Multi-Objective Cross-Layer Optimization for Selection of Cooperative Path Pairs in Multihop Wireless Ad hoc Networks

Transcription

1 7 JOURNAL OF COMMUNICATIONS SOFTWARE AND SYSTEMS, VOL. 9, NO. 3, SEPTEMBER 3 Muli-Objecive Cross-Layer Opimizaion for Selecion of Cooperaive Pah Pairs in Mulihop Wireless Ad hoc Neworks Nyoman Gunanara, and Gamanyo Hendranoro Original scienific paper Absrac This paper focuses in he selecion of an opimal pah pair for cooperaive diversiy based on cross-layer opimizaion in mulihop wireless ad hoc neworks. Cross-layer performance indicaors, including power consumpion, signal-o-noise raio, and load variance are opimized using muli-objecive opimizaion MOO wih Pareo mehod. Consequenly, opimizaion can be performed simulaneously o obain a compromise among hree resources over all possible pah pairs. The Pareo mehod is furher compared o he scalarizaion mehod in achieving fairness o each resource. We examine he saisics of power consumpion, SNR, and load variance for boh mehods hrough simulaions. In addiion, he complexiy of he opimizaion of boh mehods is evaluaed based on he required compuing ime. Index erms muli-objecive opimizaion, Pareo mehod, scalarizaion mehod, selecion of he pah pair, mulihop wireless ad hoc neworks I. INTRODUCTION An ad hoc nework is a collecion of nodes ha communicae dynamically wihou a fixed infrasrucure. Each node can ac as a source, relay, and desinaion. The nodes have limiaions in erms of ransmission range and baery capaciy []. To overcome aforemenioned limiaions, i requires cooperaive communicaion echniques. Cooperaive communicaion is a sysem he source nodes cooperae and coordinae wih he nodes funcioned as relay before reaching he desinaion node o improve ransmission qualiy. Cooperaive communicaion using a single anenna in mulinode scenario can make beneficial use of anenna from each node so ha i can creae muliple anenna communicaion sysems such as he muli inpu muli oupu MIMO []. Selecion of nodes ha will ac as relays is a problem ha mus be solved by considering several crieria. In [] and [3-6] relay selecion is based only on he resources a he physical layer. Selecion of relay ha mees arges and consrains on muliple layers need o ake ino accoun he resources in he higher layers, so i is necessary o apply a cross-layer opimizaion [7-]. If he opimizaion problem involves he compromise of more han one resource, some of hem Manuscrip received April 3, 3, revised Sepember, 3. Auhors are wih he Deparmen of Innovaion Engineering a Insiu Teknologi Sepuluh NopemberITS, Indonesia, gunanara, are conradicory, i is needed o apply MOO muli-objecive opimizaion []. The applicaion of MOO o opimize wireless neworks in [-3] is solved by scalarizaion. However, he problem of resources opimizaion can no be done separaely because he problems are iner-relaed wih each oher. An alernaive o overcome his weakness is he Pareo mehod. Runser e al [] is one of he firs o apply he Pareo mehod in solving MOO problems in wireless ad hoc neworks. The resul is a radeoff characerisic of hree parameers, namely robusness, energy consumpion and delay for hop ad hoc neworks. Gunanara and Hendranoro [] furher develop opimal relay selecion for single mulihop pahs based on cross-layer opimizaion for power consumpion, hroughpu, and load variance. In [], he work deals wih finding he opimum single pah wih muliple hops, as he problem a hand is on finding a pair of mulihop pahs ha is opimum for cooperaive diversiy applicaions. This paper is moivaed by hose resuls, as well as o address he limiaions of he sudy in [9] for wireless neworks wih relays energy efficiency and load balance can no be achieved a he same ime. To deermine he performance of Pareo mehod, we compare i wih he scalarizaion mehod. izaion mehod has been applied on he manipulaor each resource is normalized by he sandard deviaion mehod [6] and