Mixed Contiguous and Aggregated Spectrum Allocation Algorithm for CR based TD-LTE System

Transcription

1 Commuicatios ad etwork, 2013, 5, Published Olie September 2013 ( ixed Cotiguous ad Aggregated Spectrum Allocatio Algorithm for CR based D-LE System Xiaoya Liu, Qigyi Qua, Peg Lei, Webi Guo, Webo Wag WSP Lab. Key Laboratory of Uiversal Wireless Commuicatio, iistry of Educatio, School of Iformatio ad Commuicatio Egieerig, eiig Uiversity of Posts ad elecommuicatios, Chia Received July, 2013 ASRAC Cogitive Radio (CR) is a promisig techology to icrease the efficiecy of spectrum by opportuistic access. However, i may cases the sesed spectrum is so broke that it caot be utilized sufficietly with traditioal cotiuous spectrum catio strategy. Spectrum regatio ca make full use of arrow spectrum fragmets while the overhead it brigs may reduce the overall performace of the system. his paper proposes a ixed Cotiguous ad Aggregated Spectrum Allocatio (ixcasa) algorithm for CR based D-LE etworks. his algorithm combies cotiguous with discotiguous spectrum catio together i order to guaratee the feasibility ad efficiecy of system. Simulatio results show that the ixcasa algorithm ca icrease the umber of cated chaels, ad reduce the overhead brought by excessive spectrum regatio as well. Keywords: Cogitive Radio; D-LE; Spectrum Allocatio; Spectrum Aggregatio 1. Itroductio As a stadard for wireless commuicatio techology, LE offers cosiderably high data rate [1]. While the spectrum usage efficiecy has bee improved, it is ecessary to icrease the trasmissio badwidth to achieve a higher data rate accordig to Shao formula. However, spectrum is oe of the mai obstacles that the developig wireless commuicatio techology must overcome. esides, traditioal spectrum catio strategy is proved to be iefficiet [2], so fidig ew algorithms to assig this scarce resource more ratioally is urget. CR is a promisig way to solve this problem. I CR etworks, able spectrum is automatically detected to w the access of Secodary Users (s) without sesible iterferece to Primary Users (PUs) [3]. However, able spectrum coditio of CR etworks has great differece compared with that of traditioal wireless commuicatio etworks. ecause the sesed spectrum may ot be stable all the time ad its character also chages. esides, i most cases the spectrum is so fragmeted that it is hard or impossible to be utilized efficietly [3]. Owig to these facts, traditioal spectrum catio protocols are ot very applicable. Spectrum regatio is a feasible way to settle the problems metioed above. y bodig multiple spectrum fragmets together, it ws multiple cotiguous or discrete blocks of spectrum to be treated as if they are oe large cotiguous block [4]. However, the regatio capacity is limited to some extet ad fragmets that are too scattered caot be boded together. oreover, regatio also brigs overhead ad it has high requiremet for hardware [5]. So excessive spectrum regatio should be avoided to reduce uecessary overhead ad guaratee the spectrum efficiecy at the same time. esides, spectrum resource is similar to computer memory resource i some aspects, for both of them may be fragmetized durig the catio, ad excessive fragmets hadicap their performace. I this paper, we are ispired by the best-fit strategy of computer memory catio, which is a good way to miimize fragmets [6]. est-fit memory catio picks out the miimum oe amog all the resources that satisfy user s demad. Ad it makes the best use of resource i this way. Cosiderig those metioed above, we propose a ixed Cotiguous ad Aggregated Spectrum Allocatio (ixcasa) algorithm. I this algorithm, spectrum or chael resource is cated i two ways. If there is sufficietly able cotiguous resource for a certai user demad, best-fit spectrum catio strategy is implied. Otherwise spectrum regatio is carried out to icrease the total umber of chaels that ca be cated as may as possible. he rest of this paper is orgaized as follows. I part 2 related works are illustrated. he the system model ad the ixcasa algorithm are proposed i part 3. After

