Inter national J our nal of Electrical and Computer Engineering (IJECE) V ol. 6, No. 5, October 2016, pp. 2403 2414 ISSN: 2088-8708 2403       I ns t it u t e  o f  A d v a nce d  Eng ine e r i ng  a nd  S cie nce   w     w     w       i                       l       c       m     An Unequal Cluster -based Routing Pr otocol Based on Data Contr olling f or W ir eless Sensor Netw ork Slaheddine Chelbi * , Majed Abdouli *** , Mourad Kaddes *** , Claude Duv allet ** , and Rafik Bouaziz * * MIRA CL Laboratory , Uni v ersity of Sf ax, T unisia ** LITIS, Uni v ersity of Le Ha vre, France *** FCIT , Uni v ersity of Jeddah, Saudi Arabia Article Inf o Article history: Recei v ed Apr 3, 2016 Re vised Aug 4, 2016 Accepted Aug 21, 2016 K eyw ord: W ireless Sensor Netw orks Unequal cluster size Sensiti v e data controlling Routing Protocol Ener gy Sa ving ABSTRA CT W ireless Sensor Netw orks (WSN) dif fer from traditional wireless communication netw orks in se v eral characteristics. One of these charac teristics is po wer a w arness , due to the f act that the batteries of sensor nodes ha v e a res tricted lifeti me and are dif ficult to be replaced. There- fore, all designed protocols must minimize the ener gy consumption and therefore preserv e the longe vity of the netw ork. In this paper , we propose (i) to f a irly balance the load among nodes. F or this, we generate an unequal clusters size where the cluster heads (CH) election is based on ener gy a v ail ability , (ii) to reduce the ener gy consumption due to the transmission by using multiple metrics in the CH jointure process and taking into account the link cost, residual ener gy and number of cluster mem bers to construct the routing tree and (iii) to min- imize the number of transmissions by a v oiding the unnecessary updates using sensiti v e data controller . Simulation results sho w that our Adv anced Ener gy-Ef ficient Unequal Clustering (AEEUC) mechanism impro v es the f airness ener gy consumption among all sensor nodes and achie v es an ob vious impro v ement on the netw ork lifetime. Copyright c 2016 Institute of Advanced Engineering and Science . All rights r eserved. Corresponding A uthor: Slaheddine Chelbi MIRA CL Laboratory Uni v ersity of Sf ax, T unisia slaheddine.chelbi@gmail.com 1. INTR ODUCTION In the last decade, with the de v elopment and adv ancement of sensor technologies and their relati v e cheapness, WSN ha v e been widely deplo yed for both ci vil and military applications, e.g., harbor container monitoring system, marine monitoring for earthquak e and tsunami, and military surv eillance in battlefields [1][2]. A WSN consists of hundreds to thousands of sensor nodes, which ha v e the ability to communicate among themselv es using radio antenna. Since sensor nodes are battery po wered and may be applied in dangerous or inaccessible en vironments, it is dif ficult and e v en impossible to replace or rechar ge the po wer supplies [3] [4]. W e need ener gy-ef ficient mechanisms to reduce ener gy consumption of nodes and maximize netw ork lifeti me. Balancing the ener gy consumption in the netw ork is an important issue for prolonging the netw ork lifetime. Therefore, ef ficient routing techniques are highly required to balance the ener gy consumption among the netw ork nodes and maximize the netw ork lifetime [5][6]. Dif ferent routing protocols for WSN proposed in literature are cate gorized based on their computational comple xity , netw ork structure, ener gy ef ficienc y and path establishment [7][8] [9]. One of the principal clas ses is hierarchical routing protocols. In this class, the main idea is that e v ery sensor node within a WSN is grouped along with some other of its neighboring nodes in order to constitute a specific clus ter . Clustering pro vides an ef fecti v e method for prolonging the lifetime of a WSN. Data, collected by the sensor nodes belonging to a cluster , are not directly transmitted to the Base Station (BS). Instead, a node of the cluster , called CH, collects these data and forw ards them to the BS after possibly ha ving performed appropriate data aggre g ation. In this w ay , the number of transmitted messages to the BS is reduced and considerable po wer conserv ation is achie v ed. Ho we v er , routing protocols rarely consider the hot spot problem in multi-hop sensor netw orks. Indeed, when CH cooperate with each other to forw ard their data to the BS, the CH closer to the BS are b urdened with hea vier relay traf fic and tend to die much f aster , lea ving the areas of the net w ork unco v ered and causing netw ork partitions. T o J ournal Homepage: http://iaesjournal.com/online/inde x.php/IJECE       I ns t it u t e  o f  A d v a nce d  Eng ine e r i ng  a nd  S cie nce   w     w     w       i                       l       c       m     Evaluation Warning : The document was created with Spire.PDF for Python.