he prioriy mehod [7]. Normalizaion using sandard deviaion and prioriy mehod ends o separae prioriized objecs and ignore oher objecs. In his sudy, each objec is given equal weigh and normalized by he square roo of average power of he performance indicaor quaniy. Normalizaion is used o provide a sense of fairness among he objecives. The main conribuion of his paper is, firsly, he opimizaion mehod for ad hoc nework model ha is dynamic ha can be done simulaneously for all opimized resources based on pah in order o obain an opimal pair of pahs wih he help from MOO wih Pareo mehod. Secondly, i describes scalarizaion mehod wih fairness for all hree resources. Thirdly, his paper describes he complexiy of boh mehods of opimizaion and also o obain cumulaive value for all hree resources. Secion II of his paper gives a descripion of ad hoc neworks, radio propagaion, and MOO. Secion III describes he model configuraion, parameer simulaion, and analysis of simulaion resuls, wih conclusions given in Par IV. 8-6/9/89 3 CCIS

2 GUANANTARA AND HENDRANTORO MILTI-OBJECTIVE CROSS-LAYER OPTIMIZATION II. COOPERATIVE COMMUNICATIONS Frequency Hz Cooperaive communicaion can be explained by graph, { } is he se of nodes and { } is he se of links/hops. In mulihop ad-hoc neworks, here are pairs of source and desinaion node ha communicae by involving oher nodes as relays o form mulihop pahs. If he oal number of nodes including he source and desinaion pair is N, hen here is one single-hop soluion, -hop soluions, 3-hop soluions, -hop soluions, and so on, for he source and desinaion pairs. In his sudy, he maximum number of hops o be considered for one pah is limied o hree. From he se of pahs wih hree hops maximum, here are several possible combinaions ha form a pair of pahs beween he source and desinaion. Suppose denoes he se of all pah pairs having and hops, saes permuaions of ou of, and specifies he number of pah pairs in he se. The number of combinaions can be obained such as = soluions consising of wo pahs, each wih one and wo hops = for each pah, soluions consising of wo pahs, each wih one and hree hops, R, = N- N-3 soluions consising of a pair of pahs, each wih soluions consising of wo hops, = wo pahs, each wih wo and hree hops, and = soluions wih a pair each having hree hops. A he receiver, he signals received from he seleced pair are combined wih maximal raio combining MRC. Broadcas rouing is assumed using amplify-and-forward AF relays, he source sends he informaion o all nodes poenial o be relays, so ha informaion can arrive a he desinaion [8]. Broadcas rouing is seleced so ha he ransmied daa can be received by all adjacen nodes simulaneously o save ransmission ime. 7 is used as in []. Each pah uses a differen sub-carrier, as each hop in a pah uses a differen ime slo. Fig. illusraes an example of frequency/sub-carrier ime slo division for wo pahs, namely pah --3 consising of wo hops and pah ha consiss of hree hops. III. PROBLEM FORMULATION A. Radio Propagaion A. Oudoor I is assumed ha he ransmi power for all nodes is idenical and gain of he ransmier and receiver anenna, and are he same. Therefore he received power hrough a wireless hop of lengh meers can be calculaed by he following equaion [] denoes shadowing loss db which is normally disribued wih a sandard deviaion of. A. Indoor In indoor condiion, he nodes in an ad-hoc nework are well posiioned in rooms separaed by walls. The walls can cause parial reflecion of he ransmied signal so ha only some porion of he energy is ransmied hrough he wall, which is represened by a ransmission coefficien []. Power received a a node from anoher node in a differen room via a link/hop can be deermined using by inroducing he influence of he ransmission coefficien and respecively denoes he ransmission coefficien of he m-h wall ha is passed by he direc propagaion pah and he number of walls