Open Access Node Localization Method for Wireless Sensor Networks Based on Hybrid Optimization of Differential Evolution and Particle Swarm Algorithm

Transcription

1 Send Orders for Reprnts to The Open Automaton and Control Systems Journal, 014, 6, Open Access Node Localzaton Method for Wreless Sensor Networks Based on Hybrd Optmzaton of Dfferental Evoluton and Partcle Swarm Algorthm Lepng Zhang 1,,*, Wenjun J and Yu Zhang 1, 1 GuangX Key Laboratory of New Energy and Buldng Energy Savng, Guln Unversty of Technology, Guln 5404, Chna College of Mechancal and Control Engneerng, Guln Unversty of Technology, Guln 5404, Chna Abstract: Regardng the node localzaton problems for wreless sensor network, a hybrd optmzaton method was proposed accordngly on dfferental evoluton(de) algorthm and partcle swarm optmzaton(pso) algorthm. Frstly, the poston and velocty of the ntal populaton were randomly generated by PSO, and the ftness functon was constructed accordng to the mean square error of estmated and measured dstance between the unknown nodes and ther adjacent anchor node. Secondly, the mutaton and selecton operaton of DE algorthm were executed to fnd out the optmum poston of the populaton. Lastly, the current veloctes and postons of all partcles of the populaton were updated, and the crossover operaton and selecton operaton of DE algorthm were executed to update the current global optmum poston of the whole populaton. Populaton global optmum soluton of teratve search algorthm s the poston coordnate of the unknown node. Smulaton results ndcate that the proposed localzaton method has smaller average localzaton error and hgher localzaton accuracy than that of DE algorthm and PSO algorthm n the same envronment. Keywords: Wreless sensor network, Dfferental evoluton algorthm, Partcle swarm optmzaton algorthm, Hybrd optmzaton, Node localzaton. 1. INTRODUCTION Poston nformaton s of great mportance to wreless sensor network (WSN) montorng actvtes. Event occurrng place or node poston of acqured nformaton accounts for the mportant nformaton that contans n wreless sensor node montorng nformaton [1]. In nature WSN localzaton s equvalent to optmzng dfferent measured dstances or routes, so t s a NP unsolvable problem []. In the meanwhle, as node localzaton method based on optmzaton algorthm has the features of non-dependence on the adopted localzaton method, hgh localzaton accuracy, etc., researchers proposed the method to mprove the performance of WSN node localzaton by usng the optmzaton algorthm. Ewa Newadomska-Szynkewcz and others made a study over WSN node localzaton method by adoptng heurstc algorthm, such as smulated annealng and genetc algorthm, whch n turn proves the node localzaton effectveness of heurstc algorthm [3]. Chehr A. and others proposed a non-lnear optmzaton localzaton algorthm based on dfferental evoluton(de) algorthm [4], whch has the advantage of low complexty. But the method has the advantages of low localzaton accuracy f the measured dstance error s bg, node wreless range s small or anchor node rato s low. Wulng Ren and others proposed a localzaton algorthm based on shuffled frog leapng algorthm and partcle swarm optmzaton(pso)algorthm [5], whch has a good localzaton accuracy. The method has a good localzaton effect, notably f the measured dstance error s bg enough, but ts consumpton and stablty reman to be mproved. A new localzaton algorthm for WSN by combnng extremum dsturbed and smple partcle swarm optmzaton was proposed to solve the dstance estmaton errors problem of DV-hop algorthm [6]. A group partcle swarm optmzaton algorthm was proposed and used to optmze the result of the maxmum lkelhood estmaton method based on tme of arrval(toa) [7]. DE algorthm s one algorthm based on populaton evoluton proposed by R. Storn and K. Prce [8]. It offers soluton to the optmum queston through cooperaton and competton among populaton ndvduals and t has the advantages of smplfcaton, hgh-effcency, hgh convergence rate, good