2404 ISSN: 2088-8708 mitig ate the hot spot problem, an Unequal Cluster -based R ou t ing (UCR) protocol is proposed in [10]. As sho wn in Figure 1, nodes are grouped into unequal sizes clusters: Clusters closer to the BS ha v e smaller sizes than those f arther a w ay from it. Thus CH closer to the BS can preserv e some ener gy for the purpose of inter -cluster data forw arding. cluster head cluster member base station Figure 1. Unequal Cluster size with the UCR mechanism [10] In this paper , we propose a ne w mechanism called AEEUC in order to enhance the WSN lifetime. AEEUC, as well as UCR [10], use an unequal clustering and multi-hop routing scheme to impro v e the netw ork lifetime. Ho we v er , the proposed mechanism to choose CH and inter -cluster communication in AEEUC is dif ferent from those of UCR. Our protocol aims to balance the amount of the ener gy consumption among CH within the netw ork. Furthermore, we propose to reduce t he consumed ener gy by minimizing the unnecessary data transmissions and hence prolong the lifetime of the system. T o achie v e this goal, we propose to a v oid the transmission of ne w v alue of data if it does not de viate from the current v alue more than Maximum Data Error (MDE). W e remind that the MDE indicates the maximum de viation tolerated between the current v alue and the ne w one. The rest of this paper is structured as follo ws: section 2 presents a brief related w ork. Section 3 outlines the design of our proposed routing protocols AEEUC. Section 4 presents simula tions and results. Finally , secti on 5 dra ws the conclusion and some future w ork. 2. RELA TED W ORK In the fe w recent years, a WSN ha v e g ained the attention of researchers in man y challenging aspects. The most important challenge in these netw orks is ener gy conserv ation. P articulary , t w o major issues are e xplored to optimize ener gy consumption: routing techniques and data aggre g ation. The routing techniques are classified into three cate gories depending on the netw ork structure: flat, hierar - chical, and location-based routing [11]. Hierarchical clustering techniques can help to reduce useful ener gy consumption. Hierarchical or cluster -based routing are well-kno wn techniques with special adv antages related to scalability and ef ficient communication [2] [12] [13]. Data aggre g ation in WSN is a data transfer technique by which se v eral pack ets from sensor nodes are combined into one [14] [15]. This technique is essential because the reduction of data pack ets reduces ener gy consumption, increases netw ork lifetime, and therefore impro v es successful data transmission ratio. LEA CH (Lo w Ener gy Adapti v e Clustering Hierarch y) is one of the first hierarchical routing protocol pro- posed for WSN [3]. It is also the first clustering protocol that w as proposed for reducing po wer consumpti on. LEA CH randomly selects a fe w sensor nodes as CH and rotates this role to e v enly distrib ute the ener gy load among the sensors in the netw ork. The idea is to form clusters of the sensor nodes based on the recei v ed signal strength and use local CH as routers to the BS. The operations of LEA CH are done into tw o phases, the setup phase and the steady state phase . In set up phase, the clust ers are or g anized and CH are selected. CH change randomly o v er time in order to balance the ener gy dissipation of nodes. In the steady state phase, the current data is transferred to the BS. Although LEA CH is able to increase the netw ork lifetime, it suf fers from a fe w limits: (1) routing in LEA CH protocol is based on the assumption that each node can transmit directly to both its CH and the BS. Ho we v er , such a single hop routing is impractical in netw orks ha ving their sensors deplo yed o v er wide re gi ons. It is not al w ays a realistic assumption for single-hop inter -cluster routing with long communication range. Besides, long-range commu- nications directly from CH to the BS can lead to high ener gy consumption. (2) Despite the f act that CH rotati on is IJECE V ol. 6, No. 5, October 2016: 2403 2414 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 2405 performed at each