F-PAC: A Novel Soft Index Based Cluster Head Validation & Gateway Election Mechanism for Ad Hoc Network

Transcription

1 F-PAC: A Novel Soft Index Based Cluster Head Validation & Eletion Mehanism for Ad Ho Network S. THIRMURUGAN 1, E. GEORGE DHARMA PRAKASH RAJ 2 1 Department of Computer Appliations, J.J. College of Engg & Teh. Tiruhirapalli, India. s_thiru_gan@rediffmail.om. 2 Department of Computer Siene and Engineering, Bharathidasan University Tiruhirapalli, India. georgeprakashraj@yahoo.om Abstrat: - In this dynami senario the ommuniation no longer happens in predetermined manner. The network as a platform for ommuniation omes with high infrastruture may likely to waste the resoures. Thus, the ad ho senario network ame into existene. This network funtionality has been enhaned through lustering mehanism. These lusters need to be perfet to sustain the effiient funtionality of the network. Thus, this paper proposes F-PAC as a fuzzy logi based luster validation tehnique to authentiate the luster head identified by the existing luster formation mehanisms. This proedure also helps to elet the gateway node for eah luster. This study has been shown using OMNET++ as simulator. Key-Words: - F-PAC, W-PAC, Fuzzy. 1 Introdution It is well known that the ad ho network stands uniquely as a tentative network in the ommuniation proess. The network funtionality never remains onstant sine the number of nodes is not onstant always. There is an expetation upon the network funtionality to be stable irrespetive of the hange happens in network size. Unfortunately this ouldn t happen due to the inadequate maintenane of the network. If this setbak on the network hasn t been set right then this may ause olossal damage on the performane of the network. This issue an be addressed and solution an be attained with the help of some unique mehanism on the ad ho network. The lustering has been realized as one suh mehanism whih an provide the expeted results. The luster formation algorithms are used in hard way of forming lusters. There should be some soft approah whih would be helpful to deide the degree of loseness of nodes in a luster. The F-PAC(Fuzzy based Partitioning Around Cluster head) will apply soft method to asertain the luster head identified by WCA[1] and W-PAC[2] proedure. This paper has been organized as follows. Setion.1 deals with the introdution. Setion.2 gives out the literature study. Setion.3 tells about existing approahes. Setion.4 speaks about W- PAC proedure. Setion.5 puts down the proposed F-PAC proedure. Setion.6 shows experimental results and analysis. Setion.7 speifies the future diretion of this work. Setion.8 ends up with the onluding remarks. 2 Literature study The power level of nodes deided the sustainability of the luster head role. The transmission power alone is not suffiient to alulate the weight of the node. The power reward[3] based weight alulation ensures the uniform power distribution. The lustering an also be formed based on the signal strength[4] between the luster head and nodes belong to luster. The luster head has been omputed using the signal strength expression. The luster head will selet the nodes for the luster on the basis of signal strength. The dominating sets are identified in the lustered network. In whih the minimum independent set[5] an be onstruted and the tree struture of the same an be formed later. The onneted dominating set algorithm has been a bakbone to form the lusters. The luster head funtionality an be tampered by the maliious node whih behaves like luster head. The SWCA[6] proposed a seured weighted lustering algorithm to keep the network away from suh maliious nodes. E-ISSN: Volume 13, 214

