Energy Efficient Clustering Algorithm in Wireless Sensor Networks using Fuzzy Logic Control

Transcription

1 11 I Colloquium on Humanities, Science an ngineering Research (CHUSR 11), Dec , Penang nergy fficient Clustering Algorithm in Wireless Sensor Networks using Fuzzy Logic Control Z.W. Siew, A. Kiring, H.T. Yew, P. Neelakantan an K.T.K. Teo Moelling, Simulation an Computing Laboratory School of ngineering an Information Technology Universiti Malaysia Sabah Kota Kinabalu, Malaysia msclab@ums.eu.my, ktkteo@ieee.org Abstract In general, environment monitoring cluster base hierarchical routing protocol is among the most common protocol being opte ue to the loa balancing among each other sensor. Sensors are ranomly eploye in a specific area to collect useful information perioically for a few months or even a few years. Therefore, battery power limitation becomes a challenging issue. It is also impractical to maintain the network lifetime by changing the battery frequently. Low energy aaptive cluster hierarchical (LACH) is one of the common clustering protocols that will t the cluster hea base on the probability moel which will possibly lea to a reuce in network lifetime ue to tion of cluster hea with a least esire location in the network. For wireless sensor networks, the istribution of cluster hea stion irectly influences the network s lifetime. This paper presents factors which will affect the network lifetime an apply fuzzy logic base cluster hea stion conucte in base station. The base station consiers two stion criteria from sensor noes which are energy level an istance to the base station to st the suitable cluster hea that will prolong the first noe ie (FND) time, ata stream guarantee for every roun an also increase the throughput receive by the base station before FND. Keywors wireless sensor network; clustering hierarchical; first noe ie; fuzzy logic I. INTRODUCTION Wireless sensor networks contain hunres or thousans of sensor noes equippe with sensing, computing an communication abilities. The sensor noes can be equippe with ifferent sensors such as temperature sensor, humiity sensor, gas sensor et cetera. When it sense elements of the surrouning environment, the processing units insie the sensor noes will convert the analog signal to igital ata by using the analog to igital converter moule an further sens the ata to the base station for processing. Wireless sensor noes not only communicate irectly to base station, it can also communicate among its peers. In the coming future it is possible to implement smart aaptive sensor noes by loaing artificial intelligence algorithm embee into the sensor noes. It is challenging to esign the wireless sensor noes ue to the purpose of cost saving an the size of the wireless sensor noes has to be compact for convenience purpose. The traeoffs for being compact are limite computation capability, limite power source an small memory storage. The noes energy is the most important issue because the noes are small in size an it may be eploye in hazarous areas, thus making battery replacement unpractical an impossible. Therefore, it is more practical to save energy an prolong the network lifetime by improving the routing algorithm. Cluster base hierarchical routing protocol is an energy efficient routing protocol. In the cluster routing, the sensor noes will be ivie into a few groups with one cluster hea te for each group. The cluster hea collects ata from member noes in the same cluster an aggregates the collecte ata so that it can be transmitte to the base station. Implementing this protocol will significantly reuce the overall energy use an reuce the network congestion by only allowing the cluster hea to communicate with the base station [1, ]. LACH protocol is one of the most common cluster routing protocols which aims to achieve the loa balancing in sensor noes so it can prolong the network lifetime. In cluster routing protocol, energy consumption is concentrate on cluster heas which collect an aggregate the sense ata from member noes an forwar the aggregate information to the base station [3, 4]. In LACH protocol each sensor noes ts itself as a cluster hea base on the probability moel. ach sensor noe will become cluster hea in every cycle to evenly istribute the work loas. The main problem of LACH protocol is it only epens on the probability moel to t the cluster hea an therefore it is possible that no cluster heas or too many are ste in a single roun [5]. Furthermore, the ste cluster heas may be locate near to each other or even near to the ege of the networks which leas to inefficient energy istribution [6]. LACH-C uses a centralize algorithm to t the cluster heas where each noe sens information about its current location an energy level to the base station an ts the cluster hea using the simulate annealing algorithm [7]. Gupta propose a cluster hea tion metho using fuzzy logic base centralize algorithm [8]. By consiering three parameters which are energy, concentration an centrality of sensor noes, the base station will t the suitable cluster hea among all the noes which show the improvement of /11/$6. 11 I 39