round to achie v e load balancing, LEA CH cannot ensure r eal load balancing because CH are elected in terms of probabilities without ener gy considerati ons. (3) LEA CH randomly selects a list of CH and it has not a con- straint on their distrib ution. Hence, probably all CH will be concentrated in the same area. Therefore, some nodes will not ha v e CH in their neighborhoods. T o o v ercome these dra wbacks, man y e xtension of LEA CH are proposed in liter - ature [16]. In [17], the authors propose a central control algorithms named C-LEA CH. The latter propose to construct the clusters in such a w ay that CH are scattered throughout the netw ork. In [18], the authors propose MODLEA CH protocol. In this protocol, the CH is changed until and unless it has more ener gy than the certain required threshold, unlik e LEA CH where the CH is changed after e v ery round. Furthermore, MODLEA CH cate gorised communication into dif ferent cate gories in order to optimize ener gy spend for communication. In [19], a multi-hop protocol entitled MH-LEA CH is proposed. This protocol increases netw ork life time by using neighbour nodes for data transmission which results in lesser ener gy consumption. Hybrid Ener gy Ef ficient Distrib uted (HEED) is a distrib uted clustering algorithm in which tw o parameters are used to determine the eligibility of a node to become a CH [20]. Since prolonging the netw ork lifetime is the main goal, residual ener gy is used as the first parameter , which allo ws nodes with higher residual ener gy to become CH, thus balancing the o v erall ener gy of the net w ork. The second f actor intra communication cost, which can be cluster density , allo ws a node to join a CH with the least number of nodes so as to r educe the load of the int ra-cluster traf fic on the CH. Ho we v er , HEED does not mak e an y assumption about the netw ork such as density or size. HEED introduces e xtra communication o v erhead because it needs to e xchange a lar ge number of messages in order to compute the communication cost with its neighbors. In [21], Y u et al. proposed a cluster -based routi ng protocol for WSN with non-uniform node distrib ution (EADC), whose cores are an ener gy-a w are clustering algorithm and a cluster -based routing algorithm. EADC is a competition based algorithm, where CH is selected on the basis of the ratio between its residual ener gy to the a v erage residual ener gy of its neighbors. T o form clusters, each node chooses t he nearest CH without tak en into account its residual ener gy . Each CH chooses a ne xt hop CH with higher residual ener gy and fe wer cluster members as its ne xt hop. In BCEE [22], CH is selected by using K-means algorithm. T o form clusters, BCEE does not require position of each node b ut uses the idea of recei v e signal strength indicator (RSSI). T o route the data to the sink, the techniques of ant colon y optimization is used to establish an optimal multi-hop route from CH to the sink using rational and hop-selecting technique. Ho we v er , BCEE introduces e xtra communication o v erhead and delay of data transmission to the sink node. In [23], the authors ha v e used the adv antages of tw o approaches i.e. fuzzy c-means (FCM) clustering and neural netw ork to mak e an ener gy ef ficient netw ork by prolonging the lifetime of netw ork. The cluster formation is done using FCM to form equally sized clusters in netw ork and the decision of choosing CH is done using neural netw ork ha ving input distance from BS, heterogeneity and ener gy of the node. The main focuses of these algorithms are to re d uc e and balance the ener gy consumption of nodes by using clustering algorithms. In f act, clustering (1) enables data aggre g ation at CH, hence reduces the number of unnecessary data transmissions, and sa v es ener gy of the sensor nodes, (2) simplify routing management because only CH need to maintain the local route setup of other CH and thus require small routing information, (3) conserv es communication bandwidth [24] . Ho we v er , CH bear some e xtra w ork load contrib uted by their member sensor nodes. In the other hand, when the multi-hop inter -cluster communication model i s adopted, the CH closer to the BS are b urdened with hea vier relay traf fic and will depletes ener gy much f aster , lea ving areas of the netw ork unco v ered. T o mitig ate the hot spot problem, Chen etal . [10] ha v e proposed UCR protocol. In UCR, CH is selected based on local information i:e: , the residual ener gy of neighboring