robustness, and so on. PSO algorthm s a knd of self-adaptve global optmzaton heurstc algorthm that has the advantages of smple algorthm, easy adopton, fast search,and son on [9]. However, as a sngle algorthm,de algorthm s also lmted to such problems as local convergence, slow late convergence rate and unstable performance when solves complcated optmzaton queston, and PSO algorthm s lmted to confned search range, vulnerablty to local optmzaton and other problems []. Obvously PSO algorthm wth good optmzaton and DE algorthm wth strong strength n mantanng populaton dversfcaton and searchng ablty are well complementary to each other [11]. Therefor, a hybrd-optmzed WSN node localzaton method based on DE algorthm and PSO algo / Bentham Open

2 6 The Open Automaton and Control Systems Journal, 014, Volume 6 Zhang et al. rthm s proposed n ths paper,and ts effectveness s proved through smulaton experments.. DESCRIPTION OF LOCALIZATION PROBLEMS WSN node localzaton problem s actually the process that poston coordnates of unknown nodes are acheved by the dstance nformaton, wreless communcatons radus accordng to some localzaton strategy [1]. Snce there s an error n dstance measurement technology, measured dstance between anchor node and unknown node s not the real value. Therefor, poston estmaton of unknown node can be treated as a knd of optmzaton whch mnmzes the target functon of localzaton error of unknown node to fnd out the poston coordnate of unknown node. The rangng error s an mportant factor affectng the unknown node locaton error, and reduce the maxmum error can effectvely mprove the localzaton accuracy. The mean square error of dstance between unknown node and adjacent anchor node s defned as the ftness functon of localzaton problem. As shown below: f ( x, y) = d ( x x) + ( y y ) (1) Wheren, ( x, y ) s the estmated coordnate of unknown node, x, y ) (=1,,,M) s the actual coordnate of the th ( anchor node, ( x x) + ( y y) s the estmated dstance between the th anchor node and that unknown node, f ( x, y ) s the error value between measured and estmated dstance of the th anchor node and that unknown node, and d s the measured dstance between the th anchor node and that unknown node. As a matter of fact, measured dstance between two nodes s not the actual dstance n realty, so the combned form of actual dstance and gaussan error should be used for measurng dstance [4], namely: d = d j (1+ randn! ") () d j Wheren,d j s the actual value between two nodes, = x x ) + ( y y ), η s the error factor related to ( j j dstance measurement accuracy, randn s the random varable subject to standard normal dstrbuton, where the average value s 0 and square varance s 1. In ths way, WSN node localzaton problem s converted to the ftness functon optmzaton problem and the optmum soluton provded by mnmzng ftness functon s therefore the estmated poston of unknown node. The mprovng localzaton error ftness functon can be shown as below, that the weghted poston nformaton of nearest hop dstance between anchor node and the unknown node are ntroduced to the prmary ftness functon. As shown below: M!! F (x, y) = " f!! # (x, y) (3) =1 Wheren, M s the number of nearest hop dstance of anchor node to the unknown node, and n ths paper, M=4,whch s the mnmum number of M. α s the accuracy weght of measured dstance between unknown node and the th anchor node, whch s n reverse proporton wth the shortest route hop count between unknown node and the th anchor node. The hop count can be calculated by Djkstra algorthm. Improved ftness functon can effectvely reduce algorthm complexty, runnng tme and node energy consumpton but extend network lfetme. Optmzaton algorthm can be adopted to optmze the ftness functon, fnd out the optmum soluton and mprove node poston accuracy. 