Time s Fig.. OFDMA Mehod for Pah --3 and The proocol mechanism of he sysem model can be described as follows - The source can idenify he desinaion posiion by each node deecing oher nodes conneced direcly via a single hop and sending informaion o all nodes wihin one hop [9]. - To avoid inerference and collisions among nodes, OFDMA orhogonal frequency division muliple access B. Performance Indicaors of Cooperaive Communicaions To opimize he performance of cooperaive communicaions, funcion or duy of each communicaion layer needs o be adaped by including he parameers and crieria on more han one layer of he archiecure of he communicaion sysem. This is known as cross-layer opimizaion. The purpose of cross-layer opimizaion depends on he quaniy of he layer o be made adapive. In his paper, he layers of ineres are he physical and he nework layer. The following describes he parameers of hese wo layers. B. Power Consumpion Power consumpion on pah is overall power requiremens needed in ransmiing daa from he source o desinaion hrough muliple relays in each pah. If i is assumed ha all

3 7 JOURNAL OF COMMUNICATIONS SOFTWARE AND SYSTEMS, VOL. 9, NO. 3, SEPTEMBER 3 nodes have he same ransmission power, hen power consumpion in he p-h pah consising of hops are 3 While he amoun of power consumpion for pah pair is obained from he following equaion 9, and denoe he power consumpion of he pah wih hops, he pah wih hops, and he pair of pahs wih and hops, respecively. The opimal pah pair is hus he one wih he smalles value of power consumpion pah pair. B.3 Load Variance Load variance is he variance of raffic load over all nodes, which is inversely proporional o he load balance or fairness [3]. In wireless ad hoc neworks, load balance is very imporan because some node may have greaer opporuniy o be chosen as a relay when energy consumpion alone is considered, bu migh no be so when he raffic load i carries is aken ino accoun. In a pah pair, node i is used as a relay, he load of node becomes wih and respecively denoing is own raffic load and he incoming raffic load ino node. Afer he load of each node is known hen he variance of raffic load of nodes in he whole nework can be evaluaed for each possible pah pair wih he following equaion [3] represens he power consumpion of he opimal Based on variances obained for all possible pah pairs, he opimal pah pair in erms of load fairness can be deermined by finding one wih he lowes raffic load variance B. Signal-o-Noise Power Raio SNR a each hop is he raio beween he received power, wih he noise power a he node, represens noise power assumed idenical for all nodes. I is assumed ha each relay does amplify and forward, so ha he overall SNR on a pah depends on he SNR of each hop [] denoes he load variance of he nework wih he opimal pah pair and denoes he load variance obained for a pah pair wih and hop. 6 IV. MULTI-OBJECTIVE OPTIMIZATION wih is he value of SNR a he -h hop. The SNR for a pah pair afer maximal-raio combining is obained from he following equaion 7,, and represen SNR of he pah wih hops, ha of he pah wih hops, and ha of he maximalraio combined pahs wih and hop. Mehods o solve MOO problems can be classified ino wo, Pareo and scalarizaion []. The following describes each of hese mehods. A. Pareo Mehod Opimizaion is he process of finding he bes soluion of a problem. For issues ha conradic each oher, such as he problems of smalles power consumpion and he larges SNR, Pareo mehod can be used in searching he bes soluion. Mahemaically, hree issues in secion III can be wrien as follows [] subjec o wih denoes he SNR of he opimal pah pair. 8 For an ad hoc nework, he opimal pair of pahs is he one giving he maximum value of SNR among all pah pairs deermined by he following equaion, represens he number of cooperaive pahs and indicaes ha he pahs consiuing a cooperaive pah pair canno share any hop.