nodes. The node’ s competition range decrea ses as its distance to the BS decreases. The result is that clusters closer to t he BS are e xpected to ha v e smaller cluster sizes, thus the CH will consume lo wer ener gy during the intra-cluster data processing, and can preserv e more ener gy for the inter -cluster relay traf fic. The UCR protocol produces a distrib ution of CH through an election process using a competition radius. The remaining sensor nodes recei v e the broadcast from one or more CH and mak e their association decision based on minimum communication cost. The inter -cluster multi-hop routing protocol considers the tradeof f between the ener gy cost of relay links and the ener gy of relay nodes. UCR e xtends LEA CH and HEED by choosing CH with more residual ener gy . In UCR protocol, clusters closer to the BS are e xpected to ha v e smaller cluster sizes. Thus the y will consume lo wer ener gy during the intra-cluster data processing, and can preserv e more ener gy for the inter -cluster relay traf fic. But, in netw orks with non-uniform node distrib ution, this mechanism is not al w ays ef fecti v e. Consequently , the higher ener gy consumption still e xists among CH closer to BS due to the non-uniform node distrib ution. Moreo v er , in UCR se v eral tentati v e CH are random ly selected to compete for final CH. Hence, the participation of some tentati v e CH An Unequal Cluster -based Routing Pr otocol Based on Data Contr olling for WSN (S. Chelbi) Evaluation Warning : The document was created with Spire.PDF for Python.
2406 ISSN: 2088-8708 may be at the e xpense of other nodes ha ving higher residual ener gy . 3. THE AEEUC MECHANISM In this section, we present the rele v ant details of Adv anced Ener gy Ef ficient Unequal Clustering mechanism in WSN. W e suppose that the WSN is composed by a BS and a set of homogeneous sensor nodes which are randomly distrib uted o v er a bounded area of interest. The sensors nodes and BS are stationary after deplo yment. The BS is a node with high capabilities and unlimited po wer b ut the sensor nodes are ener gy constrained. The main idea of AEEUC is based on the creation of ener gy-ef ficient clusters for a gi v en number of transmissions. The task of being a CH is rotated among sensors in each round (a set of transmissions) to distrib ute the ener gy con- sumption across the netw ork. AEEUC is a distrib uted CH competiti v e algorithm where nodes with higher ener gy will be elected as CH. T o mitig ate the hot spot problem, we use in the AEEUC protocol unequal sizes clusters. Thus, clusters closer to the BS ha v e smaller cluster sizes, and, consequently , the y will consume less ener gy during the intra-cluster data processing, and can conserv e some more ener gy for the inter -cluster relay traf fic. By producing cluster with unequal sizes, UCR succeeds in making the ener gy consumption of CH balanced. Ho we v er , in some cases and due to non-uniform distrib ution of nodes, it may happen that CH with a small competition range will ha v e more members’ nodes. In this case, the y ha v e high intra-cluster ener gy consumption [21] . F or this, in our w ork, each CH will select t he neighbor CH with higher residual ener gy , minimum link cost and a smaller number of cluster members as the ne xt hop to balance the ener gy consumption among CH. T o a v oid the unnecessary updates and subsequently optimize the use of resources, in [25] data is allo wed to de viate, with a certain de gree, from their corresponding v alues in the e xternal en vironment. The y introduce the notion of data error , denoted DE, which gi v es an indication of ho w much the v alue of stored data de viates from the corresponding real-w orld v alue. This de viation has a threshold named MDE (Maximum Data Error) . The transmission of data is dis- carded if the de viation between the current data v alue and the stored v alue is less or equal to MDE (if D E M D E ). In the same w ay to reduce the number of transmission, we allo w data error in AEEUC. The whole p r ocess is di vided into three phases: CH competition phase; cluster formation phase and data transmission phase. 