Balanced K Means Based Clustering Algorithm for Energy Efficient in Wireless Sensor Networks

Transcription

1 Balanced K Means Based Clustering Algrithm fr nergy fficient in Wireless Sensr Netwrks Heena Chawla 1, Parveen Verma 2 Department f lectrnics & Cmmunicatin ngineering Maharishi Ved Vyas ngineering Cllege, Jagadhri, India Abstract-In wireless sensr netwrks, technlgy has prvided the availability f small and lwcst sensr ndes with capability f sensing varius types f physical and envirnmental cnditins. Clustering is an imprtant and widely used methd, which can prlng the lifecycle f wireless sensr netwrk. xisting clustering algrithm des nt assure the maximum prlngatin f the verall netwrk lifetime. T address the abve issues, this paper presents a balanced k-means prtcl n the basis f space equivalent distributin and energy equivalent distributin. The simulatin result shws that balanced k-means is mre effective in prlnging the netwrk life time cmpared t LACH[1], LACH-C[2] and LACH K-Means[3]. Keywrds-LACH, cluster head, Balanced K- Means I. INTRODUCTION In wireless sensr netwrks, the tiny, lw cst and lw pwer sensrs are able t cmmunicate within a shrt range and wrk tgether t frm a sensr netwrk fr gathering data frm a field. These sensrs have data prcessing and cmmunicatin capabilities. They als have enabled us t mnitr and cllect data in any envirnment. They sense the cnditin in which they are surrunded and transfrm their data t trnic signals. The trnic signals are transmitted ver radi waves t the base statin (BS). Wireless sensrs netwrks ruting prtcls can be divided int flat and hierarchical ruting prtcls accrding t the tplgy f netwrk. LACH (Lw nergy Adaptive Clustering Hierarchy) is the simplest hierarchical prtcl which pssess clustering apprach. LACH is a prtcl that tends t reduce energy cnsumptin in a wireless sensr netwrk. Clustering is the main factr respnsible fr energy cnservatin in LACH algrithm. Main bjective f clustering are equal distributin f energy and equal distributin f ndes in space s that less energy is cnsumed and early death f ndes can be delayed[4]. In LACH bth f these bjectives cannt be achieved. Accrding t the deficiency f the LACH,Max- nergy LACH and LACH K-Means are almst imprved based n the LACH agreement. Amng them LACH K-Means algrithm imprves the perfrmance f netwrk by the ptimal number f cluster head and the mechanism f cluster head stin, making nde energy cnsumptin reduced and extending the netwrk life cycle. This simulatin result shws that LACH K-Means have bvius advantage ver LACH and MAX-nergy LACH agreement in verall perfrmance but LACH K-means has a prblem as it can have unbalanced number f ndes in cluster. On the basis f clustering algrithm LACH, this paper put frward a clustering ruting algrithm Balanced K- Means based n re-ting cluster head methd fr balancing lcal clusters and uses the infrmatin which the cluster head have. This is called balanced energy based clustering. II. LACH PROTOCOL Lw nergy Adaptive Clustering Hierarchy (LACH) first prpsed by Wendi.B.Heinzelman f MIT is a clustering based prtcl that utilizes randmized rtatin f lcal cluster t evenly distribute the energy lad amng the sensrs in the netwrk[1]. The LACH uses lcalized crdinatin t enable scalability and rbustness fr dynamic netwrks. LACH incrprates data fusin int the ruting prtcl t reduce the amunt f infrmatin that must be transmitted t base statin. It rearrange the netwrk clustering dynamically and peridically, making it difficult fr us t rely n lng lasting nde t nde trust relatinship t make the prtcl secure. LACH assumes every nde can directly reach a base statin by transmitting with sufficiently high pwer. We have chsen fur algrithms fr clustering purpse in WSN ruting in ur experimentatins. These are LACH, LACH-C, K-Means LACH and balanced K-Means LACH. These are explained as belw: 1540