4 GUANANTARA AND HENDRANTORO MILTI-OBJECTIVE CROSS-LAYER OPTIMIZATION Pareo opimizaion mehod mainains he soluions of boh problems in he Pareo Opimal Fron POF apar during opimizaion. In POF, here is he dominance concep o disinguish he dominaed inferior and he non-dominaed soluion non-inferior. For he opimizaion of wo problems, non-dominaed soluion can be described on a POF plane wo dimensions, as illusraed in Fig. for wo problems Z and Z [6]. As for he opimizaion of hree problems, nondominaed soluion can be described in POF surface hree dimensions [7]. l Z Non dominaed soluion Dominaed soluion POF of fairness all objecives are given equal weigh and are each normalized by is square roo of average power SRAP. For example, SNR is normalized by he SRAP of SNR, which simply can be seen in he denominaor of equaion, namely. izaion of he hree objecives becomes V. NUMERICAL RESULTS Z Fig.. POF for Two Objecives In searching for he opimal value of a POF, he uopia poin should be se firs. For he case involving wo objecive funcions ha should be minimized and maximized, respecively, he uopia poin is he inersecion of he minimum value of he firs objecive funcion and he maximum value of he oher. The opimal value can be deermined by finding he shores Euclidian disance [8] by equaion [9] A. Model Configuraion We review ad-hoc neworks in wo condiions, i.e. oudoor and indoor. Resuls discussed in his and he nex par are aken from one ou of configuraions generaed wih randomly posiioned nodes in our simulaions. The exemplary configuraion can be seen in Figs. 3 and. For oudoor condiion, all he nodes are in an open space wih an area of m m. As for indoor condiion, he building area of m m is divided ino 6 rooms bounded by walls. In boh configuraions here are 3 nodes wih random posiions. Node acs as a source, as node 3 as desinaion, and he oher nodes migh ac as relays if considered necessary. Simulaion parameers are aken based on he applicaion of WLAN in ad-hoc wireless neworks as shown in Table I. denoes he finess funcion,,, and denoe he s, nd and 3rd objecive funcion, respecively, and,, denoe he corresponding weighs.,, and are respecively calculaed by equaion, 7, and. Weighs are deermined randomly, seleced, and changed gradually and periodically [3]. In our sudy,,, and are all se equally o /3. Due o he large number of searches over exising cooperaive pah pairs, opimizaion mehods such as geneic algorihm GA can be applied o deermine he opimal value. Uopia Poin Euclidian Disance } is he coordinae of he uopia poin for { variable Z ha should be minimized and variable Z ha } is he coordinae of he should be maximized and { poins on POF on he objecives plane. The normalizing value is deermined based on he mínimum value of, while is deermined by he maximum value of. In he simulaion resuls repored in secion V, his mehod is applied o hree problems in. B. izaion Mehod In he scalarizaion mehod, all objecives are organized ino a scalar by giving weigh o each of hem. Objecive funcions ha should be minimized are marked negaive, while hose ha should be maximized are marked posiive. To gain a sense node posiion m node posiion m 3 3 Fig. 3. Oudoor Configuraion To calculae he load variance of a pah, i is assumed ha aside from node acing as he source ha send daa o a