2 The NWCA[7] has been proposed to improve the weight based algorithms through hanging methodology of parameter alulation for weight. The degree omputation has been hanged to mean onnetivity degree. This novel method also onsiders the energy level of the nodes to play the role of luster head. The DWCA[8] onsiders the luster formation based on weight, mobility fator and luster maintenane. The new node addition to luster has been handled through distint approah by this protool. The weight based lustering has been improved[9] on the basis of reduing the load of the luster head with the help of threshold value. This limits the luster size to ensure the luster head to last longer. The role of lustering in ad ho networks has been realized when AODV[1] inorporates the lustering mehanism. This added mehanism enhanes the funtionality of the network. The appliation of the lustering mehanism[11] with AODV as a routing protool in the real world senario has been indispensible. This shows that lustering tehnique makes the network to be suitable for various appliations. The lustering mehanism PAC[12] over the k-means approah tells the purpose of parameters in luster formation. These parameters deide the effiieny level of lusters. This study also onfirmed that k-means takes more time when the number of nodes are high in ount. This work laks in implementation and also the sample set of nodes are small in size. The PAC has shown good results when the number of nodes are less. It leaves many nodes as non lustered nodes. The Ex-PAC [13] ame out as an extension to PAC whih takes entire nodes and produes the maximum lusters. The luster formation proess ultimately improved in Ex-PAC proedure. This approah onludes that Ex-PAC has outperformed k-means in terms of omputational speed. The multi-parametri swarm intelligene based lustering mehanism PSO-PAC[14] takes the neessary parameters to identify the luster head of the luster. This parameter optimization will suit speifi appliation. The re-lustering should be based on identifying the strength of the existing lusters. The role of various indies[15] on evaluating the luster should be understood very well. The luster lassifiation[16] also plays key role in determining the perfetness of the lusters. Those lassifiations are of numeri, disrete and partitioned types. It also finds out the preferred lustering method for a given sample set of nodes. The luster formation proedure will not onfirm the perfetness of the lusters. The validation[17] proess onfirms the perfetness of the luster to determine the stability fator of the luster. This fator helps in finding out the re-lustering time based on the measured value. 3 Existing Approahes The highest degree algorithm[18] onsiders only degree of the node to form the lusters. This doesn t take other parameters into aount to deide the members of the partiular luster. Thus, δ-degree lustering algorithm[19] has improved highest degree algorithm by onsidering speed and link fator into aount. The stability of the link has been deided based on fuzzy membership degree. The fuzzy set has been onsidered with near, far and medium as fuzzy names to deide the membership of the nodes. The fuzzy logi based lusterhead eletion algorithm proposed[2] laks in obtaining the experimental results instead speifies the fuzzy rules to elet the lusterhead. The multi-parametri algorithm WCA takes more time to form the lusters. Thus, W-PAC mehanism has been onsidered to form the lusters sine this elets the lusterhead by onsidering multiple parameters. 4 W-PAC Fig.1 Clustered Network The Fig.1 shows the lustered network struture of W-PAC proedure. The W-PAC algorithm takes multiple parameters together in the name of weight to identify the luster head. Node E-ISSN: Volume 13, 214