2 11 I Colloquium on Humanities, Science an ngineering Research (CHUSR 11), Dec , Penang network lifetime compare to LACH. These cluster hea tion mechanism nees the location information of the sensor noe (possible etermine using a GPS receiver). To use the algorithm, the sensor noes nee to know the location information. At the first roun, location information will be transmitte to the base station an the algorithm assumes all sensor noes uner minimal mobility constraint. GPS receivers consume large amounts of power since GPS is one of the communication raios. At current stage, it is too expensive to inclue GPS in the sensor noes for wireless sensor networks. This project aims to maximize the FND perio, ata collection is guarantee for each roun an the propose metho oes not require the location information of the noes. A cluster hea stion mechanism in base station using fuzzy logic is propose. Fuzzy logic system can manipulate the linguistic rules into mathematical form; it can make real time ecision ue to the algorithm s simplicity. Moreover extra parameters such as battery charging rate, ata queue length are among the parameters that can be ae into fuzzy logic system to prouce an optimize solution [9, 1]. The paper is organize as follows. In the next section, overview of cluster base hierarchical routing protocol LACH an fuzzy logic cluster base are emonstrate. Section III explains the system moel followe by the analysis of the simulation results in section IV. Lastly, section V conclues the finings. II. RLATD WORK A. Cluster Base Hierarchical Routing Protocol In cluster routing, noes with more energy will become the cluster hea in the networks. By forming few clusters, the surrouning noe only nee to communicate to cluster hea in the same cluster to reuce the energy consumption use to communicate among other noes at ifferent clusters. LACH [] is one of the clustering mechanisms aim to istribute the energy concentration in the cluster hea. This protocol operates in rouns; each roun consists of two phases which are set-up phase an steay phase. Set-up phase inclue avertisement phase an cluster set-up phase while steay phase inclue scheule creation an ata transmission. In each roun, each noe inepenently generates the ranom number between an 1. If the generate number is less than the threshol value T (n) which efines by (1), the noe wills self t to become the cluster hea for the current roun. p, if n G 1 T ( n) = 1 p ( r mo ) (1) p, otherwise r is the roun which alreay ene; p is the proportion of the noes to cluster heas; G is a set of noes which have never been cluster hea in the last 1 p rouns. Although LACH protocol istributes the energy concentration in the cluster hea but it has some isavantages. LACH epens on only the probability moel to t the cluster hea, te cluster hea may be very close to each other. In each roun the number of the cluster hea is ynamic an it cannot guarantee the ata stream receive at the base station. Cluster hea will appear in the ege of the network or in the place where the noe ensity is very low. It can be observe that the isavantages of LACH protocol are mainly ue to only using the local information [3, 4]. By using centralize algorithm in the base station with global information, it will possibly st the suitable cluster hea in the network. B. Cluster-Hea lection using Fuzzy-Logic In the centralize algorithm, base station has the global view of the overall networks. The base station is more powerful than the sensor noes in term of computation power, sufficient memory, unlimite power an storage. Gupta prove that, by consiering three fuzzy parameters which are energy, concentration an centrality the network lifetime can be improve. nergy level is available in each noe, concentration is the number of neighbor noes an centrality is a value base on how central the noe to the cluster. In this project, expert knowlege is represente base on the following two escriptors: Noe nergy remaining energy level in the noe. Distance the istance between the noe an the base station. The remaining battery level can be obtaine from each noe via analog to igital converter (ADC) moule which is built in the microcontroller. The istance information can be acquire when the base station broacasts the te cluster hea information to the whole network an each noe calculates the istance to the base station base on the receive signal strength inicator (RSSI). III. SYSTM MODL Fig. 1 shows the basic linear topology use to stuy the effect of various parameters on the networks lifetime. BS represents the base station whereas noe A, B an C are the sensor noes. The relative istance between noe A an BS is set as 7R whereas noe B an C are locate 6R an 5R from the BS. The inputs of fuzzy inference system (FIS) are energy an istance of a particular noe to the base station. Base on the rule base, FIS will compute an ecie the output which is the fitness of a noe to become a cluster hea. Base station will st p total number of noes to become cluster heas base on their fitness. Sensor noes will be liste in a table from highest to lowest fitness. The table restores every single roun to keep the information up to ate. In short, the ste cluster hea shoul be the best leaer in every roun. 393

