A clustering protocol using multiple chain

525 views

Published on

Published in: Education, Technology, Business
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
525
On SlideShare
0
From Embeds
0
Number of Embeds
2
Actions
Shares
0
Downloads
0
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

A clustering protocol using multiple chain

  1. 1. 586 JOURNAL OF NETWORKS, VOL. 5, NO. 5, MAY 2010 A Clustering Protocol Using Multiple Chain Strategy in WSNs Zheng Gengsheng, Hu Zhengbing School of Computer Science & Engineering, Wuhan Institute of Technology, China Department of Information Technology, Huazhong Normal University, China Email: zhenggengsheng@sina.comAbstract—In WSNs, energy efficiency and low latency are The rest of the paper is organized as follows: First weconsidered as two key issues in designing routing protocol. give an overview of the related works in section 2. ThenThis paper introduces two schemes to combine clustering the network and communication models of our proposalstrategy with chain routing algorithm in order to satisfy are discussed in section 3. Two scheme of clusteringboth energy and delay constraints in WSNs. Furthermore protocol using chain routing algorithm are introduced inby one scheme, we propose a two layer hierarchical routingprotocol called Chain Routing Based on Coordinates- section 4. A detail description of our approach CRBCC isoriented Clustering Strategy (CRBCC), which gives a good presented in section 5. Simulation results make acompromise between energy consumption and delay. First, comparative analysis between PEGASIS and CRBCC inCRBCC makes balanced clustering according to y section 6. Finally, section 7 presents a short conclusioncoordinates where each cluster has approximately equal with future work.number of nodes. Second, CRBCC makes chain routing bysimulated annealing algorithm (SA) inside the cluster and II. RELATED WORKSelects chain leader in the order of x coordinates. Third, Low-Energy Adaptive Clustering Hierarchy (LEACH)CRBCC makes chain routing again by SA method amongchain leaders. Simulation results show that CRBCC ([13], [14]) is a clustering protocol. LEACH reduces theperforms better than PEGASIS in terms of energy efficiency number of nodes communicating directly with BS. Theand network delay. protocol achieves this by forming different clusters in a self-organizing manner, where each cluster-head collectsIndex Terms—WSNs; PEGASIS; CRBCC; simulated the data from all nodes in its cluster, fuses and sends theannealing algorithm information to BS. This decision is made by the node n choosing a random number between 0 and 1. If the number is less than a threshold T(n), the node n becomes I. INTRODUCTION a cluster-head for the current round. The requirement for designing routing protocol inWSNs is different from in the other network. The main  Pconstraint of WSNs is their very limited battery energy, which has great influence on the lifetime and the quality T(n) =  1 − P * [r mod(1/P)] ,n ∈ G  (1)of the network ([1], [2], [3], [4], [5], [6]). On the other 0,otherwise hand, transmission time is also another important factorto a routing protocol in the real-time environment. For where P is the desired percentage of cluster-heads, r is thethat reason, the protocols running on sensor networks current round, and G is the set of nodes that have notmust consume the energy efficiently while transmit been cluster-heads in the last 1/P rounds. Using thisinformation in a low delay. There are many routing threshold, each node will be a cluster-head at some pointprotocols designed for wireless sensor networks to satisfy within 1/P rounds. LEACH can distribute the energyenergy efficiency or low latency requirement ([7], [8], [9], among the nodes in the network and enhance system[10], [11], [12]). lifetime. Many algorithms have been proposed addressing the PEGASIS [15] is a chain-based protocol, which makesapplication and architecture requirements. Most of the further improvement on LEACH. PEGASIS makes aprotocols can be classified as either flat or hierarchical chain routing through Greedy algorithm, which is alsobased. In hierarchical routing protocol, the network is called the nearest neighbor algorithm. In the chain, eachdivided into clusters with one cluster-head acting as node receives from and transmits to the closest neighbor.sender to base station (BS). Data are gathered and fused The elected