3 WSEAS TRANSACTIONS on COMPUTERS W-PAC Cluster Creation Proedure (1) Initialize set of nodes as M. (2) Compute the degree of node Ni. (3) Deg (Ni) =. (4) j = 1. (5) If ( i not equal to j) begin Manhattan Dist (Ni, Nj) = If ( Manhattan Dist( Ni, Nj) < Radious ) begin Add ( Ni, Cm) // add to luster Deg(Ni) = Deg(Ni) + 1 j = j + 1 end else Add ( Ni, NCn) // add to Non luster end (6) Repeat the step 5 until j = M. W-PAC Cluster Head Eletion Proedure (1) Create Clusters using W-PAC luster reation. (2) Cluster = Ci, P = Number of nodes in Ci. (3) j = 1; Ni = (U, V ); Nj = (U, V ; (4) If ( i not equal to j) begin If ( Manhattan Dist(Ni,Nj) < Radious ) begin Compute the Mobility speed of Node Ni of Ci. 1 M(Ni) = T MODU U Compute the Distane between Ni and Nj. D(Ni) = MODU U MOD X X Y Y V V V V (8) Repeat the step 7 for all nodes belong to Ci. (9) k = Max { W(N 1 ),W(N 2 ),W(N 3 ) W(N M ) }. (1) Repeat the step 2 through 9 for i = 1..no of lusters. W-PAC algorithm ertainly improves WCA and shows performane improvement in obtaining effetive results while the number of nodes and their mobility level is high. The nodes whih are identified as part of lusters have to onfirm their identity within the speifi lusters and their ommuniation with the luster head. 5 F-PAC This proedure is based on fuzzy logi and fuzzy set. The Fuzzy logi deals with the possible value or truth value or approximate value rather than identifying the fixed hard values. The values are neither 1 and nor whih lies between and 1.This approah is so alled as soft approah on luster formation. It is against the binary logi whih says the value ould be either or 1. This fuzzy logi has been expressed in terms of fuzzy set onstrution omprise of nodes as set members. A fuzzy set is a pair (M,K) where M is set of nodes and K is alled the degree of membership of the nodes denote as M : K [,1]. For a finite set ontains the elements as shown(1). M N1, N2 Nn Dij Di1, Di2 Din i 1,2. Ci (1) The degree of the nodes Dij an have the following values as shown (2). When Dij value is either or 1 then the node will not belong to the set or definitely ontained in the set. But the value when lies between and 1 then the node will beome fuzzy member and value determines how far the node is truthful in its property. j = j + 1 end end (5) Repeat the step 4 until j = P. (6) Assume the Energy of nodes E(Ni) for all the nodes. (7) The weight of node Ni omputed as follows, W(Ni) = q1*deg(ni) + q2*m(ni) + q3*d(ni) + q4*e(ni) Dij Dij Fuzzy set K (2) Dij 1 Dij Fuzzy set K 1 The set { Ni M Dij > } is alled the support of (M,K) and the set {Ni M Dij =1 } is alled its E-ISSN: Volume 13, 214

4 WSEAS TRANSACTIONS on COMPUTERS kernel. The degree of node an take values between and1 has been realized (3). These nodes are part of Dij 1, i 1,2 C, j 1,2 n Dij 1, luster, will have the membership degree. The sum of those membership degrees of all nodes within the luster will not exeed the maximum value 1. This is not only true for intra luster but also for inter luster. F-PAC algorithm: Cluster Head Identifiation 1. Input the lusters formed using WCA or W-PAC algorithm. 2. Input the Cluster Ci 3. Compute the degree D ij of the node N i. Dij = j 1,2 n Dij, 1,2 1 dist, Ni (3) 4. Compute the luster Center based on membership degree Dij. Dij dist, Ni Center = dist, Ni 5. Repeat the steps 2 and 3 for all the nodes in the luster. 6. { To Find the number of nodes from luster head less than or equal to Computed Center } For i = 1.. n { n number of nodes in Cluster Ci} If ( dist (,Ni) < Center) Count[i] = Count[i] {To find the Max ount in a luster } For i = 2 n Max=Count[1]; { tentative maximum } If ( Count[i] > Max) Max = Count[i]; 8. Repeat the steps 2 through 6 for i = 1..No of lusters. 9. Identify the luster head for maximum ount. 1. If the luster head = luster head identified in the ase of WCA or W-PAC then luster head is valid. This algorithm takes output from WCA or W-PAC as input for further proessing. It finds the degree of eah node and then Center for eah luster. The Center will be taken as a referene to find the number of nodes within the luster. This Center signifies the maximum distane the luster head and nodes an have while their existene in the same luster. The Count[i] finds the maximum number of nodes in a luster based on Center value. The luster head of the luster whih has max Count[i] will be hosen. If the luster head identified by this fuzzy proedure is same as luster head identified by the WCA or W-PAC proedure then say luster head validation has reahed suess. F-PAC algorithm: Eletion 1. Input the lusters formed using WCA or W-PAC algorithm. 