5 7 JOURNAL OF COMMUNICATIONS SOFTWARE AND SYSTEMS, VOL. 9, NO. 3, SEPTEMBER 3 desinaion, here are five oher nodes ha ransmi daa simulaneously o heir respecive desinaion nodes. As a resul, here migh be some nodes wih beer chance o become a relay due o heir relaively low raffic loads. In his example, hese five node pairs are using pah --9-3, 7-9-, -9--3, 6---, and I is assumed ha he sources, i.e., nodes, 7,, 6 and, each send daa a a rae of Mbps, 3 Mbps, 8 Mbps, 7 Mbps, Mbps, and Mbps, respecively. Whereas oher nodes are each assumed o have a random load of Mbps, 7 Mbps, Mbps, or 7 Mbps As for he indoor configuraion, cooperaive pah pair -3 and are obained wih he smalles Euclidean disance of.67. The values achieved of performance componens are power consumpion of W, SNR of.3 db, and load variance of 8.9 Mbps. In our reviewed example, opimizaion wih scalarizaion for all hree performance indicaors oudoors produces finess value of.88. The seleced cooperaive pah consiss of and As for indoor configuraion, he cooperaive pah pair are found o be and ---3 wih he finess value of -9.. The resul of he enire imes simulaion is shown in Figs. hrough. Beside he comparison beween Pareo and scalarizaion mehod, we also compare he resuls for oudoor and indoor configuraions. node posiion m PDF node posiion m Pareo 3 3 Fig.. Indoor Configuraion 3 Power Consumpion W TABLE I PARAMETERS OF SIMULATION Fig.. PDF of Power Consumpion Oudoor Value 6 Indoor pah loss exponen, Sandard deviaion of shadowing, 8 db Wall ransmission coefficien,.3 Power Transmi, W Transmi anenna gain, db Receive anenna gain, db Frequency,. GHz Bandwidh, MHz Noise, - dbm PDF Parameer Oudoor pah loss exponen, 3 Pareo 3 Power Consumpion W Fig. 6. PDF of Power Consumpion Indoor B. Opimizaion Resuls In deermining he resuls of his opimizaion we perform simulaions imes. This secion describes one of he simulaion resuls. Opimizaion by Pareo mehod for all hree performance indicaors in oudoor configuraion resuls in cooperaive pah pair -3 and ---3 having he smalles Euclidean disance of.699. Performance componens produced in he process are power consumpion of 3 W, SNR of 3. db, and load variance of 6.9 Mbps. Fig. shows he PDF probabiliy densiy funcion of power consumpion for oudoor configuraion. From Fig. i can be seen ha he larges value of power consumpion wih Pareo mehod in oudoor configuraion is 3 W while he resul from scalarizaion mehod varies beween 3 W, W, and W. On he oher hand, Fig. 6 shows ha he power consumpion in indoor configuraions based on he Pareo mehod varies beween 3 W and W, while he scalarizaion resuls in an accumulaion a W.

6 Prob[SNR <= abscissa] Prob[Load Variance <= abscissa] Prob[SNR <= abscissa] Prob[Load Variance <= abscissa] GUANANTARA AND HENDRANTORO MILTI-OBJECTIVE CROSS-LAYER OPTIMIZATION 7 From Figs. and 6, i is known ha Pareo mehod for he oudoor configuraion resuls in seleced cooperaive pah pair consising of one and wo hops, while for he indoor configuraion he seleced pair may consis of pahs having one o hree hops. This is because he received power a nodes obsruced by walls for indoor configuraion is under he hreshold power so ha more hops are required in selecion of cooperaive pah pair. A similar sory also happens wih he scalarizaion case, ha is, he number of hops consiuing he seleced cooperaive pah pair is greaer for he indoor han ha for he oudoor configuraion. Consequenly, cooperaive diversiy in he indoor scenario ends o consume more energy han in he oudoor, which can be expeced due o he presence of walls separaing rooms inside he building. The CDF cumulaive disribuion funcion of SNR in oudoor configuraion for boh mehods can be seen in Fig. 7. I shows ha opimizaion by Pareo mehod produces values of SNR slighly greaer han hose obained by scalarizaion mehod. However, boh mehods have he same range of SNR, ha is,. - db. The SNR median difference beween Pareo and scalarizaion mehod for he oudoor configuraions is approximaely. db Pareo 6 8 SNR db.9 Pareo Fig. 7. CDF of SNR Oudoor values are greaer compared o hose from he scalarizaion mehod. The range of SNR for he Pareo mehod is beween -. db, while for he scalarizaion mehod, SNR value is in he db range. In his case, he median difference of SNR beween