2 A. LACH LACH cmprises f 2 phases:- Set-Up Phase and Steady State Phase (i)set-up Phase At the beginning every single nde cmputes a randm number between 0 and 1, and then cmputes a threshld frmula T(n). If the randm number is less than the cmputed threshld the nde becmes a cluster head. After its stin each cluster head will bradcast an advertisement message t the rest f the ndes by using CSMA MAC prtcl. After that each nde sts a cluster head based n the Received Signal Strength Indicatin (RSSI) f the advertisement. ach nde uses CSMA MAC prtcl t transmit its stin. Clusters are frmed, each cluster head creates a TDMA schedule accrding t number f ndes in the cluster. ach nde during its allcated transmissin time. ach sensr nde senses and transmits data t its cluster head based n the TDMA schedule. The cluster heads receive all the data aggregate it and sends it t the base statin. After transmissin that netwrk starts next rund and again execute the setup and steady state phase B. LACH-C When nde is sted as cluster head, the cluster bradcast the signal (Advertisement message,adv) t the ther ndes. ach nde receive ADV frm different cluster heads, accrding t the strength f the signal, it chses t jin prper cluster, and rely t the crrespnding cluster head. The cluster head receives the jining infrmatin, prducing a TDMA schedule.the cluster head receives data frm members f cluster, and transmits t base statin after merging int a packet. After data transmissin, the netwrk chses the cluster head again. Start Cmpute K-means fr each cluster center determinatin Fr each nde, chse the prximate cluster. (i) Set Up Phase During the set-up phase, each nde sends infrmatin abut its current lcatin using a GPS receiver and energy level t the base statin. Alng with determining gd cluster, the base statin needs t ensure that the energy nde is evenly distributed amng all the ndes. Base statin cmputes the average nde energy, and whichever nde has energy belw this energy cannt be the cluster head fr the current rund. Using the remaining nde as pssible cluster heads, the base statin finds ptimal cluster itself. Once the cluster heads and assciated clusters are fund, the base statin bradcasts a message that cntains the cluster head ID fr each nde. If a nde cluster head ID matches its wn ID, the nde determines its TDMA slt fr data transmissin and ges t sleep until it is time t transmit data. C. LACH K-MANS (i) Set-Up Phase In set up phase, it first chses k ndes as K clusters. Then assign ther ndes t these clusters. It then calculates the every nde s cst. Nde having the smallest cst is sted as cluster head. This cycle desn t stp until K-cluster heads fix dwn. If cluster is full with permissible number f ndes, chse next prximate cluster Steady Phase Stp Figure 1. Flw graph fr Balanced K-Means Algrithm 1541

3 D. PROPOSD PROTOCOL(BALANCD K- MANS) (i) Set Up Phase During set up phase, the sensr ndes are divided equally int clusters which are sted by K-Means. Depending n the size f the netwrk and number f sensr ndes, the sensr ndes can be equally divided int n numbers f clusters while sted cluster by each nde if in a cluster number f ndes exceed than the maximum permissible nde than they chse anther cluster. Prpsed prtcl s steady state phase is identical t that f LACH-KMANS. Table 1 Cmparisns f LACH, LACH-C, LACH-KMANS and Prpsed Prtcl Name f the