3.1. CH competition phase AEEUC is a distrib uted CH competiti v e algorithm, where the CH selection is primarily based on the residual ener gy of tentati v e CH and rotating CH periodically to distrib ute the ener gy consumption among nodes in each cluster . Only nodes which ha v e not been CH in N late rounds are eligible to be a tentati v e CH for the current round to ensure that only nodes with suf ficient ener gy are selected as CH. Ordinary nodes which ha v e not been CH in N late round become tentati v e CH with the same probability T which is a predefined threshold. It should be noted that T is used to reduce the message o v erhead on the dense netw ork. S tate ( s i ) = " T entativ e " if s i 2 G and < T " M ember " O ther w ise (1) Where is the probability of being selected as tentati v e CH and G is the group of the nodes which ha v e not been CH in N late rounds. Each tentati v e CH s i has a competition range R i . Dif ferent competition ranges are used t o produce clusters of unequal sizes. Only one final CH is allo wed in each competition range. If s i becomes a CH at the end of the competition, there will not be another CH s j in s i s competition range. The task of being a CH is rotated among sensors in each round to distrib ute the ener gy consumption across the netw ork. Similar to UCR, CH closer to the BS should support smaller cluster sizes. Thus, more clusters need to be produced closer to the BS. That i s to say , the tentati v e CH’ s competition range should decrease as its distance to the BS decreases. W e need to select a proper scope of competition ranges in the netw ork. R is the maximum competition range and the minimum competition range is set to (1 c ) R correspondingly , where c is a constant coef ficient between 0 and 1. Thus the tentati v e CH s i s competition range R i can be e xpressed as a l inear function of its distance to the BS (cf. F ormula 2): R i = (1 c d max d ( s i ; B S ) d max d min ) R (2) IJECE V ol. 6, No. 5, October 2016: 2403 2414 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 2407 Where d max and d min denote respecti v ely the maximum and minimum distance between sensor nodes in the netw ork and the BS and d ( s i ; B S ) denotes the distance between s i and the BS. Each tentati v e CH maintains a set S C H of its adjacent tentati v e CH. T entati v e CH s j is an adjacent node of s i if s j is in s i s competition diameter or s i is in s j s competition diameter . In the CH competition phase, the broadcast radius of e v ery control message is R . Each tentati v e CH broadcasts a C ompeteH eadM sg which contains it s competition radius and residual ener gy . After the construction of S C H has finished, each tentati v e CH checks its S C H and mak es a decision as to whether it can act as a CH. If the node s i has the lar gest residual ener gy among all tentati v e CH in its S C H , it is elected as CH and broadcasts a F inal H eadM sg message. Otherwise, i t gi v es up the competition. The follo wing pseudo-code gi v es the details of this phase. Algorithm 1 CH competition algorithm if s i 2 G then   R AN D (0 ; 1) else   1 end if if < T then beT entativ e = T r ue end if f or each tentati v e s i do C ompeteH eadM sg ( id ; R comp ; E r es ) end f or On recei ving a C O M P E T E H E AD M S G form node s j if d ( s i ; s j ) < s j :R comp or d ( s i ; s j ) < s i :R comp then Add s j to s i :S C H end if while S tate == T entativ e do if s i :E r es >s j :E r es ; 8 s j 2 s i :S C H then State = Head F inal H eadM sg ( id; E r es ) Exit end if On recei ving a F I N ALH E AD M S G form node s j if s j 2 s i :S C H then State = Member QuitE l ectionM sg ( id ) Exit end if On recei ving a QU I T E LE C T I O N M S G form node s j if s j 2 s i :S C H then Remo v e s j from s i :S C H end if end while After the CH has been selected, sleeping nodes no w w ak e up and each CH broadcasts a C H AD V M S G across the netw ork field which contains its id and its residual ener gy . 3.2. Cluster f ormation phase In UCR, each ordinary node chooses i ts closest CH. If node i w as to mak e a decision bas ed on a single parameter it could result in a bad choice o v er all. F or e xample, selecting the closest CH will lead to choosing a CH which is at a smaller residual ener gy , thus resulting in more CH load. Hence, when a node mak es a decision about associating with the CH, it is necessary that man y parameters should be considered. In the proposed technique, the first focus is optimizing the ener gy usage in cluster formation; i.e. t he decision process An Unequal Cluster -based Routing Pr otocol Based on Data Contr olling for WSN (S. Chelbi) Evaluation Warning : The document was created with Spire.PDF for Python.