he wo mehods is roughly db. Comparing he median differences from he oudoor and indoor configuraions, i can be observed ha he indoor case benefis more han he oudoor case does from he use of Pareo mehod over he scalarizaion. Fig. 9 shows he CDFs of load variance for oudoor configuraion. The values of load variance resuling from he use of Pareo mehod is found o be smaller han hose produced by he scalarizaion mehod. For he Pareo mehod he load variance ranges from. - 6 Mbps, as using scalarizaion mehod, i ranges from Mbps. The CDFs of load variance for he indoor scenario are given in Fig., which shows again ha he load variance acquired by employing he Pareo mehod ends o be smaller han ha produced by he scalarizaion mehod. The range of load variance obained by Pareo is from. - 6 Mbps, as he scalarizaion mehod resuls in he range beween Mbps. This observaion confirms ha he Pareo mehod ouperforms he scalarizaion in balancing he raffic loads among he nodes Pareo Load Variance Mbps Fig. 9. CDF of Load Variance Oudoor SNR db Pareo Fig. 8. CDF of SNR Indoor Fig. 8 shows he CDF of SNR for indoor configuraions and demonsraes ha by using Pareo mehod he achieved SNR Load Variance Mbps Fig.. CDF of Load Variance Indoor

7 76 JOURNAL OF COMMUNICATIONS SOFTWARE AND SYSTEMS, VOL. 9, NO. 3, SEPTEMBER 3 In addiion, from Figs. 9 and, i is known ha he median difference in load variance beween he Pareo and scalarizaion mehod in oudoor configuraion is equal o 3 Mbps, while in indoor configuraion, he median difference is 7 Mbps. As in he case of SNR, he indoor scenario appears o benefi more han he oudoor from he use of he Pareo mehod in reducing he load variance. C. Compuaion Time The Pareo mehod is found o ake a longer ime o complee in he simulaion compared o scalarizaion mehod. For a oal of imes simulaion, he Pareo mehod akes 6. hours o complee 7.3 minues per simulaion, while he scalarizaion mehod only akes abou 3.96 hours abou.7 minues per simulaion. I means ha Pareo mehod akes on average. imes longer han he scalarizaion mehod. This is because Pareo mehod akes ino accoun all possible cooperaive pairs in he opimizaion. On he oher hand, wih he scalarizaion mehod, he opimizaion is done ieraively and randomly, depending upon he populaion and he number of ieraions. This compuaional resuls are obained for simulaions on Malab R9a run on a compuer wih Core CPU GHz and GHz RAM. A compuer wih higher specificaions can be used o ge faser compuaion. VI. CONCLUSIONS From he analysis of he opimizaion resuls, several poins can be highlighed. Firsly, in selecing cooperaive pah pair using MOO wih he Pareo mehod, performance indicaors under consideraion are aken care of separaely. Wih he scalarizaion mehod, performance indicaors of ineres are incorporaed in he scalar finess funcion. I is herefore expecable ha he resuls of he Pareo mehod give a beer compromise of he performance indicaors. Secondly, he opimizaion resuls obained wih he Pareo mehod are beer han hose obained using scalarizaion, as shown by he hree performance indicaors of cooperaive diversiy neworks considered herein, i.e., power consumpion, signal-o-noise power raio and load variance. Thirdly, he advanage of he Pareo mehod over scalarizaion is more prevalen for indoor cooperaive diversiy neworks han for heir oudoor counerpars. This is suppored by he finding ha he median difference of SNR beween he Pareo and scalarizaion is greaer for indoor han for oudoor scenario, and similarly so for load variance. Lasly, Pareo mehod requires a longer compuing ime han scalarizaion does because Pareo mehod is enumeraive while scalarizaion mehod is random. Hence, if he problem of compuaion ime can be alleviaed by employing a fas compuing processor, he use of Pareo mehod in MOO for