chain-leader is responsible for transmittingat each cluster-head before transmission to the BS. the final information to the base station.Among the hierarchical protocols, Low-Energy Adaptive Greedy chain algorithm begins at a farthest node fromClustering Hierarchy (LEACH) ([13], [14]) and Power BS, which is the only node in the chain at first. EachEfficient Gathering in Sensor Information Systems terminal node of the chain finds a closest node from the(PEGASIS) [15] are very representative. remaining nodes set which are not in the chain. Then the closest node will join the chain and be the new terminal node of the chain. The process repeats till all nodes join© 2010 ACADEMY PUBLISHERdoi:10.4304/jnw.5.5.586-593
  2. 2. JOURNAL OF NETWORKS, VOL. 5, NO. 5, MAY 2010 587the chain. The Greedy chain algorithm in PEGASIS is as (1) Each sensor node has power control and the abilityfollows. Here, N is the nodes set, which includes all to transmit data to any other sensor node or directly to thenodes at first. END is the terminal node of the chain. BS.Variable chain is the required chain to be constructed. (2) Our model sensor network contains homogeneousThe function FindCloseNode(N,END) means to find the and energy constrained sensor nodes with initial uniformclosest node to END node in the remaining nodes which energy.are still not in the chain. The function (3) Every node has location information.Append(chain,END) means to add the terminal node (4) There is no mobility.END into the chain chain tail. B. Radio Model C0 C1 C2 C3 C4 In our work, we used the first order radio model. In the radio model, the energy dissipation of the radio is to run the transmitter or receiver circuitry. Depending on the Base Station distance between transmitter and receiver, the free space propagation channel model is taken into account to Figure 1. PEGASIS chain measure the energy loss due to wireless transmission. Therefore, the energies expended to transmit a k-bit packet to a distance d and to receive that packet with this Procedure ConstructGreedyChain(N,END) radio model are as follows. 1.Begin For transmitter, 2. N={all nodes}; 3. END = farthest node from BS;  E tx (k,d) = E tx − elec (k) + E tx − amp (k,d)  (2) 4. chain= {END}; 5. N=N-{END}; 6. if(N!=NULL) E tx (k,d) = E elec *k + ε amp *k*d 2 (3) 7. { 8. END=FindCloseNode(N,END); For receiver, 9. Append(chain,END); 10. goto 5. E rx(k) = E rx − elec(k) (4) 11. } 12.END E rx(k) = Eelec*k (5) Here, Eelec are energies required in running transmitter Figure 2. Greedy chain algorithm. and receiver electronics. ε amp is the energy consumed in The main advantages of PEGASIS are: (1) The transmitter amplifier for transmission.transmission distances between nodes are minimized. (2) We make the assumption that the radio channel isThe number of sensor nodes that must send packets to the symmetric such that the energy required to transmit abase station is minimized. The main drawbacks of message from node A to node B is the same as the energyPEGASIS are: (1) It has excessive delay introduced by required to transmit a message from node B to node A forthe single chain. (2) Greedy algorithm using in PEGASIS a given SNR. For our environment, we also assume thatis a local search, which cannot provide an approximate all sensors are sensing the environment at a fixed rate andglobal optimal route. thus always have data to send to the end-user. In this work we propose a coordinates-based clusteringrouting protocol, which satisfies the requirements stated IV. CLUSTERING SCHEME BASED ON CHAIN ROUTINGabove. By using multiple short chains, we can alleviate We propose two schemes to combine clusteringthe excessive delay caused by PEGASIS while achieving strategy with chain routing algorithm in order to satisfymore energy gain by a more global optimal routing. both energy and delay constraints in WSNs. III. THE NETWORK AND RADIO MODELS Clustering strategy is useful for low latency while chain routing algorithm is beneficial to energy efficiency.A. Network Model There are two schemes for clustering protocol based on chain routing. We consider a 100-node network with randomlydistributed nodes in a (100 x 100) meter area. The BS is A. Scheme 1located at (x=50, y=300). The length of each signal is Clustering strategy + Direct transmission inside cluster2000 bits and the energy required for data aggregation is + Chain routing among cluster-heads5nJ/bit/signal. Moreover, data processing time per node is First, all nodes make clustering. Then chain routing istaken as 5-10 milliseconds. The radio speed is considered made among cluster-heads from each cluster. When dataas 2 Mbps. transmission begins, each node in cluster sends data to its In the paper, we assume the following: cluster-head directly. Then each cluster-head fuses and transmits its data along the chain. At last, one chain- leader sends data to BS directly.© 2010 ACADEMY PUBLISHER