Prtcl LACH LACH-C LACH- KMANS Mdified LACH III. Organizatin f clusters First Cluster Heads are chsen, then cluster are frmed. Clusters and Cluster Heads are chsen by the Base Statin First, Clusters are made, then Cluster Heads are chsen. First, Cluster Heads are chsen, then cluster are frmed. Randm and Centralised and fixed Randm ven XPRIMNTATION STUP AND SIMULATION (i) xperimental Setup T simulate LACH, we have used randm 200 ndes netwrk fr ur simulatin with parameters used in[3]. We placed the base statin at a far distance frm all the ndes. Fr a plt, ur base statin is lcated at(50,200) s that the base statin is at least 100m frm the clsest sensr nde. All the ndes in the netwrk have limited energy and hmgenus. The initial energy f all the ndes are same. (ii) nergy Mdel fr LACH This paper adpts the same energy cnsumptin mdel f[5].in the prcess f transmissin k thrugh lng distance d, the energy cnsumptin f the sending end is: TX k, d k k d am p d 2, 4 d d, d d (1) The receiving energy cnsumptin f receiving k data is: Rx In(1), k k (2) is the pwer cnsumptin f sending and receiving, decided by the circuit itself.if the transmissin distance is less than threshld d, the pwer amplifier f free space mdel is adpted. Cnversely, when the transmissin distance is equal r greater than the value, the pwer amplifier f multi path attenuatin mdel is adpted. d (3) am p amp In[5], nde energy cnsumptin f cmputing is far less than energy cnsumptin f data transmissin, s in this paper we ignre nde calculatin energy cnsumptin. (iii) Parameter Stin Fr Simulatin Fllwing table presents the list f parameter required fr the simulatin f LACH with encryptin strategies. Table 2: Simulatin Parameter Netwrk Architecture Parameter Netwrk size Value N. f ndes x50 m Base Statin (BS) 25, 200 Initial energy f nde d 0 nergy fr transferring f each bit (TX) nergy fr receiving f each bit (TX) Data Aggregatin nergy (DA) Packet length (number f bits in packet sent frm CH t BS) ctrpacketlength (number f bits in packet sent frm nde t CH) Pwer amplifier f Free space() Pwer amplifier f Multi Path (mp) 0.5 J 87.7 m 50e-9 J 50e-9 J 5e-9 J 6400 Bit 200 Bit 1.0e-11 J 1.3e-15 J 1542

4 IV SIMULATION RSULTS AND ANALYSIS Fllwing table shws the result btained frm the experimentatin dne as per the setup. The perfrmance f the prpsed prtcl is evaluated and cmpared with existing LACH,MAX-LACH and LACH-K Means in terms f number f runds, packet sent t the base statin, energy and number f ndes. statistics in the fllwing graph 1 t 3.LACH and MAX_LACH and marginally better than k-means as cmpared t fr each and every criterin except the number f packet sent t base statin. Table 3: Simulatin Results Name f the Prtcl Mdified LACH- KMANS LACH- KMANS Rund Packet nergy Dead ndes t BS Nde Rund Start: 20 1 st 6 nd: 50 th th Start: 20 1 st 6 nd: 50 th Graph 1 Cmparisn f fur algrithms fr n f packets sent t BS Vs N. f Runds. 80 th 534 M A X _ Start: 20 1 st 248 L A C H nd: 50 th LACH Start: 20 1 st 144 nd: th th 418 Graph 2 Cmparisn f fur algrithms fr Residual nergy Vs N. f Runds In abve table 3 and graphs, it is clearly shwn that mdified K-means perfrm far better as cmpared t randm LACH and MAX_LACH and marginally better than K-means as cmpared t fr each and every criterin except number f packets sent t BS. The balanced K-means algrithm perfrms nearly tw times better than randm LACH and MAX_LACH and as cmpared t K-means LACH perfrmance is nearly 20%. If we cnsider a netwrk, dead if 50% ndes are dead then LACH and MAX_LACH are perfrming better than k-means based LACH but they survive lnger time as a whle. Randm LACH has perfrmed