2 X. Y. LIU E AL. 299 givig the simulatio results ad aalyzig the performace i part 4, we make our coclusios i part Related Work I order to utilize the scarce spectrum resource more effectively ad efficietly, Dyamic Spectrum Allocatio (DSA) has bee well studied these years. Previous works such as [7] ited to icrease spectrum efficiecy ad guaratee user s demad i a cotiguous spectrum level. However, i most cases the sesed spectrum may ot be that perfect ad may have may arrow fragmets [8], thus it is difficult to make full use of this scarce resource with traditioal cotiuous spectrum catio strategy. Wireless radio techology such as Discotiguous Orthogoal Frequecy Divisio ultiplexig (DOFD) [9] makes spectrum regatio possible, i which may detached spectrum holes ca be oit together ad the utilized as if they were a whole block [10]. It is a promisig techology for its flexibility ad dyamic characteristic ad low iterferece betwee adacet subchaels whe catig spectrum resource. Spectrum regatio works exactly this way [11]. Spectrum regatio makes it possible to utilize more tha oe carrier ad i this way icrease the overall trasmissio badwidth. Although techique has bee already quite well ivestigated at theoretical level, their practical implemetatio is ot immediate ad also raises umerous challeges. Ituitively speakig, the regatio capacity caot be ulimited [12], otherwise the correspodig hardware may ot satisfy its requiremet ad the overhead regatio brigs may hadicap the overall performace of the system as well. Aggregatio Aware Spectrum Assigmet (AASA) proposed i [13] is a greedy algorithm. his algorithm gives a regatio capacity amed regatio spa, which performs i the form of slidig. It ca achieve maximum access umber by assigig able chaels withi the slidig ad the movig the from lower to upper frequecy successively. However this greedy algorithm is ot very practical because every i this model is set to have the same badwidth demad ad it does ot take overhead brought by chael regatio ito cosideratio. aximum Satisfactio Algorithm (SA) preseted i [12] cosiders differet badwidth requiremets of s. I this algorithm with larger badwidth requiremet ca be cated preferetially. While it oly focus o maximizatio of the total access umber of s badwidth requiremets, ad still eglects the load regatio brigs to system. ixcasa algorithm we propose i this paper aims to improve spectrum efficiecy, ad avoid excess regatio at the same time by combie cotiguous ad discrete spectrum catio together. So it ca guaratee a relatively high umber of cated chaels ad avoid excessive overhead brought by spectrum regatio. Experimet results show that this algorithm ca achieve a much better performace whe takig both spectrum efficiecy ad regatio overhead ito cosideratio. 3. System odel ad Algorithm 3.1. Sceario Defiitio We assume a CR etwork that has s deoted by U, {1, 2, }. Cosiderig the most commo wireless commuicatio sceario, we make some assumptios as follows. here are eough s, ad every chael ca oly be occupied by oe or PU. Every has a badwidth requiremet R, {1, 2, }. Ad all of these requiremets compose a array R. R R 1, R 2,, R (1) where R is radom distributed so that it ca lie with the practical situatio, i which s have various badwidth requiremets. he coditio of spectrum remais uchaged, which meas we have a stable spectrum that has already bee sesed durig the catio, ad the total badwidth is Hz. he maximum regatio capacity is, ad the width of slidig is, which is show i Figure 1. As moves forward, the whole spectrum will be searched. Assume that spectrum is uitized, so we ca label these chaels i iteger umbers. he total spectrum total ca be deoted by. Ad the total able spectrum is deoted by ail. Every able spectrum av fragmet ad its legth ca be deoted Available spectrum fragmets Uable spectrum fragmets L U Figure 1. Fragmet fragmets ad catio schematic. L 1 Frequecy