cooperaive diversiy pahs selecion is recommendable. ACKNOWLEDGMENT The graduae sudy and research work of N. Gunanara has been suppored by a Posgraduae Scholarship BPPS from he Indonesian Minisry of Educaion and Culure. REFERENCES [] D. D. Perkins, H. D. Hughes, and C. B. Owen Facors Affecing he Performance of Ad Hoc Neworks, Proceedings of he IEEE Inernaional Conference on Communicaions ICC, pp.8-,. [] A. Nosrainia, T. E. Huner, and A. Hedaya Cooperaive Communicaion in Wireless Neworks, IEEE Commun Magazine, Vol., No., pp. 7-8,. [3] M. Kubisch, S. Mengesha, D. Hollos, H. Karl, and A. Wolisz Applying Ad-hoc Relaying o Improve Capaciy, Energy Efficiency, and Immission in Infrasrucure-based WLANs, Proc. of Kommunikaion in Vereilen Sysemen, Leipzig, Germany, 3. [] X. Edward Li Hop Opimizaion and Relay Node Selecion in Muli-Hop Wireless Ad-hoc Neworks, LNCS Social Informaics and Telecommunicaions Engineering, Vol., pp 6-7, 9. [] A. Blesas, A. Lippman, and D. P. Reed A Simple Disribued Mehod for Relay Selecion in Cooperaive Diversiy Wireless Neworks, based on Reciprociy and Channel Measuremens, Proceedings of he IEEE Vehicular Technology Conference Spring, Sockholm, Sweden,. [6] J. Huang, Z. Han, M. Chiang, and H. V. Poor Aucion-Based Resource Allocaion for Cooperaive Communicaions, IEEE Journal on Seleced Areas in Comm. Vol. 6, No. 7, 8. [7] Y.Shi, Y.T. Hou, H. D. Sherali, and S. Kompella Cross Layer Opimizaion for UWB-Based Ad hoc Neworks, Miliary Communicaions Conference MILCOM, pp. -7, 6. [8] L. Le, and E. Hossain Cross Layer Opimizaion Frameworks for Mulihop Wireless Nework Using Cooperaive Diversiy, IEEE Trans. On Wireless Communicaion, Vol. 7, No. 7, 8. [9] W. Chen, L. Dai, K. B. Leaief, and Z. Cao A Unified Cross Layer Framework for Resource Allocaion in Cooperaive Neworks, IEEE Transc. on Wireless Communicaions, Vol. 7, No. 8, 8. [] L. Ding, T. Melodia, S. N. Baalama, and J. D. Mayjasy Disribued Rouing, Relay Selecion, and Specrum Allocaion in Cogniive and Cooperaive Ad Hoc Neworks, Proceedings of he Sevenh Annual IEEE Communicaions Sociey Conference on Sensor, Mesh and Ad Hoc Communicaions and Neworks, Boson,. [] M. Elmusrai, H. El-Sallabi,and H. Koivo Applicaions of Muli- Objecive Opimizaion Techniques in Radio Resource Scheduling of Cellular Communicaion Sysems, IEEE Transc. on Wireless Communicaion, Vol. 7, No., 8. [] A. S. Karkkainen, K. Mieinen, and J. Vuori Bes Compromise Soluion for a New Muliobjecive Scheduling Problem, Elsevier Compuers & Operaions Research, Vol. 33, pp , 6. [3] A. de Baynas, P. Mahonen, and M. Perova ARQ-Based Cross Layer Opimizaion for Wireless Mulicarrier Transmission on Cogniive Radio Neworks, Elsevier Compuer Neworks, Vol., pp , 8. [] K. J. Runser, C. Comaniciu, and J. M. Gorcey A Muliobjecive Opimizaion Framework for Rouing in Wireless Ad Hoc Neworks, IEEE Inernaional Symposium on Conference Modeling and Opimizaion in Mobile, Ad Hoc, and Wireless Neworks WiOp,. [] N. Gunanara and G. Hendranoro Muli-Objecive Cross-Layer Opimizaion wih Pareo Mehod for Relay Selecion in Mulihop Wireless Ad hoc Neworks, WSEAS Transacion on Communicaions, Vol., Issue 3, 3. [6] S. McGhee, T. F. Chan, and R. V. Dubey Probabiliy-Based Weighing of Performance Crieria for a Redundan Manipulaor, IEEE Iner. Conf. on Roboics and Auomaion, pp , 99. [7] W. Chen, Q. Zhang, Z. Zhao, and W. A. Gruver Opimizing muliple Performance Crieria in Redundan Manipulaors by Subask-Prioriy Conrol, IEEE Iner. Conf. on Sysems Man and Cyberneics, pp. 3-39, 99. [8] R. Ranjan Roy Handbook of Mobile Ad hoc Neworks for Mobiliy Models, Springer, London,. [9] S. Capkun, M. Hamdi, and J. P. Hubaux GPS-free posiioning in mobile Ad-Hoc neworks, Proceedings of he 3h Hawaii Inernaional Conference on Sysem Sciences,.