2408 ISSN: 2088-8708 used by an ordinary sensor node to associate itself with a CH is based on minimum o v erall communication cost. F or the node j , the cost of joining the CH k is computed by equation (3): C ost k j = d ( j ; k ) d max + (1 ) E max E k R es E max (3) d max = max f d ( j ; k ) g ; K 2 S (4) E max = max f E R es g ; K 2 S (5) Where is the weighted f actor for the trade-of f between the distance to the CH and the residual ener gy of the CH, and S is a candidate CH set of the node j . Each ordinary node chooses its CH and then informs the CH by sending a J O I N C LU S T E R M S G which contains the id and residual ener gy of this node. The CH sets up a time di vision multiple access (TDMA) schedule and transmits it to the nodes in its cluster . Each CH collects the messages from its cluster members and sa v e them. 3.3. Data transmission The data transmission is di vided in tw o phases. The first one is intra-cluster where each non CH nodes send data to CH and the second is inter -cluster where the CH transmits aggre g ated data to BS. 3.3.1. Intra-cluster communication In our protocol, the task of being a CH is rotated among sensors in each round to balance the ener gy con- sumption across the netw ork. Each round is composed of a predefined number of iterations in which each cluster will k eep its structure. But, a ne w clustering will be triggered only when one CH is dead or the end of a round is reached. This mechanism reduces the o v erhead traf fic through reducing control messages. Thus, the ener gy will be sa v ed. At the be ginning of each round (first iteration), e v ery non-CH node w aits for their TMD A slot to transmit data. When the time slot arri v es, the node transmits the data to CH. From the second iteration, each node, during its allocated trans- mission time, sends to the CH quantitati v e data concerning the sens ed e v ents. Data are transmitted by a node to its CH only when this v alue changes by an amount equal to or greater than the MDE (if D E M D E ). Hence, only parts of nodes passing data v erification need to transmit data to CH. This reduces the transmission ener gy consumption. In the second phase, each CH recei v es the data from its cluster nodes. When, all the data are recei v ed, each CH aggre- g ate the data it has recei v ed along with its o wn data into a single composite message. When CH recei v es a pack et of an y type, it updates, in its table of cluster member , the residual ener gy field and the data v alue of the sender node to be a w are of the death of a node. When the time slot of one node arri v es and this node doesn’ t transmit the data to CH, this later will use the v alue of data stored and the de viation is allo wed. 3.3.2. Inter -cluster communication After recei ving and aggre g ating data from dif ferent cluster members, the CH be gin, in this phase, the routing of data to the ne xt hop nodes. As the ener gy dissipation is directly proportional to transmission distance and due to their ener gy constraint, WSN usually ha v e a limited transmission range making multi-hop data routing to w ard the BS more ener gy-ef ficient than one-hop transmissions. The CH sent a ll collected data of the cluster members to BS by multi-hop manner , which sa v es the ener gy consumption of CH when BS is v ery f ar in WSN. Each CH s i selects the ne xt CH s j . If a s i s distance to the BS is smaller than TD-MAX , it transmits its data to the BS directly; otherwise it should find a relay node which can forw ard its data to the BS. Before selecting the ne xt hop node, each CH broadcasts a beacon message across the netw ork at a fix ed po wer which consists of its node id , residual ener gy , number of cluster member and distance to the BS. The multi-hop forw arding algorithm considers nodes on the CH backbone in the forw ard direction (i.e., closer to the BS) only . The neighboring node set R C H of CH s i is defined by equation 6 where x is the minimum inte ger that lets s i :R C H contain at least one item. s i :R C H = f s j =d ( s i ; s j ) xR i ; d ( s j ; B S ) < d ( s i ; B S ) g (6) IJECE V ol. 6, No. 5, October 2016: 2403 2414 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 2409 In UCR, in order to reduce inef ficiencies of ener gy consumption, a tradeof f is made between the tw o criteria of residual ener gy and link cost E r el ay . E r el ay ( s i ; s j ) = d 2 ( s i ; s j ) + d 2 ( s j ; B S ) (7) s i first chooses k eligible neighbor nodes from s i :R C H , denoted as the set S el ig ibl e : s i :S el ig ibl e = f s j =s j 2 s i :R C H ; E r el ay is the k smal l est g (8) In this mechanism, s i chooses as its relay node the neighbor in s i :S el ig ibl e that has