  3. 3. 588 JOURNAL OF NETWORKS, VOL. 5, NO. 5, MAY 2010B. Scheme 2 algorithm uses Metroplis rule to get the probability of Clustering strategy + Chain routing inside cluster + acceptance Pa as followsChain routing among chain-leaders First, all nodes make clustering. Then chain routing is 1,∀f(xnew ) < f(xold ) pa =  (6)made among all nodes in each cluster and among all exp((f(x ) − f(xnew ))/T),∀f(xnew ) ≥ f(xold ) oldchain-leaders respectively. When data transmissionbegins, each node in cluster sends data along the lowest where f(x) is object function which represents the totalchain. Then each chain-leader from each cluster fuses and distance along the route x. But in CRBCC, objecttransmits its data along the top chain. At last, one selected function f(x) means the total energy consumption alongtop-leader in top chain sends data to BS directly. the route x. In WSNs, we want to find an approximate By the contrast between two schemes, the second lowest energy consumption route x to minimize f(x)scheme performs better using Energy*Delay metrics. through SA algorithm. So, we can setThis is mainly for chain routing among all nodes incluster in the second scheme is more energy efficient than f(x) = ∑ (E tx (k,d i,j ) + E rx (k))  (7)direct transmission to cluster-head in cluster in the first d i,j ∈xscheme. In the following sections, a protocol called CRBCC where di,j belongs the asked route x and represents thebased on the second scheme is proposed in detail. distance between node i and j. That is,Simulations show it can give a good compromisebetween energy consumption and delay. f(x) = ∑ (2*E d i,j ∈x elec *k + εamp*k*d i,j )  2 (8) V. PROTOCOL OF CRBCC Since Eelec, k and ε amp are all constants, f(x) can be as This paper proposes a two layer hierarchical routingprotocol called CRBCC, which ensures the minimumenergy consumption and low latency. Network routing f(x) = ∑d d i,j ∈x 2 i,j  (9)begins with the formation of clusters. Several clusters areformed with the approximately equal size according to ycoordinates. All nodes within a cluster form a chain The cooling schedule is given byamong them using a simulated annealing algorithm. Then T(n + 1) = α*T(n) (10)a chain leader of each cluster is elected in x coordinatesorder. After that a higher-level chain is formed including where α is constant with α < 1 , and n is the times ofall leader nodes from every cluster using SA too. Among cooling temperature.this leader-to-leader route only one randomly chosen topleader sends data to BS. The operation of CRBCC can be Proceduredivided by three phases: route formation phase, data ConstructSAChain(current_route,threshold)transmission phase and route maintenance phase. In the 1.current_cost = energy_cost(current_route);following subsections we discuss them in details. 2.while(iterations < threshold1) 3. new_route = SwapNode(current_route);A. TSP Problem in CRBCC 4. new_cost = energy_cost(new_route); One of the TSP problems in WSNs is the chain routing 5. diff = abs(new_cost - current_cost);where we want to find an approximate optimal route to 6. if(new_cost < current_cost)minimize total energy consumption in gathering data. The 7. current_route = new_route;traveling salesman problem (TSP) is as follows: Given a 8. current_cost = new_cost;finite number of “cities,” along with the cost of travel 9. if(temperature_iterations>= threshhold2)between each pair of cities, find the cheapest way of 10. cooling(temperature);visiting all the cities and returning to the starting point. 11. temperature_iterations = 0;There is several well know classical methods for TSP 12. endincluding Simulated Annealing (SA) algorithm ([16], 13. temperature_iterations ++;[17]). Simulated Annealing is based on heuristics from 14. iterations ++;annealing process. During the initial search phase (when 15. elsetemperature is high) local search is carried out and routes 16. if(rand(1) < exp(-diff/(temperature)))that minimize the total distance are always accepted. To 17. current_route = new_route;escape the problem of getting stuck in local minima 18. current_cost = new_cost;occasionally routes with distance more than the current 19. temperature_iterations ++;route is also accepted but with a probability similar to the 20. iterations ++;probability in the dynamics of the annealing process. As 21. endthe temperature decreases, this probability of accepting a 22. endbad solution is decreased and in the final stages it 23.endbecomes similar to gradient-based search. The SA Figure 3. SA Chain algorithm.© 2010 ACADEMY PUBLISHER