2. Find the distane between node N and luster head Dist(,Ni) = xi yi 3. Compute the degree Dij of the node N i. Dij = 1 dist, Ni 4. Repeat the steps 2 & 3 for all nodes in luster. 5. = least ( of node Dij). 6. Repeat the steps 2 through 5 for eah luster. This Proedure onsiders the non overlapping lusters to onstrut distributed gateways. These gateways will be identified based on their degree values. Higher the degree values indiate the loseness of the node towards the luster head. Thus, the lower value nodes will be near to the edge of the luster boundaries to at as gateway nodes. The maximum membership value lies between and 1. 6 Experimental Results The F-PAC algorithm has been implemented using OMNET++ Tool and the results are tabulated. This work has been arried out with the system onfiguration of 64bit AMD proessor, 2GB RAM and windows XP as an operation system. This simulation has been done for 1 nodes and 25 nodes. Table.I shows the simulation parameters. The luster heads identified using WCA and W-PAC have been mentioned. These lusterheads need to be validated using F-PAC to onfirm their plae within eah luster. If F-PAC results are showing different lusterheads then the lusters are imperfet in their existene. E-ISSN: Volume 13, 214

5 Clusterhead Node Table.II Cluster heads of F-PAC, WCA and W-PAC Parameter N (Number of ) Spae (area) Table.I Simulation Parameters Tr (Transmission range) WCA,W-PAC Cluster heads (1 nodes) WCA, W-PAC Cluster heads (25 nodes) Cluster Existing Methods WCA C1 C2 C1 C W-PAC C1 N2 N2 N2 C2 N7 N7 N7 C1 N3 N3 N3 C2 N19 N19 N19 Number of nodes and Clusters Values 1, m N2, N7 N3,N19 Fig.2 F-PAC Vs WCA and W-PAC results Proposed F-PAC WCA W-PAC F-PAC Table.II shows that the luster head eleted using WCA and W-PAC proedure and validated using F- PAC proedure at time T1. This validation proess onfirms that the luster heads are properly identified by WCA and W-PAC proedure. But WCA takes more time to form lusters than W- PAC. Fig.2 shows the lusterheads eleted for the existing methods and proposed methods. WCA and W-PAC give the same results as for as the lusterheads are onerned. These two algorithms differ on time what it takes to identify the lusterheads. This study has onsidered two sets of sample for the results. The first sample of 1 nodes forms two lusters C1 and C2 with the lusterhead nodes N2 and N7 respetively. The sample of size 25 nodes ontains two lusters with the lusterhead nodes N3 and N19 respetively. These lusterheads are eleted by the WCA and W-PAC algorithms. The results are same sine the two algorithms are applying same methodology in identifying lusterheads but they differ in luster formation time. These lusterheads need to be validated to onfirm the proper seletion of lusterhead at time T1. The proposed F-PAC proedure has identified the lusterhead of eah luster. These lusterheadss are same as WCA and W-PAC lusterheads. Thus, the lusters formed an sustain without any hanges. The hange of lusterhead is needed while the lusterhead makes a move. This may end up with the re-eletion of lusterhead. Fuzzy logi based Clusterhead Re-eletion: Fuzzy logi based Validation deides the perfetness of the lusterhead seleted by multi- The forthoming parametri lustering algorithms. analysis desribes the validation results at time periods T1 and T2. Time T1 has been initial validation of the lusterhead seleted by the luster formation proedures. Time T2 has been validation after the mobility of the lusterhead eleted. Time T1: Y C2 C1 Fig.3 Clusters C1 and C2 at time T1: 25 Fig.3 shows the initial level of lusters formed using W-PAC algorithm exluding outliers at time T1. The lusterhead eleted using W-PAC proedure has been validated using F-PAC proedure. This validation proess onfirms that the luster heads are properly identified by W-PAC proedure. The luster heads of lusters C1 and C2 have mathed with the luster heads of same lusters identified by X E-ISSN: Volume 13, 214