the biggest residual ener gy . This method has some dra wbacks: In UCR, it is goi ng to choose k nodes with the smallest E r el ay ( k = 2 ). Then, it chooses the one with the highest res idual ener gy . Practically , in some cases, the choice based on residual ener gy will be at t he e xpense of distance. In f act, the choice of the second node is not al w ays v alid mainly when the g ap between the first E r el a y and the second one is lar ge. Thus, it can results in more ener gy e xpenditure. In UCR, if CH s j s distance to the BS is smaller than TD-MAX , and CH s i selects s j as its relay node according to the approach described before and if the residual ener gy of s j is smaller than that of s i it l et s i communicate with the BS directly rather than aggra v ating the load of s j : this h ypothesis, in some cases, is not the best choice when the dif ference between s i :E r es and s j :E r es is too small and at the same time the distance between s i and BS is too long. In our solution, if a node’ s distance to the BS is smaller than TD-MAX , it transmits its data to the BS directly; otherwise it’ s better to find a relay node which can forw ard its data to the BS. Therefore, a tradeof f is made between the three criteria of residual ener gy , number of cl uster members and link cost E r el a y . CH selects the neighbor CH with higher residual ener gy and a smaller number of cluster members and minimum link cost E r el ay as the ne xt hop in order to balance the ener gy consumption among CH and reduce the ener gy cost of the link in the relay process. In this case, the CH s i calculates the cost function to choose the best relay node among all candidates S . The node that has the minimum v alue of C ost j will be selected as the ne xt relay . The cost function is defined by equation 9: C os t j = E r el ay ( s i ; s j ) R el ay max + E cons ( s j ) E max + s j :d N (9) R el ay max = max f E r el ay ( s i ; s j ) g ; s j 2 S (10) Where N is the number of ali v e node and s j :d is the number of cluster member of s j . E cons ( s j ) represents the ener gy consumed by the node j . W e can s ee from formula 9, the CH with minimum consumed ener gy (i.e., higher residual ener gy), minimum link cost and fe wer cluster members will ha v e smaller C os t . CH s i chooses the neighbor CH with the smallest C ost as its ne xt hop. 4. SIMULA TIONS AND RESUL TS In order to e v aluate the proposed mechanism, we use our simulator and we ha v e compare results to the UCR mechanism [10]. Our simulations in v olv e tw o parts. In the first part, data is sent to CH without controls. In the second part, we control data before sending it to CH. 4.1. Experiment settings and metrics Our paper adopts the same radio ener gy model with [17]. The ener gy consumed in the transmitter node ( E tx ) and in the recei v er node ( E r x ) with distance d for transmitting a b -bit data pack et can be calculated as follo ws: E tx ( b; d ) = b E e + b E f d 2 if d d c b E e + b E tr d 4 if d > d c (11) The parameters E f and E tr are the amount of ener gy dissipates per bit in the radio frequenc y amplifier according to the distance d c . d c is the threshold distance that depends on the en vironment. In our w ork we consider , both the free space ( d 2 po wer loss) and multi-path f ading ( d 4 po wer loss) channel models depending on the distance between the transmitter and the recei v er . W e assume free space model if ( d < d c ), otherwise multi-path f ading model is intended. The ener gy for recei ving a b -bit message is calculated as follo ws: E r x ( b ) = b E e (12) Where E e represents the electronics ener gy . It depends on such electronic f actors as digital coding, modulation, filtering, and spreading of the signal. Simulation parameters are gi v en in T able 1. F or simplicity , we assume: An Unequal Cluster -based Routing Pr otocol Based on Data Contr olling for WSN (S. Chelbi) Evaluation Warning : The document was created with Spire.PDF for Python.