3. DESCRIPTION OF NODE LOCALIZATION METHOD 3.1. DE Algorthm Smlar to genetc algorthm, DE algorthm repeats teratve computaton from one randomly generated ntal swarm n accordance wth some operaton rules, such as selecton, hybrdzaton and mutaton, and gudes the search close to the optmum soluton by survvng good ndvduals whle excludng the bad ones n lne wth every ndvdual ftness value. However, dfferent from genetc algorthm, DE algorthm uses dfferental strategy n operatng mutaton ntended for ndvdual mutaton concretely through dsturbng ndvduals by use of dfferental vectors among populaton ndvduals. DE algorthm can be extended as follows [13]. Step 1: Intalzed populaton. Defne populaton scale as NP and maxmum teraton as t max, and randomly generate the ntal populaton meetng restrant condton n D- dmensonal space. In whch, =1,,,NP, j=1,,,d, rand j (0,1) stands for random number evenly dstrbuted n the nterval (0,1). b j,, U b j, respectvely stands for the L X j, upper and bottom lmt. As shown below: X j,,0 = rand j (0,1)! (b j,u " b j,l ) + b j,l (4) Step : Mutaton operaton. Mutaton s the key step of DE algorthm and the common strategy of DE algorthm s ntended for changng through mutaton operaton. The ndvdual generaton strategy of DE algorthm can be expressed as shown: DE/x/y/z. Wheren, x stands for mutated ndvdual vector type, y stands for dfferental vector quantty and z stands for crossover method. In ths paper, DE/rand/1/bn strategy s used n the mutaton operaton,whch s a wdely used combnaton strategy. Mutaton operaton for Gth generaton evolved target vector NP]) s shown as below: X, G ( [1, V,G+1 = X r1,g + F! ( X r,g " X r3,g ) (5) j,, G+ 1 Step 3: Crossover operaton. Mutaton vector wll randomly change wth the target vector. when a new vector experment vector U would be output after crossover. Target vector X and mutaton vector V n teraton j are defned n the followng equaton. As shown below:

3 Node Localzaton Method for Wreless Sensor Networks The Open Automaton and Control Systems Journal, 014, Volume 6 63 # % V j,,g+1 f (rand! CR) or( j = rnbr()) U j,,g+1 = $ X j,,g+1 f (rand > CR)and( j " rnbr()) & % Step 4: Selecton operaton. Through the above operatons combned comparng ftness values of the experment vector and target vector, the populaton s therefore selected to be passed on to the next generaton. As shown below: " $ U,G+1 f F(U,G+1 )! F( X,G ) X,G+1 = # X,G otherwse % $ 3.. Partcle Swarm Optmzaton Algorthm PSO algorthm takes every ndvdual as partcle wthout any volume and weght n the search space. Every partcle wthn the search space has ts own poston and velocty as well as ftness value determned by optmzaton functon through regulatng operaton rules for each partcle. Partcle fles at one velocty n the search space, durng whch the optmum poston the partcle once passed s the optmum soluton searched by ths partcle. All partcles are searchng wthn the solutons space of the optmum partcle, and gradually the optmum soluton s searched out through teratve search, or called as ndvdual best value, namely P best. The best poston that the entre populaton (or some partcles) once passed s the optmum soluton found out to the entre populaton (or some partcles), or called as global (or local) best value, namely G best. Partcle wll be updated wth ndvdual best value P best and global (or local) best value G best durng every teraton. In the populaton composed of m partcles wthn the n-dmensonal space, the th partcle X wll change ts own velocty and poston based on equaton (8) and (9). V d (t +1) = w!v d (t) + C 1! rand()! (P d " X d ) + C! rand()! (P gd " X d ) X d (t +1) = X d (t) +V d (t +1) (9) Among whch, 1!! m,1! d! n and t s the number of teraton for current populaton, and w s the nerta weght, C 1 and C are postve constants called as learnng factors, and rand() s the random number evenly dstrbuted n the nterval [0,1] WSN Node Localzaton Method Based on Hybrd Optmzaton of Dfferental Evoluton and Partcle Swarm Algorthm DE algorthm and PSO algorthm are both random heurstc algorthms based on swarm ntellgence. The evoluton process wth randomness would result n some blndness n search for the optmum soluton. Notably n offerng soluton