3 11 I Colloquium on Humanities, Science an ngineering Research (CHUSR 11), Dec , Penang R Figure 1. Basic linear moel. Below are some assumptions use in the system moelling. All the sensor noes can transmit ata with enough power to reach the base station an the base station have a wie transmission range which is capable of using a single broacast to reach all the sensor noes. The istance between sensor noes an base station can be measure base on the RSSI. All the sensor noes are homogenous, with the similar noes moel an same energy resources. A. Raio Moel Fig. shows the block iagram of the raio moel use in this project. The moel consists of the transmitter moel an receiver moel for energy calculation purpose. The raio moel referre from [3]. The expene energy uring transmission an reception for k bits message to a istance between transmitter an receiver noe is given by (). R A B C Tx ( k, ) = ( ( k, ) () k = k Tx + + kε + kε fs mp 4 Tx if < if Distance is the threshol for swapping lification moel, it can be calculate as = ε fs ε mp. To receive k -bit message, the raio moel expene as in (3). 5R ;, BS B. Simulation Parameters The simulation is carrie out in MATLAB with the raio moel parameters shown in Table I. The network topology an base station are varie from each case which will be iscusse later. C. Cluter Hea Stion Base on Distance Distance is one of the factors that may affect the energy consumption as shown in (). Distance refers to the relative istance among the cluster hea, member noes an the base station. quation (4) illustrates noe A being te as cluster hea, using the moifie linear topology as shown in Fig. 1. Distance 1, an 3 in (4) represent the total transmit istance by each sensor noes. = n( ε = 5n( k( + n( ) ) + ( + ( k + ε k + ε + ( n 1 k( ) ) + n( k( 1 3) + n ) + 3 k + In the case of n equal to 1 an n represents numbers of sensor noes, where noe C becomes the cluster hea, the overall energy consume is less than if noe B an noe A being ste as cluster hea.. This is ue to the fact that noe C is locate nearer to the base station, an also the istance to base station is far larger by other member noes. Therefore the istance to the base station is significantly affecting the energy consumption. The previous statements become invali when n not equal to 1. It can be prove by the following cases where can be calculate base on (4). Case 1: A become cluster hea, with n numbers of sensor noes in B an in C. = (5n + 1)( + (5nR + 49R Case : B become cluster hea, with n numbers of sensor noes in A an in C. (4) k bit packet ( k (3) Rx = Transmit lectronics k bit packet Tx () Tx Amplifier k k Rx Receive lectronics k Figure. Raio energy moel. = (5n + 1)( + (nr + 36R Case 3: C become cluster hea, with n numbers of sensor noes in A an in B. = (5n + 1)( + (5nR + 5R TABL I. PARAMTRS OF TH RADIO MODL Parameter Value Initial nergy,.5j Packet Size, k 4bit nergy of transceiver, 5nJ/bit nergy of ata aggregate, DA 5nJ/bit/message ε fs 1pJ/bit/m ε mp.13pj/bit/m 4 394

4 11 I Colloquium on Humanities, Science an ngineering Research (CHUSR 11), Dec , Penang Referring to Case if n = 5, noe B become cluster hea, an the overall energy consumption is less than Case 3 an follow by Case 1. It can be observe that noe B is more suitable to become cluster hea compare to noe C, since the n numbers of noes from A an B spen more energy to transmit to noe C. As a result, sting cluster hea in the mile of linear topologies will prouce less energy consumption per rouns. Accoring to equation above, cluster hea B will shows less energy consumption per rouns compare to cluster hea A an C, but the traeoff of it is FND occurre faster than noe C. D. Balance in nergy Distribution Although proper cluster hea tion base on istance can achieve minimum energy use per roun, but it will rain energy faster which leas to FND faster because it overloas the cluster hea. FND can be prolonge by istributing the workloa to maintain the level of remaining energy in each noe. This paper examines three protocols, namely irect communication, ranom cluster hea stion protocol an energy constraint cluster hea