  4. 4. JOURNAL OF NETWORKS, VOL. 5, NO. 5, MAY 2010 589 The SA chain algorithm is as figure 3. The function 1) Communication in clustercost=energy_cost(route) evaluates the energy During this data communication phase each normalconsumption cost along the route route. The function node is responsible to collect and fuse data from neighbornew_route=SwapNode(old_route) makes a new route nodes in the chain of the cluster. At the end of this phase,new_route based on the old route old_route. The one chain leader in each cluster gets the information of allfunction cooling(T) cools the temperature T down. nodes in that chain. The schedule can use Token or TDMA scheme.B. Route Computation 2) Communication among clusters The main activities in this phase are cluster setup, In this collection phase, each chain leader node fuseschain routing in cluster, chain leader election, leader-to- and transmits its data along the top leader-leader chain.leader chain formation, top chain-leader election and The schedule assignment can also use Token or TDMAschedule creation. method. Last, one randomly chosen top leader fuses the The CRBCC routing algorithm is as follows. information of all nodes in WSNs and transmits to BS directly. Procedure Routing(void) To reduce the delay, multiple chains make parallel 1. Begin communication using different CDMA codes. When a 2. if ( in routing formation phase ) node decides to become a chain leader, it chooses 3. { randomly from a list of spreading codes. It informs all the 4. make balanced clustering by Y coordinates; nodes in the cluster to transmit using this spreading code. 5. SA chain routing in cluster; The chain leader then filters all received energy using the 6. elect chain leader in X coordinates order; given spreading code. Thus neighboring clusters’ radio 7. SA chain routing among chain leaders; signals will be filtered out and not corrupt the 8. elect top leader; transmission of nodes in the cluster. 9. } 10. if ( in data transmission phase ) y BS 11. { 12. communication in cluster; 13. communication among chain leaders; 14. } 15. if ( node is failed ) 16. make route maintenance; 17. End Figure 4. CRBCC routing algorithm. x First, CRBCC makes balanced clustering according to Normal nodey coordinates where each cluster has approximately equal Chain leadernumber of nodes. Normally in many literature p=5% Top leadernodes acting as cluster-heads is considered as an optimal Figure 5. CRBCC sketch map.result. Thus, in our experiment, 100 nodes of WSNs aredivided into 5 clusters with 20 nodes in each. Second,CRBCC makes chain routing with SA algorithm inside D. Route Maintenancecluster. After that, a chain leader of each cluster is elected Every a periodic time, neighbor nodes in chain willin x coordinates order which can ensure the short distance contact with each other using short detection messages. Ifbetween leader-leader. Till now, there are 5 chains with 5 a node does not reply, the neighbor node considers it tochain leaders. Last, CRBCC makes chain routing again be dead. The neighbor node will skip the dead node onwith SA method among chain leaders. Then a top chain the routing table and send the information to the nextleader is randomly chosen to forward the data to BS. BS sensor node.without energy limitation can make route computation. When 10% of the sensor nodes are found to be dead inThen the whole routing table is broadcasted to all nodes the WSN, the base station will move into the routingin WSNs. Each node keeps the routing table information computation phase again.related to it. AC. Data Gathering B F The data gathering phase is divided into two parts, H Dead E Cwhich are communication phase in the cluster andcommunication phase among clusters. D At the end of this phase, BS should broadcast the endof communication in current round and initiate the next Figure 6. Chain routing failure.round communication.© 2010 ACADEMY PUBLISHER