to complex optmzaton queston, DE Algorthm s lmted to the convergence rate of solvng complcated optmzaton n the late optmzaton teraton, whch would easly result n local optmzaton. As for PSO algorthm, premature (6) (7) (8) convergence appearng as a result of reduced partcle dversfcaton would also easly result n local optmzaton. In order to cover defects for both DE algorthm and PSO algorthm, a hybrd optmzaton algorthm of dfferental evoluton and partcle swarm algorthm (DEPSO) would be proposed through an effectve combnaton of DE algorthm and PSO Algorthm. Ths algorthm adds dsturbance on the current poston of PSO partcle by use of DE algorthm mutaton operaton for the beneft of mantanng swarm dversfcaton, mprovng PSO algorthm space search capacty and avodng partcle nto local optmzaton. As hybrd optmzaton DEPSO algorthm features better global convergence, t would mprove node localzaton accuracy by applyng t to WSN node localzaton. In ths way, ths paper ntends to propose a WSN node localzaton method based on DEPSO algorthm. Frstly, t randomly generates the ntal swarm, ntalzes all ndvdual postons and veloctes, calculates and stores ndvdual best value P best as well as global optmum soluton G best, executes mutaton and selecton operaton for dfferental evoluton, and eventually fnds out the swarm optmum poston. Secondly, t renews veloctes and postons of all partcles n the swarm by usng PSO velocty and poston equaton. Thrdly, t renews the swarm optmum poston by use of crossover and selecton operaton of DE algorthm. Lastly, t outputs the global optmum soluton for swarm untl all teraton termnaton condtons are met or maxmum teratons are reached. The poston of partcle n the search space through PSO algorthm s the poston of unknown node, swarm global optmum soluton renewed through algorthm teraton s the partcle optmum poston, or estmated poston coordnates for the unknown node, and the estmaton process of unknown node poston s the process of mnmzng localzaton error ftness functon. The node localzaton desgn mentoned n ths paper s based on DEPSO hybrd algorthm, whch s specfed as follows: Step 1: Intal parameter setup for algorthm: swarm scale NP, maxmum teratons t max, scalng factor F, crossover rate CR, maxmum velocty v max, learnng factor C 1 and C, ntal nerta weght w max, fnal nerta weght w mn, teratons t=0, and error factor η. Step : Intalzed swarm: randomly deploy N unknown nodes and M anchor nodes wthn the specfc network area, randomly generate the ntal poston and velocty for each partcle, calculate the ftness value for each partcle and set the current partcle poston as ndvdual best value P best, among whch the best s kept as global best value G best. Step 3: Calculate the dstance matrx D as composed of the respectve dstance between N unknown nodes to M anchor nodes. Step 4: Calculate the shortest route and hops between each unknown node and anchor node by Djkstra algorthm, and fnd out the dstance matrx d and ts hops N hop between the unknown node and ts 4 nearest anchor nodes. It can be converted to fnd out the node poston [ x, y ] that mnmzes the ftness functon by use of DEPSO algorthm, equvalent! to solve [x!, y! ] = argmn(f (x!, y ))

4 64 The Open Automaton and Control Systems Journal, 014, Volume 6 Zhang et al Localzaton error(m) Iteraton tmes Fg. (1). When error factor η=5%, the nfluence of teraton tmes on localzaton error. Step 5: Parameter nput: anchor node coordnate matrx Beacon, dstance matrx d, and weght α I. Step 6: Iteratons t=t+1. Step 7: Proceed on mutaton and selecton operaton for the swarm n lne wth equaton (6) and equaton (8). Step 8: Calculate ftness functon value for each ndvdual by use of equaton (4), and compare t wth the optmum soluton to current hstorc swarm, among whch the ftness functon wth the smallest value shall be passed to the next generaton, equvalent to renew the optmum poston for current swarm. Step 9: Renew the postons and veloctes of all ndvduals n the swarm n lne wth equaton (9) and equaton (). Step : Execute crossover and selecton operaton for each ndvdual poston n lne wth equaton (7) and equaton (8), recalculate the ftness values for all ndvduals and compare correspondng ftness value for current optmum poston and the ftness value for swarm hstorc optmum poston, among whch the ndvdual wth the smallest ftness value shall be passed to the next generaton, namely ths ndvdual poston s the optmum poston of the current swarm. Step 11: Judge f the maxmum teraton t max s reached or not, f so, t outputs the correspondng ndvdual poston [ x, y ] wth global optmum soluton, or the estmated coordnate for unknown node poston, or otherwse return to Step SIMULATION EXPERIMENT AND ITS RESULT ANALYSIS 4.1. Setup of Smulaton Envronment Smulaton experment s carred out n MATLAB. Set the total network nodes as 0 randomly dstrbuted n 0m 0m square area, where unknown nodes and anchor nodes are randomly generated and deployed. Through experence and many repeated experments to determne the smulaton parameter by DEPSO localzaton method: node unlmted range R=40m, swarm scale NP=0, maxmum teratons tmax=0, maxmum velocty v max =6, ntal nerta weght w max = 0.9, fnal nerta weght w mn =0.4, C 1 = C =, scalng factor F=0.5, crossover rate CR=0.6, and let error factor as η=5%. Compare average localzaton error from the perspectve of algorthm teratons, anchor node densty, dstance measurement error and anchor wreless range by employng three methods DE, PSO and DEPSO through 0 smulaton experments. 4.. Smulaton Result and tts Analyss DEPSO localzaton method has been proved to be a knd of localzaton method wth hgh localzaton accuracy that only needs a few anchor nodes and smaller node wreless range by correspondngly changng teratons, anchor node densty, network connectvty and dstance measurng error. Compare DEPSO localzaton method wth DE localzaton method and PSO localzaton method for analyzng DEPSO localzaton method performance. The average localzaton error equaton used to evaluate DEPSO localzaton method performance n smulaton s shown as below: error = 0 N # (x N! R " x) # $ + ( y " y ) % () =1 Among whch, R s node communcatons radus, N s the number of unknown nodes, ( x, y ) s the estmated poston of the unknown node, and ( x, y ) s the actual poston of the unknown node. Algorthm convergence s the key to algorthm effectveness. Fg. (1) shows the correlaton between teraton and localzaton error by DEPSO hybrd optmzaton method when the error factor η=5%. It follows that the localzaton

5 Node Localzaton Method for Wreless Sensor Networks The Open Automaton and Control Systems Journal, 014, Volume DE PSO DEPSO Average localzaton error(%) Anchor node densty(%) Fg. (). The nfluence of anchor node densty on average localzaton error. error by DEPSO localzaton method also contnuously decreases wth the ncrease of teratons f there s dstance measurement error. When the teratons reach 0, localzaton error curve tends to be flat wth the localzaton error convergng approxmately at 1.0m. Consequently, DEPSO method featurng strong convergence can be used to accurately estmate the poston of unknown node. The proporton of anchor nodes s another mportant ndex for evaluatng localzaton effects, because the number of anchor nodes drectly determnes the consumpton. In smulaton experments, the anchor node densty ncrease from % to 30% at 5% nterval f other parameters stay unchanged; the nfluence of changng anchor node densty upon the average localzaton error usng all localzaton methods s observed. Fg. () shows the nfluence of anchor node densty on average localzaton error. Clearly, the average localzaton errors by use of these three methods decrease wth the ncrease of anchor node densty, but DEPSO hybrd optmzaton method has hgher localzaton accuracy at the same anchor node densty or wth the same anchor nodes. In other words, anchor nodes quantty used by DEPSO localzaton method s the lowest under the requrement of the same localzaton accuracy. However the quantty of