stion protocol using the same raio moel an same network topology (base station locate at x = 1m an y = 16m) as shown in Fig. 3. Fig. 4 shows the simulation results of the three protocols with 5 ranomly eploye sensor noes for uration of 15 rouns. Fig. 4 illustrates the network lifetime of the three protocols, it is observe that FND time for energy constraints cluster hea stion protocol is longer than ranom cluster hea stion protocol an followe by irect communication protocol. In irect communication protocol, each sensor sens its ata irectly to the base station. It will quickly rain the battery when the sensor noes are far away from the base station because it requires a large amount of transmitting power. On the other han, since only one ranomly ste cluster hea sens ata (gathere from sensor noes) to the base station for ranom cluster hea stion protocol, therefore it outperforms the irect communication protocol. Y-coorinate (m) Base station Sensor noes X-coorinate (m) Figure 3. Moifie topology moel. Number of Alive Noes energy ranom irect Rouns Figure 4. Network lifetime 1. In ranom cluster hea stion protocol, the base station will ranomly st one cluster hea among the noes. The ranom protocol will not consier the remaining energy in sensor noes, therefore the energy constraint cluster hea stion protocol which consiers energy level will perform better than ranom protocols in terms of FND time ue to the remains of energy for each sensor. However, energy constraint protocols will not guarantee minimum energy consume per roun.. Fuzzy Logic There are pros an cons in using either energy or istance as the factor to t the cluster hea. Using fuzzy logic to combine an consier both factors will make the cluster hea stion process simpler. The optimum results from both factors can be obtaine base on the fuzzy rule base system. Fuzzy inference technique via Mamani metho is use ue to its simple structure. There are four important steps in the Mamani metho: Fuzzification of the input variables, transforms the system inputs which are crisp values into fuzzy sets. Rule evaluation, taking the fuzzifie inputs an evaluate them to the anteceents of the fuzzy rules. Aggregate conclusions, the process of unification of the outputs of all rules. Defuzzication, transforms the fuzzy set obtaine by the inference engine into a single crisp value. In the efuzzification moule, the fuzzy set obtaine from the inference engine nees to be transforme to the single crisp value. COA (centre of area) is use to calculate the crisp value from the aggregate output membership function using the (5) where μ A (x ) is the membership function of the fuzzy sets: fitness = ( x. u A ( x) x xx) (5) 395

5 11 I Colloquium on Humanities, Science an ngineering Research (CHUSR 11), Dec , Penang The following membership functions are set for the simulation stuy. The linguistic variables use to represent the noe s energy an istance are ivie into three levels; high, meium, an low for energy as in Fig. 5; then far, mile an near for istance as in Fig. 6. The outputs to represent the noe cluster hea stion is name as fitness which is ivie into nine levels: very small, small, rather small, meium small, meium, meium large, rather large, large an very large as in Fig. 7. Table II shows the fuzzy rule base, consists of 3 = 9 rules. Figure 5. Membership functions of istance. Figure 6. Membership functions of energy. Table II shows the rule base set in the fuzzy inference system (FIS), the rule set as if the energy is low an the istance is far then the fitness to be cluster hea is very small. On the other han, if the energy is high an the istance is mile, the fitness becomes very large. Referring to the istance rule base, mile is set to higher priority than near because the cluster hea in the mile zone will perform better than the near zone which has been explaine in section III. In fuzzy base centralize algorithm, base station will calculate the fitness for every noe to become cluster hea via FIS where the base station will st the cluster heas accoring to p total number of noes. The ata transmission phase is similar to the LACH steay state phase. IV. RSULTS AND DISCUSSIONS The propose metho nees extra information to t the cluster heas. Therefore, an aitional 4 bits of energy an istance information are ae to the ata transmitte from every member noe to the cluster hea. On the other han, 48 bits is ae to the transmission from the cluster hea to the base station as extra energy an istance information for each cluster members an the cluster hea. The simulation was carrie out in MATLAB with the parameters show in Table III. In Fig. 8, the simulation results show that the propose fuzzy logic metho performs better than the LACH approach before FND in both 5% an 7% chosen as cluster hea cases. However, the propose metho loses to LACH after 6% of total alive noes. In LACH, cluster hea is te base on a ranom probability moel causes the cluster hea to exhaust the energy rapily compare to the propose metho which consiers critical factors in ting the cluster hea. TABL III. SIMULATIONS PARAMTRS Parameter Value Topology size 3 1m Base Station Position 15m, 15m Simulation Roun Number of Noe, n % 7% Figure 7. Membership functions of fitness. TABL II. FUZZY RUL BAS No nergy Distance Fitness 1 low far very small low near small 3 low mile rather small 4 meium far meium small 5 meium near meium 6 meium mile meium large 7 high far rather large 8 high near large 9 high mile very large Number of Alive Noes propose LACH Rouns Figure 8. Network lifetime. 396