6 the F-PAC proedure at the time T1. The WCA results are same as W-PAC. This situation may prolong till the lusterheads are stationary. Y Time T2: The lusterhead N19 of luster C2 has moved. Under this assumption the fuzzy logi based validation happens at time T2. Now, the position of lusterhead N19 has moved fromm (34,56) to new position (59,47). This drives the hange to happen within the luster C2. At this time T2, F- PAC has been applied and new results are obtained. The node N2 has been onsidered as new enter point(clusterhead) of the luster C2. This reeletion has been ahieved using F-PAC proedure. Y Fig.5 shows the old and new lusterheads after applying the F-PAC proedure. The old has gone out of the ommuniation range of the luster X Fig.4 Mobilty of Node N2 from C2 at time T2: 25 nodes New Old X Fig.5 Node 2 as new of C2 at time T2: 25 nodes C2. This node has lost its membership status of C2. It will be onsidered as outlier. 5.1 Eletion Fuzzy logi based proedure not only finds the valid lusterhead and re-elet the lusterhead but also identifies the gateway node for eah luster of the ad ho network. The gateway nodes are identified for eah luster to have inter-luster ommuniation between the lusters. To identify this gateway the fuzzy membership degree values have to be observed. The node whih is far away from luster and lies at the edges of the luster boundary will have least degree. The node whih is loser to lusterhead will have high degree while it is ompared with other nodes degree. This gateway is known as distributed gateway. Sine eah luster will fix its own gateway node. F-PAC produe plaes a key role in determining the gateway node. 1 Table. III Eleted s Cluster Node (WCA) Node (W-PAC) of Node C1 N4 N4.7 C2 N8 N8.6 Table. III shows the eleted gateways of luster C1 and Cluster C2 for the sample of 1 nodes. This gateway node of the W-PAC has been same as WCA. It differs in gateway node identifiation time. In luster C1, the node N4 has the least degree value to be onsidered as gateway whereas for the luster C2 node N8 holds the minimum membership degree value to be eleted as gateway Fig.6 Graphial Illustration: 1, Cluster C1 The Fig.6 shows the nodes and their degree values. It is obvious that the node N4 arries lower degree while this is ompared with other degrees. This least value degree represents the degree of loseness of the nodes towards the luster boundary. The Higher E-ISSN: Volume 13, 214

7 degree node will be loser to the luster head. This will be helpful to deide the gateway node whih would be loser to boundary of luster and keep themselves away from the luster head node. The Fig.7 shows that node N8 has been the least degree node omparatively to the other nodes of luster C The Table.IV shows the eleted gateways and their respetive degree values for eah luster where 25 nodes are onsidered as sample size. The node N16 has been onsidered as gateway sine this holds the least membership degree value. The luster C2 elets N23 as gateway node beause of its minimum degree value. Fig.7 Graphial Illustration: 1, Cluster C Cluster Table. IV Eleted s Node(WCA) Node(W- PAC) of Node C1 N16 N16.7 C2 N23 N Fig.9 Graphial Illustration: 25, Cluster C2 The Fig.8 pitorially shows that node N16 has been the least degree node belongs to the luster C1. This node will be onsidered as gateway node as far as the luster C1 has been onerned. The Fig.9 shows that node 23 has been the least degree node belongs to the luster C2. This node plays the role of gateway for the luster C2 to ensure inter-luster ommuniation. The gateway nodes eleted also hanges based on their mobility property. This gateway re-eletion ould be handled as separate task of F-PAC. While the gateways are going out of ommuniation range of lusterheads then the lusterhead ould be able to sense it sine the absene of the periodi signal from gateway to lusterhead. This may invoke the F-PAC gateway eletion proedure to re-elet the gateway node. 7 Future Diretion The fuzzy logi an be further utilized on the ad ho network to identify the nodes falling into the overlapping region. The ommon gateway node identifiation has to be done. The experimental results an be further enhaned with the help of more number of nodes. 