anchor nodes drectly nfluences the network cost and energy consumpton, the overall performance of DEPSO localzaton method featurng the lowest cost and consumpton s better than other two localzaton methods. Network connectvty refers to the quantty of adjacent nodes of unknown nodes. Generally, strong network connectvty wll brng about more localzaton nformaton used by nodes, thus mprovng node localzaton accuracy and node localzaton rate. Smulaton could change network connectvty by changng the wreless range of nodes. Therefore the wreless range of anchor nodes s used as one performance ndex to evaluate localzaton error. In smulaton experment, network connectvty s changed wth the change of anchor node wreless range R f node-deployng area and other parameters stay unchanged. Ths paper observes the localzaton error by settng anchor node wreless range R ncrease from 0m to 45m at 5m nterval usng all knds of methods. Fg. (3) compares the nfluence of anchor node rado range R on the average localzaton error respectvely based on DE, PSO and DEPSO localzaton method. From whch the average localzaton errors based on these three methods nvarably decrease wth the ncrease of anchor node wreless range.in the same anchor node rado range, the localzaton error based on DEPSO algorthm s the lowest whle node rado range corresponds to node transmttng power. Consequently, DEPSO localzaton method could be used to satsfy localzaton accuracy and extend network lfetme wth a slower varaton tendency of error curve, thus verfyng ts better stablty and robustness than other two methods. In vew of the naccuracy regardng dsturbance and dstance measurng technology n the external envronment, there s unavodable measured dstance error between anchor node and unknown node. In smulaton experment, the proposed localzaton method and relevant performance are compared through changng error factor η when other parameters stay unchanged. Fg. (4) shows the nfluence of dstance error on average localzaton error when the error factor η ncreases from 0% to 0% at 5% hop length. From t the average localzaton errors of three methods respectvely ncrease wth the ncrease of dstance measurement error, but DEPSO algorthm has better performance than other two methods. Wth the same dstance measurement error, the

6 66 The Open Automaton and Control Systems Journal, 014, Volume 6 Zhang et al DE PSO DEPSO Average localzaton error(%) Anchor node rado range(m) Fg. (3). The nfluence of anchor node rado range on average localzaton error. Average localzaton error(%) DE PSO DEPSO Fg. (4). The nfluence of dstance error on average localzaton error Dstance error(%) localzaton error based on DEPSO algorthm s the lowest, and ts average localzaton error curve slope s also the lowest among these three methods. In other words, ths localzaton method wth better localzaton effect receves the mnmal nfluence of dstance measurement error. For the convenence of comparng three localzaton effects, the unknown node localzaton effects by three methods are put under the unform ordnate scale for comparson as shown n Fg. (5), under the condton that the ntal chosen anchor node densty s 0%, and network unknown nodes amounts to 80. From t the error curve fluctuaton for unknown nodes based on PSO algorthm s much weaker than that of DE algorthm, provng PSO algorthm has better stablty than DE algorthm. However, both DE method and PSO method has the sngle case of bgger node localzaton error, provng local mnmal value when localzaton at ths node. Among three localzaton error varaton curve, DEPSO localzaton method features lowest error, and ts