  5. 5. 590 JOURNAL OF NETWORKS, VOL. 5, NO. 5, MAY 2010 TABLE I. NORMAL ROUTING INFORMATION TABLE Node Routing Information Table H HABCDEF A HABCDEF B HABCDEF C HABCDEF D HABCDEF E HABCDEF F HABCDEF F A Figure 8. Placement of WSNs. B E H D Figure 7. Recover chain routing. TABLE II. RECOVER ROUTING INFORMATION TABLE Node Routing Information Table H HABDEF A HABDEF B HABDEF C Dead D HABDEF E HABDEF F HABDEF Figure 9. Routing map by PEGASIS. VI. SIMULATION RESULTS We make a comparative analysis of route structure,energy consumption and delay between PEGASIS andCRBCC. The parameters of experiment is the same intwo protocol tests. With energy radio model, we setk=2000bit, Eelec=50nJ/bit and ε amp =100pJ/bit/m2. InWSNs, 100 nodes place randomly in a 100m*100m areawith BS (50,300). We assume v=2Mbps and p=5%. We run the simulations for at least 100 times. At last,we make a conclusion that CRBCC shows a betterperformance in both energy efficency and low latencymetric than PEGASIS.A. Route Structure Comparison We can see that routing map by PEGASIS shows manycrosses, and many of these crosses are long-range. Butthere is no similar case in routing map by CRBCC. That Figure 10. Chain of 1st cluster in CRBCC.means in CRBCC, the transmission distances betweennodes are greatly minimized.© 2010 ACADEMY PUBLISHER
  6. 6. JOURNAL OF NETWORKS, VOL. 5, NO. 5, MAY 2010 591 Figure 11. Chain of 2nd cluster in CRBCC. Figure 14. Chain of 5th cluster in CRBCC. Figure 12. Chain of 3rd cluster in CRBCC. Figure 15. Top leader-leader chain in CRBCC. Figure 13. Chain of 4th cluster in CRBCC. Figure 16. Routing map by CRBCC.© 2010 ACADEMY PUBLISHER
  7. 7. 592 JOURNAL OF NETWORKS, VOL. 5, NO. 5, MAY 2010B. Energy Consumption Comparison TABLE III. DELAY COMPARISON (ONE ROUND) We can see that CRBCC spends less energy than Protocol PEGASIS CRBCCPEGASIS, which can prolong the lifetime of WSNs. It isnoted that one energy unit in y-axis is ε amp *k=2*10-7J. Delay(unit) 100 24This improvement is mainly achieved by SA chainalgorithm in CRBCC instead of greedy chain algorithm inPEGASIS. Figure 18. Delay comparison. Figure 17. Energy consumption comparison. VII. CONCLUSION This paper proposes a hierarchical routing approach forC. Delay Comparison WSNs in an energy and time constraint scenario. The creativity of the work is its clustering strategy and its We can see that CRBCC reduces the data delivery time chain routing algorithm in the whole routing protocol.greatly than PEGASIS. This advantage is achieved by Simulation results show that our protocol offers bettercoordinates-based clustering method in CRBCC. performance than PEGASIS in terms of energy Assume 1 unit time is needed for one node to transmit consumption and network delay. Later, we will make usek bit message with v bps to the neighbor node. On a of very fast simulated annealing algorithm (VFSA) and2Mbps link, a 2000 bit message can be transmitted in 1ms. more fine clustering method in the routing protocol ofTherefore, each unit of delay will correspond to about WSNs.1ms time for the case of a single channel and non-CDMA In the future, we will study the scheme aboutsensor nodes. clustering protocol based on tree routing algorithm. Tree delay(unit) = k / v  (11) routing algorithm has better performance than chain routing algorithm in both energy efficiency and latency The delay value will depend on the effective data rate metrics. Four tree routing schemes can be:for the CDMA sensor nodes. For example, as to the DS- (1)Scheme 1: Tree routing among all nodes in WSNs.CDMA system, suppose m=8 chips sequence per bit, then (2)Scheme 2: Clustering strategy + Direct transmission1 unit time is needed for one node to transmit k bit inside cluster + Tree routing among cluster-heads.message with v bps as (3)Scheme 3: Clustering strategy + Chain routing inside cluster + Tree routing among chain-leaders. delay(unit) = m * k / v (12) (4)Scheme 4: Clustering strategy + Tree routing inside In CRBCC, there only need 19 unit time to collect data cluster + Tree routing among cluster-heads.in the lowest chain for balanced clustering. After thatnodes need 4 unit time to collect information in the top ACKNOWLEDGMENTchain. At last, the top leader sends data to BS using 1 unit The project is supported by the Research Foundationtime. So, in total 24 unit time may require for of Education Bureau of Hubei Province, China (No.transmission in CRBCC. B20081506) and Youths Science Foundation of Wuhan But in PEGASIS, there is only one single chain. There Institute of Technology, China (No. Q200808).need 99 unit time to collect data in the chain and 1 unittime for chain leader to BS. Therefore, in total 100 unit REFERENCEStime of delay may occur. Moreover, latency advantage inCRBCC is more obvious in large-scale WSNs. [1] K. Akkaya and M. Younis, “A Survey of Routing Protocols in Wireless Sensor Networks”, Elsevier AdHoc Network Journal, Vol. 3/3 , 2005, pp. 325-349.© 2010 ACADEMY PUBLISHER