8 GUANANTARA AND HENDRANTORO MILTI-OBJECTIVE CROSS-LAYER OPTIMIZATION [] M. Salem, A. Adinoyi, M. Rahman, H. Yanikomeroglu, D. [] [] [3] [] [] [6] [7] [8] [9] Falconer, Kim Young-Doo, E. Kim, Y. C. Cheong An Overview of Radio Resource Managemen in Relay-Enhanced OFDMABased Neworks, IEEE Communicaions Surveys & Tuorial, Vol., No. 3,. W. Honcharenko, H. L. Beroni,, and J. Dailing Mechanisms Governing Propagaion Beween Differen Floors in Buildings, IEEE Transc. On Anennas and Propagaion, Vol., No. 6, 993. S. M. Hussain, M. Alouini, and M.O. Hasna Performance Analysis of Bes Relay Selecion Scheme for Amplify-andForward Cooperaive Neworks in Idenical Nakagami-m Channels, IEEE Elevenh Inernaional Workshop on Signal Processing Advances in Wireless Communicaions SPAWC,. J.W., Wong, J.P., Sauve, and J.A., Field A Sudy of Fairness in Packe Swiching Neworks, IEEE Transacions on Communicaions, Vol. 3, No., 98. O. L. D. Weck Muliobjecive Opimizaion Hisory and Promise, Proceedings of 3rd China-Japan-Korea Join Symposium on Opimizaion of Srucural and Mechanical Sysems, Kanazawa, Japan,. M. Ehrgo Mulicrieria Opimizaion, Springer, Germany,. E. K. P. Chong, and S. H. Zak, An Inroducion o Opimizaion, Third Ediion, John Wiley & Sons, USA, 8. F. Pernode, H. Lahmidi, and P. Michel Use of Geneic Algorihms for Mulicrieria Opimizaion of Building Refurbishmen, Elevenh Inernaional IBPSA Conference, Glasgow, Scoland, 9. T. Ozcelebi Muli-Objecive Opimizaion for Video Sreaming, PhD Thesis, Graduae School of Sciences and Engineering, Koc Universiy, 6. B. E. Cohanim, J. N. Hewi, and O. de Weck The Design of Radio Telescope Array Configuraions Using Muliobjecive Opimizaion Imaging Performance versus Cable Lengh, The Asrophysical Journal Supplemen Series, Volume, Issue, pp. 7-79,. 77 [3] T. Muraa and H. Ishibuchi Muli-Objecive Geneic Algorihm and is Applicaion o Flow-Shop Scheduling, Inernaional Journal of Compuers and Engineering, Vol. 3, No., 996. Nyoman Gunanara received he B.Eng. degree in elecrical engineering from Universias Brawijaya UB, Indonesia, in 997. In 997 -, he worked in SIEMENS as The Leader of Cable Teser Uni. He was responsible for he qualiy of he cable nework and cooperaion wih PT. TELKOM Indonesia. Since, he joined wih Universias Udayana Unud, Bali as lecurer. He received M.Eng degree in elecrical engineering from Insiu Teknologi Sepuluh Nopember ITS, Indonesia, in 6. He is currenly working oward he Ph.D. degree in elecrical engineering from ITS. His research ineress include wireless communicaions, ad hoc nework, qualiy nework, and opimizaion. He is a Suden Member of he IEEE. Gamanyo Hendranoro received he B.Eng degree in elecrical engineering from Insiu Teknologi Sepuluh Nopember ITS, Indonesia, in 99, and he M.Eng and Ph.D degrees in elecrical engineering from Carleon Universiy, Canada, in 997 and, respecively. He is presenly a Professor wih ITS. His research ineress include radio propagaion modeling and wireless communicaions. Dr. Hendranoro is a Senior Member of he IEEE.