7 Node Localzaton Method for Wreless Sensor Networks The Open Automaton and Control Systems Journal, 014, Volume DE PSO DEPSO Localzaton error(m) Node number Fg. (5). The comparson of node localzaton errors of three localzaton methods. fluctuaton range and varaton tendency of localzaton error curve are lower than that of other two methods. Although there s sngle case of bgger localzaton error, t nearly can be gnored when compared wth the above two localzaton methods. Overall speakng, the localzaton method proposed n ths paper featurng hgher localzaton accuracy and better global convergence would effectvely prevent the occurrence of local optmum soluton. CONCLUSION Ths paper ntends to propose a hybrd-optmzed method for WSN node localzaton by adoptng an effectve strategy and analyzng the respectve characterstcs of DE algorthm and PSO algorthm. Reducng the square error of estmated and measured dstance between the unknown node and ts adjacent anchor node can guarantee a better localzaton accuracy. Smulaton results prove that ths method features smaller localzaton error, hgher localzaton accuracy and better stablty performance, but t also requres more calculaton quantty whle mprovng the accuracy. In ths concern, future researches wll focus on how to mnmze the node energy consumpton and extend network lfetme. CONFLICT OF INTEREST The author confrms that ths artcle content has no conflct of nterest. ACKNOWLEDGEMENTS Ths work was supported by the key project of Guangx educatonal scence foundaton (No.0ZD051), the research fund of Guangx key laboratory of new energy and buldng energy savng (No ), and the doctoral research ntal foundaton of Guln Unversty of Technology. REFERENCES [1] P. K. Sngh, B. Trpath, and N. P. Sngh, Node Localzaton n Wreless Sensor Networks, Internatonal Journal of Computer Scence and Informaton Technologes, vol., no. 6, pp , 011. [] J. Aspnes, D. Goldenberg, and Y. R. Yang, On the Computatonal Complexty of Sensor Network Localzaton, In: Proceedngs of Frst Internatonal Workshop on Algorthmc Aspects of Wreless Sensor Networks, pp. 3-44, 004. [3] E. N. Szynkewcz, M. Marks, and M. Kamola, Localzaton n wreless sensor networks usng heurstc optmzaton technques, Jouranl of Telecommuncaton and Informaton Technology, vol. 4, pp , 011. [4] A. Chehr, P. Forter, and P. M. Tardf, Geo-locaton wth wreless sensor networks usng non-lnear optmzaton, Internatonal Journal of Computer Scence and Network Securty, vol. 8, no. 1, pp , 008. [5] W. Ren, and C. Zhao, A localzaton algorthm based on SFLA and PSO for wreless sensor network, Informaton Technology Journal, vol. 1, no. 3, pp , 013. [6] Q. Zhang, and M. Cheng, A node localzaton algorthm for wreless sensor network based on mproved partcle swarm optmzaton, Lecture Notes n Electrcal Engneerng, vol. 37, pp , 014. [7] D. Feng, and X. Jang, WSN node localzaton technology research based on GPSO, Computer Smulaton, vol. 31, no., pp , 014. [8] R. Storn, and K. Prce, Dfferental evoluton-a smple and effcent heurstc for global optmzaton over contnuous spaces, Journal of Global Optmzaton, vol. 11, pp , [9] J. Kennedy, and R. C. Eberhart, Partcle Swarm Optmzaton, In: Proceedngs of IEEE Internatonal Conference on Neural Networks, , 1995 [] Y. Ch, J. Fang, and G. Ca, Hybrd optmzaton algorthm based on dfferental evoluton and partcle swarm optmzaton, Computer Engneerng and Desgn, vol. 30, no. 1, pp , 009.

8 68 The Open Automaton and Control Systems Journal, 014, Volume 6 Zhang et al. [11] B. Xn, and J. Chen, A survey and taxonomy on hybrd algorthms based on partcle swarm optmzaton and dfferental evoluton, Journal of System Scence and Mathematcal Scence, vol. 31, no. 9, pp , 011. [1] S. Zhao, M. Sun, and Y. Tang, GASA-Based localzaton algorthm for wreless sensor networks, Computer Applcatons and Software, vol. 6, no., pp , 009. [13] A. K. Qn, V. L. Huang, and P. N. Suganthan, Dfferental evoluton algorthm wth strategy adaptaton for global numercal optmzaton, IEEE Transactons on Evolutonary Computaton, vol. 13, no., pp , 009. Receved: November 4, 014 Revsed: January 07, 015 Accepted: January 19, 015 Zhang et al.; Lcensee Bentham Open. Ths s an open access artcle lcensed under the terms of the Creatve Commons Attrbuton Non-Commercal Lcense ( whch permts unrestrcted, non-commercal use, dstrbuton and reproducton n any medum, provded the work s properly cted.