  8. 8. JOURNAL OF NETWORKS, VOL. 5, NO. 5, MAY 2010 593[2] I. F. Akyildiz et al., “Wireless Sensor Networks: A Survey”, Elsevier Sci. B. V. Comp. Networks, vol. 38, no. 4, Mar. 2002, pp. 393-422.[3] R. Szewczyk, J. Polastre, A. Mainwaring and D. Culler, “Lessons From Sensor Network Expedition”, 1st European Workshop on Wireless Sensor Networks (EWSN 04), Germany, Jan 19-21, 2004.[4] Jamal N. Al-Karaki and Ahmed E. Kamal, “Routing Gengsheng Zheng received the B.S. degree in Techniques in Wireless Sensor Networks: A Survey”, communication engineering from Xi’an Technology University, IEEE Wireless Communications, vol. 11, no. 6, Dec 2004. China, in 1994 and the M.S. and Ph.D. degrees in computer[5] K. Sohrabi, J. Gao, V. Ailawadhi, and G. J. Pottie, science from Wuhan University, China, in 1999 and 2006, “Protocols for self-organization of a wireless sensor respectively. Since January 2007, he has been an Associate network,” IEEE Pers. Commun., vol.7, no. 5, pp. 16–27, Professor in the School of Computer Science and Engineering, Oct. 2000. Wuhan Institute of Technology, China. His current research[6] Cohen, R. Kapchits, B. “An Optimal Algorithm for interests lie in the area of network and embedded system. Minimizing Energy Consumption while Limiting Maximum Delay in a Mesh Sensor Network”, Proceedings of IEEE International Conference on Computer Communications, INFOCOM, pp. 258–266, 2007.[7] C. Q. Hua and T. S. Yum, “Optimal routing and data aggregation for maximizing lifetime of wireless sensor networks”, Inf. Eng. Dept., Chinese Univ. of Hong Kong, Tech. Rep., 2005.[8] S. Kim, S. Park, S. Park, and S. Kim, “Energy Efficient Online Routing Algorithm for QoS-Sensitive Sensor Networks”, IEICE Transactions on Communications, vol. E91-B, no. 7,pp. 2401.2404, July 2008.[9] Choi, Dongmin; Moh, Sangman; Chung, Ilyong, “Variable area routing protocol in WSNs : A hybrid, energy-efficient approach”, Proceedings of 10th IEEE International Conference on High Performance Computing and Communications, HPCC 2008, p 397-403, 2008.[10] Rajiullah, Mohammad; Shimamoto, Shigeru, “An energy- aware periodical data gathering protocol using deterministic clustering in Wireless Sensor Networks (WSN)”, Proceedings of IEEE Wireless Communications and Networking Conference, WCNC, p 3016-3020, 2007.[11] Nahdia Tabassum, A K M Ahsanul Haque and Yoshiyori Urano, “GSEN: An efficient energy consumption routing scheme for wireless sensor network”, ICNICONSMCL06, pp. 117–122.[12] N. Tabassum, Q.E.K. Mamun and Y. Urano, “COSEN: A Chain Oriented Sensor Network for Efficient Data Collection”, Proceedings of the third International Conference on Information Technology: New Generations, ITNG 2006, v 2006, p 262-267, 2006.[13] W.R. Heinzelman, A. Chandrakasan, and H. Balakrishnan, “Energy-efficient Communication Protocols for Wireless Microsensor Networks”, Proceedings of the 33rd Hawaii International Conference on System Sciences, Jan. 2000.[14] W.B. Heinzelman, A. Chandrakasan, and H. Balakrishanan, “An Application-Specific Protocol Architechture for Wireless Microsensor Networks”, IEEE Trans. Wireless Commun., vol. 1, no. 4, Oct. 2002, pp. 660-70.[15] S. Lindsey and C. Raghavendra, “PEGASIS: Power- Efficient Gathering in Sensor Information Systems”, international Conf. on Communications, 2001.[16] Bohachevsky. Generalized simulated annealing for function opti2 mization[J] . Techwometrics , 1986, 28 (3): 209.[17] J.M. Zhang; Y.P. Lin; C.H. Zhou; J.C. Ouyang, “Optimal model for energy-efficient clustering in wireless sensor networks using global simulated annealing genetic algorithm”, Proceedings of 2nd 2008 International Symposium on Intelligent Information Technology Application Workshop, IITA 2008 Workshop, p 656-660, 2008.© 2010 ACADEMY PUBLISHER

×