TX,pp.7-75, 5 [9] WE C Y, MORRI R D, ETHARE W A. Dstance estmaton usng bdrectonal communcatons wthout synchronous clockng, IEEE Transactons on gnal Processng, vol.55, no.5, pp , 7 [] H. Zhang, J.C. Hou, Mantanng sensng coverage and connectvty n large sensor networks, Ad Hoc and ensor Wreless etworks, vol., no., pp.89-4, 5 [] J. Hll, R. zewczyk, A. Woo,. Hollar, D. Culler, and K. Pster, ystem Archtecture Drectons for etworked ensors, ACM IGPLA otces, vol.35, no., pp.93-4, [] Tan D, Georganas, Locaton and calculaton-free node-schedulng schemes n large wreless sensor networks, Ad Hoc etworks, vol., no., pp.65-85, 4 [3] Fan Gao-uan, un L-uan, Wang Ru-chuan, et al, on-unform dstrbuton node schedulng scheme n wreless sensor networks, Journal on Communcatons, vol.3, o.3, pp.-7, [4] Fan Gao-uan, Wang Ru-chuan, Huang Ha-png, et al, Tolerable Coverage Area Based ode chedulng Algorthm n Wreless ensor etworks, ACTA ELECTROICA IICA, Vol. 39, o., pp , [5] Jang Chang-ang, h We-ren, Tang Xan-lun, et al, Energy-Balanced Unequal Clusterng Routng Protocol for Wreless ensor etworks, Journal of oftware, vol. 3, o. 5, pp.-3, [6] hao-feng Jang, Mng-hua Yang, Han-tao ong, et al, An Enhanced permeter coverage based densty control algorthm for wreless sensor network, Proceedngs of the Thrd Internatonal Conference on Wreless and Moble Communcatons (ICWMC 7), Washngton: IEEE Computer ocety, 7

8 JOURAL OF COMPUTER, VOL. 8, O. 6, JUE 3 67 han-shan Ma, was born n 978, s currently a lecturer n Chna Unversty of Mnng and Technology. he receved the B.. n Electronc and Informaton Technology from Chna Unversty of Mnng and Technology, Xuzhou, Chna, n and the M.. n Communcaton and Informaton Engneerng from Chna Unversty of Mnng and Technology, Xuzhou, Chna, n 3. he s currently pursung the Ph. D. degree at Computer Applcaton Technology n College of Computer cence and Technology, Unversty of Mnng and Technology, from 7. Her research nterests nclude wreless sensor network and nformaton processng. Yan-Jng un, was born n 977, s currently a professor n Chna Unversty of Mnng and Technology. He receved hs Ph.D. degree n Communcaton and Informaton ystem from Chna Unversty of Mnng and Technology n 7. Hs research nterest ncludes wreless sensor network and embedded real-tme system. Jan-sheng Qan, was born n 964, s a professor and Ph.D. canddate tutor n Chna Unversty of Mnng and Technology currently. He receved the Ph. D degree n Control Theory and Control Engneerng from Chna Unversty of Mnng and Technology, Chna, n 3. Hs research nterest ncludes mne communcaton and wreless sensor network.