wrst in energy and dead ndes criterin. If we cnsider 80% nde criterin fr netwrk life then again k-means based LACH perfrm better than ther tw algrithms. If we cmpare the number f dead ndes as per ur simulatin results MAX_LACH and perfrms better, but there ndes nce start dying accelerates netwrk decay very fast We have shwn this Graph.3 Cmparisn f fur algrithms fr n f dead ndes Vs N. f Runds The balanced k-means algrithm perfrms nearly tw times better than randm LACH and MAX_LACH and as cmpared t K-MANS perfrmance is nearly 20%. In these graphs,it is clearly shwn that mdified k-means perfrm fr 1543

5 better as cmpared t randm. If we cnsider n f packets sent t BS then the LACH is clear winner. Later has sent highest number f packets t BS. This may be due t uneven and large number f cluster frmatin. Large numbers f clusters are nt desired but a fair amunt f data is desired in which balanced k-means seems t perfrm better. V CONCLUSION In this paper, we put frward an imprved ruting algrithm Balanced K-Means. This algrithm imprve the perfrmance by making equidistributin f space in the ndes fr better clustering in k-means and balancing f cluster lads have increased the energy efficiency by increasing the numbers f runds.the simulatin results shw that Balanced k-means have bvius advantages ver LACH,MAX-LACH and LACH-K Means in verall perfrmance. VI RFRNCS [1] Stefans, A. N., Dinisis, K., Dimitris, D. V. and Christs D. (2013), nergy fficient Ruting in Wireless Sensr Netwrks Thrugh Balanced Clustering, Algrithm, 6, [2] Heinzelman, W. B., Chandrakasan, A. P. and Balakrishnan, H. (2002), An applicatinspecific prtcl architecture fr wireless micrs sensr netwrks. I Transactins n Wireless Cmmunicatins, Vl. 1, Issue 4, pp [3] Park, G. Y., Kim, H., Jeng, H. W. and Yun, H.Y. (2013), A Nvel Cluster Head Stin Methd based n K-Means Algrithm fr nergy fficient Wireless Sensr Netwrk, 27th Internatinal Cnference n Advanced Infrmatin Netwrking and Applicatins Wrkshps, pp [4] Yunjie, J., Ming, L., Sng, Z. and Pengta, D. (2012), A clustering ruting algrithm based n energy and distance in WSN, Internatinal Cnference n Cmputer Distributed Cntrl and Intelligent nvirnmental Mnitring, pp [5] Natarajan, M., Arthi, R. and Murugan, K. (2013), nergy Aware Optimal Cluster Head Stin in Wireless Sensr Netwrks, 4th ICCCNT, pp. [6] Ahmed, A., Qazi, S. (2013), Cluster Head Stin Algrithm fr Mbile Wireless Sensr Netwrks, Internatinal Cnference n Open Surce Systems and Technlgies, pp [7] Anitha, R. U., Kamalakkannan, P. (2013), nergy fficient Cluster Head Stin Algrithm in Mbile Wireless Sensr Netwrks, Internatinal Cnference n Cmputer Cmmunicatin and Infrmatics, pp [8] Ran, G.,Zhang, H. and Gng, S. (2010), Imprving n LACH Prtcl f Wireless Sensr Netwrks Using Fuzzy Lgic, Jurnal f Infrmatin & Cmputatinal Science, Vl. 7, Issue 3, pp [9] Ahlawat, A., Malik, V. (2013), An extended vice-cluster stin apprach t imprve v leach prtcl in wsn, Third Internatinal Cnference n Advanced Cmputing & Cmmunicatin Technlgies, pp [10] Firuzbakht, A. R., Buyer, A. (2013), An ptimal Algrithm based n gridding and clustering fr decrease energy cnsumptin in WSN, MIPRO, pp [11] Rahmanian, A., Omranpur, H., Akbari, M. and Raahemifar, K. (2011), Nvel genetic algrithm in Max_LACHruting prtcl fr sensr netwrks, I CCC, pp [12] Shi, S., Liu, X. and Gu X. (2012), An nergy- fficiency Optimized MAX_LACH fr Wireless Sensr Netwrks, 7th Internatinal ICST Cnference n Cmmunicatins and Netwrking, pp [13] Raj,. D. (2012), An fficient Cluster Head Stin Algrithm fr Wireless Sensr Netwrks drleach, IOSR Jurnal f Cmputer ngineering, Vl. 2, Issue 2, pp