2410 ISSN: 2088-8708 T able 1. Simulation parameters. P arameters V alues Netw ork field 400 x 200 BS location (500, 100) E e 50 nJ/bit E f r iss amp 10 pJ =bit=m 2 E tw o r ay amp 0 ; 0013 pJ =bit=m 4 d C 87 Data pack et size b 500 bytes Initial Ener gy of sensor E 0 1 J Number of sensors 600 TD-MAX 150 An ideal MA C layer and error -free communication links. Nodes are GPS-enabled and each node is a w are of its geographic location. Each node is assigned a unique id to help us identifying one node from other neighboring nodes. Sensors can use po wer control to modify the amount of transmission po wer according to the distance to the desired recipient. Each node can communicate directly with an y other node on the netw ork. Sensors are capable of operating in an acti v e mode or a lo w-po wer sleeping mode. In this paper , the performance will be e v aluated via simulations with respect to the follo wing metrics: First Node Dies (FND) and Half of the Nodes Ali v e (HN A). W e note that a node is considered ”dead” if its remaining ener gy is less than the v alue for the transmission task. 4.2. AEEUC without data contr olling Figure 2 sho ws that, under AEEUC, the le v el of residual ener gy of CH is lar ger than under UCR. In f act, in UCR protocol, CH are randomly selected to compete for final CH. Y et , in AEEUC protocol, only nodes which ha v e not been CH in N late round are el igible to be a tentat i v e CH for the current round. Thus, only nodes with suf ficient ener gy are selected as CH. This e xplains which leads to a v oiding the premature death of CH. Our protocol tends to minimize the unbalanced communication cost which is e v aluated for each CH as follo ws: Figure 2. A v erage of residual ener gy of elected CH. C unbal anced = M ax ( C i ) M in ( C i ) (13) IJECE V ol. 6, No. 5, October 2016: 2403 2414 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 2411 Where C i is the communication cost of CH i and is calculated as follo ws: C i = E cons E r es (14) Where E cons and E r es represent the consumed ener gy and the residual ener gy respecti v ely . The higher v alue of C unbal anced means the more unbalanced communication cost. Figure 3 plots the le v el of the unbalanced communication cost o v er time for dif ferent algorithms. Figure 4 sho ws the a v erage of the unbalanced communication cost. Figure 3. Unbalanced communication cost As sho wn in Figure 3 and 4, we find that the unbalanced communica tion cost in the proposed algorithm is lo wer than Figure 4. A v erage of Unbalanced communication cost the other algorithm. Our protocol minimize the unbalanced ener gy consumption among CH within the netw ork. Figure 5 sho ws the comparison between UCR and AEEUC in term of netw ork lifetime. Under AEEUC, the first node of the netw ork is died aft er 425 rounds while under UCR the first node die much before, i.e. 380 rounds. Moreo v er , HND is reached later under AEEUC (480 rounds) than under UCR (465 rounds). AEEUC gi v es better performances than UCR in prolonging netw ork lifetime. This can be e xplained firstly by the f act that the CH elected is the node with higher ener gy . Second, when nodes join clusters, the y consider both the distance to CH and the residual ener gy of CH. Third, CH choose those nodes as relay node, which ha v e minimum ener gy consumption for forw arding, fe wer number of cluster member and maximum residual ener gy to a v oid early death. An Unequal Cluster -based Routing Pr otocol Based on Data Contr olling for WSN (S. Chelbi) Evaluation Warning : The document was created with Spire.PDF for Python.
2412 ISSN: 2088-8708 Figure 5. Netw ork lifetime 4.3. AEEUC with data contr olling Our w ork aims to reduce the number of unnecessary data transmissions. T o achie v e this goal, a Maximum Data Error (MDE) technique is used to reduce the number of transmissions and thus considerable ener gy conserv ation is achie v ed. Figure 6 sho ws the change in the number of messages transmitted by the nodes to CH in a gi v en period. In the UCR Figure 6. V ariance of number of sent messages to CH protocol, e v ery non-CH node must transmit its sensed data to CH at e v ery iteration. In AEEUC protocol, before each transmission, all the nodes are entitled to v erify their data, b ut only parts of them will transmit their data to CH. This ob viously reduces the transmission ener gy consumption. As sho wn in Figure 7, the ener gy consumption of our protocol is less than that of UCR. Less ener gy consumption means longer lifetime for the netw ork. Figure 8 sho ws the v ariation in the number of all ali v e nodes based on the number of messages recei v ed by the BS. As sho wn in the figure, the number of dead nodes in the proposed protocol is al w ays less than that of UCR protocol. AEEUC clearly impro v es the netw ork lifetime o v er UCR. Simulations sho wed that AEEUC reduces the en- er gy dissipation and thus e xtending the o v erall netw ork lifetime. IJECE V ol. 6, No. 5, October 2016: 2403 2414 Evaluation Warning : The document was created with Spire.PDF for Python.