6 11 I Colloquium on Humanities, Science an ngineering Research (CHUSR 11), Dec , Penang Numbers of Alive Noes propose LACH 5% % Data Receive by Base Station (bits) Figure 9. Network lifetime. x 1 6 This is because the propose metho uses energy level as one of the factor to be consiere in ting the cluster hea. All the sensor noes istribute the energy consumption evenly accoring to the FIS. After FND, the remaining noes ie faster compare to the LACH since the energy form the remaining noes are almost equally less. Fig. 9 shows the ata receive by the base station for the LACH an fuzzy approaches in the case of 5% an 7% of chosen cluster hea. Receive ata by the base station can be further increase by increasing the number of cluster hea from 5% to 7%. It can be observe that the ata receive by fuzzy approach are more than the LACH in both cases. Increasing the number of cluster hea will not much affect the network lifetime as shown in Fig. 8. However, it can significantly improve the throughput to the base station. xtra information for the propose metho is not consiere into the throughput for fair comparison. Proper number of cluster heas shoul be carefully ecie to achieve optimum in ata collection. Using too many cluster heas will lea to aitional overhea an cause the sensor noes sacrifice faster without proviing significant information. V. CONCLUSION As a conclusion, the propose metho achieve its objective to prolong the first noe ie (FND) time an also improve the throughput to the base station before FND. It shows improvement compare to the LACH protocol because the base station have partial global knowlege of the networks an the use of fuzzy logic to t the suitable cluster hea base on two factors which are energy an istance. The propose metho can be further improve by aopting some intelligent algorithm to moify the shape of each fuzzy set to achieve optimum result in specific cases. With the global knowlege of the network, multi-hop routing can be implemente to reuce the energy consumption in the cluster hea. ACKNOWLDGMNT The authors woul like to acknowlege the financial assistance of the Ministry of Higher ucation of Malaysia (MoH) uner Funamental Research Grant Schemes (FRGS), no. FRG14-TK-1/7. RFRNCS [1] J.N. Al-Karak, an A.. Kamal, Routing techniques in wireless sensor network: A survey, I Wireless Communication, vol. 11, issue 6, pp. 7-34, 4. [] I. F. Akyiliz,W. Su, Y. Sankarasubramaniam, an. Cayirci, Wireless sensor networks: a survey, Computer Networks, vol. 38, pp ,. [3] W. Heinzelman, A. Chanrakasan, an H. Balakrishnan, nergyefficient communication protocol for wireless microsensor networks, Proceeings of the 33 r Annual Hawaii International Conference on System Sciences (HICSS), Maui, Hawaii, pp ,. [4] I. Mahgoub, an M. Ilyas, Sensor Network Protocols, CRC Press, Taylor & Francis Group, pp , 6. [5] J. Zheng, an A. Jamalipour, Wireless Sensor Networks A Networking Perspective. Wiley, USA, pp , 9. [6] J.M. Kim, S. Park, Y. Han, an T. Chung, CHF: Cluster hea tion mechanism using fuzzy logic in wireless sensor networks, Proceeings of 1 th Internatinal Conference on Avance Commmunication Technology (ICACT), Phoneix Park, Korea, pp , 8. [7] W. Heinzelman, A. Chanraksan, an H. Balakrishnan, An applicationspecific protocol architecture for wireless sensor networks, I Transcation on Wireless Communications, vol. 1, no. 4, pp ,. [8] I. Gupta, D. Rioran, an S. Salli, Cluster-hea tion using fuzzy logic for wireless sensor networks, Proceeings of Communication Networks an Services Research Conference (CNSR), Halifax, Nova Scotia, Canaa, pp. 55-6, 5. [9] C.S. Chin, P. Neelakantan, H.P. Yoong, an K.T.K. Teo, Optimisation of fuzzy base maximum power point tracking in PV system for rapily changing solar irraiance, Global Journal of Technology an Optimisation (GJTO), vol., no., pp , 11. [1] C.S. Chin, P. Neelakantan, H.P. Yoong, S.S. Yang, an K.T.K. Teo, Maximum power point tracking for PV array uner partially shae conitions. Proceeings of 3r International Conference on Computational Intelligence, Communication Systems an Networks (CICSyN),Bali, Inonesia, pp