8 Conlusion This study learly exposes the role of fuzzy logi in luster formation in ad ho network. It onfirms the luster head seleted by the existing WCA and W- PAC proedure with the help of F-PAC proedure. In this way F-PAC has been highly helpful in validating the luster head eletion. The purpose of F-PAC proedure has been learly explained for lusterhead re-eletion and gateway eletion proedure to ensure inter-luster ommuniation in ad ho network. Fig.8 Graphial Illustration: 25, Cluster C1 E-ISSN: Volume 13, 214

8 Referenes: [1] Anna Gorbenko, Vladimir Popov. Clustering Algorithm in Mobile Ad Ho Networks. Adv. Studies Theor. Phys., Vol 6, pp , 212. [2] S.Thirumurugan, E.George Dharma Prakash Raj W-PAC: An Effiient weighted Partitioning Around Cluster Head Mehanism for ad ho network, CCSEIT 12, 212, pp [3] Na Zhang, Ningqing LIU, Weixiao MENG, Qiyue YU. A Novel Weighted Clustering Algorithm Based on Power Distribution for Cooperative Communiation Network. PCSPA 1, 21, [4] Dao-QuanLI, Wei-HuaXUE, Qi-Guang CAO, Huai-Cai WANG. Researh on Ad Ho network lustering algorithm based on signal strength. ICIS 1, Aug 21, [5] Dao-QuanLI, Wei-HuaXUE, Qi-Guang CAO, Huai-Cai WANG. A New Distributed CDS Algorithm of Ad Ho Network Based on weight. ISDEA 1, 21, [6] Yingpei Zeng, Jiannong Cao, Shanqing Guo, ai Yang, Li Xie. SWCA: A Seure Weighted Clustering Algorithm in Wireless Ad Ho Networks.WCNC 9, April 29, 1-6. [7] Yang Wei-dong. Weight-Based Clustering Algorithm for Mobile Ad Ho Network. CSQRWC 11, 211, [8] Naveen Chauhan, Lalit Kumar Awasthi, Narottam Chand, Vivek Katiyar and Ankit Chugh. A Distributed Weighted Cluster Based Routing Protool for MANETs. International Journal of Wireless Sensor Network, Vol.3(2), Feb 211, [9] Sudhakar Pandey, Narendra Kumar Shukla. Improved Weighted Clustering Algorithm for Mobile Ad Ho Networks. IJEMS,Vol 2(1), Jan 211, [1] S.Thirumurugan, Diret sequened C-IAODV Routing Protool, International Journal of omputer siene and Tehnology, vol.1(2), 21, pp [11] S.Thirumurugan. C-AODV: Routing Protool for Tunnel s Network, International Journal of omputer siene and tehnology, vol.2(1), 211, pp [12] S.Thirumurugan, E.George Dharma Prakash Raj, PAC - A Novel approah For Clustering Mehanism in Adho Network, ICSCCN 11,211,pp [13] S.Thirumurugan, E.George Dharma Prakash Raj, Ex-PAC : An Improved Clustering Tehnique in Ad Ho network, RACSS 12, 212,pp [14] S.Thirumurugan, PSO-PAC: An Intelligene Clustering Mehanism in Ad ho Network, NETCOM 12, 212,pp [15] Mahesh Visvanathan Adagarla, B Srinivas Gerald, H Lushington Peter Smith, Cluster validation: An integrative method for luster analysis, IEEE International Conferene on Bioinformatis and Biomediine workshop, pp238, 29. [16] Rama,B., Jayashree,P., Salim Jiwani, A Survey on lustering urrent status and hallenging issues, International Journal on omputer siene and engineering, 2(9), 21, pp [17] S.Thirumurugan, An Extended Weighted Partitioning Around Cluster head Mehanism for Ad Ho Network, International Journal of Ad Ho, Sensor & Ubiquitous Computing, 3(6): 15-27, 212. [18] Hyun Jeong Lim. A Study on the Clustering Sheme for Node Mobility in Mobile Ad-ho Network, Advaned in Computer Siene and its Appliations, pp , 214. [19] ZHAO Chun-xiao, WANG Guang-Xing. Fuzzy-Control-Based Clustering Strategy in MANET, Proeedings of the 5th World Congress on Intelligent Control and Automation, pp , 24. [2] K. Venkata Subbaiah, Dr. M.M. Naidu. Cluster head Eletion for CGSR Routing Protool Using Fuzzy Logi Controller for Mobile Ad Ho Network, Int. J. of Advaned Networking and Appliations, 1(4): , 21. E-ISSN: Volume 13, 214