3 300 X. Y. LIU E AL. as follows. L, L 1,, U (2) U L 1, 1,2, J (3) where L ad U are the lower ad upper uitized chael umber of th able spectrum, which satisfy the followig relatioship. L U L 1, 1, 2, J (4) So the total uitized able chael ad its legth is J 1 (5) J. 1 (6) I this model, we assume that with larger badwidth requiremet has priority to be cated firstly. Let C deotes the total umber of chaels assiged after the whole catio process. If K s requiremet are accepted i all, the equatio of C ca be deoted as follows. K C R, 1,2, 1 k k (7) k Ad cated chael umbers C caot be larger tha the umber of total able chaels. So we have C (8) Ad our target is to maximize C ad avoid excessive regatio at the same time Descriptio of Proposed Algorithm I this paper, we propose a ixcasa algorithm to solve the problem described above. he key idea of ixcasa is to cate spectrum resource cotiguously i best-fit method if there are sufficietly eough cotiguous vacat chaels for a s demad. Otherwise spectrum regatio is applied to guaratee the umber of chael that ca be cated. All s are cated oe by oe accordig to their badwidth requiremets, which are sorted decreasigly. Steps of this algorithm ca be simplified as follows. 1) Sort all the chael requiremets of s decreasigly. So with larger demad has a priority to use the limited spectrum resource. For a certai, spectrum assigmet ca be divided ito two stages. 2) Durig the cotiguous assigmet, best fit strategy is applied. Ad the whole sesed spectrum is searched to pick out the smallest oe amog all the spectrum fragmets that are big eough for this s requiremet. 3) Durig the discotiguous catio, able chaels withi slidig are searched ad boded together to satisfy the s demad. If o sufficiet chaels ca be foud after these two stages, this s requiremet will be refused. As metioed above, the proposed algorithm cotais cotiguous chael assigmet ad discotiguous chael assigmet. It is ot hard to discover that steps of this algorithm ca be simplified by settig the width of slidig as follows. Durig cotiguous chael assigmet stage, R, {1, 2, } (9) Ad durig regatio stage, we have (10) he model of slidig is show i Figure 1. Differet values of mea differet catio modes. As the slidig moves successively, all the chaels ca be checked ad cated if it satisfies ay oe of these two modes. Detailed procedure of ixcasa is demostrated i Algorithm 1 ad Algorithm 2. Algorithm 1: ixcasa Iput: s badwidth requiremet vector R, ormalized able chaels, regatio, ad the width of slidig Output: Chael state after the catio of all s deoted by 0 (ucated) ad 1(cated) for = 1 to do tag = 0 for = 1 to if do >0 ad R <= S1 = /* S1 is a temporary array to store */ tag = 1 ed if ed for if tag = 1 the else R the y = mi ( S1 ) /* y is the locatio of the miimum of S1 */ L L 1 y y Spectrum Aggregate ed if ed for

4 X. Y. LIU E AL. 301 Algorithm 2: Spectrum Aggregatio Iput: Slide- width, request R, able chaels th s badwidth Output: Chael state after the assigmet of th /* p1, p2 are the start ad stop locatio of slide */ for p1 = 1 to do p2 = p1+ 1 S2 = fid( ( p1: p2) 0 ) /* S2 stores the locatio of able chaels withi slide */ if legth(s2) >= R do L L 1 p1 p1 else p1 = p1+1 ed if ed for 4. Simulatio ad Performace Aalysis We cofigure the simulatio as show i able 1. Accordig to curret samplig speed of aalog-to-digital coverters, the regatio capacity, or regatio we call here, is set at 40 Hz. Spectrum rage is fixed at VHF ad UHF withi broadcastig V bad from 100 to 600 Hz. s badwidth requiremets are set to rage from 4 Hz to 20 Hz. I the simulatio, chael fragmet umber chages from 1 to 100. Ad whe it is 1, it meas we have a ubroke spectrum, which is a theoretical coditio that we take as a referece. We ru these algorithms for 150 times respectively ad the variables are re-geerated i every ru. he we get the average values of simulatio as follows. Figure 2 shows the cated chael umber, i.e. the umber of chaels that ca be cated by algorithms metioed above, icludig AASA from [12], ixcasa proposed i this paper ad traditioal cotiguous spectrum catio. Here we deote them with array A, ad C. AASA is a greedy algorithm, which slides the from low to high oe by oe without cosiderig the other factors. So it is a theoretical strategy but ot very practical. Whe the umber of fragmets is ot very big, i.e. the sesed spectrum is relatively complete, the cated chael umber of these algorithms early stay the same. As spectrum fragmets icreases, performace of ootiguous spectrum cate strategy goes dowhill quickly while the performace of ixcasa algorithm oly has a slight decrease. Figure 3 shows the regatio umber, i.e. how may times regatio occurred durig the catio. hey are deoted by array ad. Here we A make the assumptio that there is a regatio as log as the width of slide- is set at the maximum regatio capability, ad chaels are assiged to certai successfully durig this assigmet. ecause there is o regatio i cotiguous spectrum assigmet, we ust preset the regatio umber of AASA ad ixcasa algorithm here. Compare Figure 2 with Figure 3 we ca see that whe the sesed spectrum is ot so fragmetary, regatio umber of ixcasa is sigificatly reduced tha that of greedy algorithm. Ad the umber of cated chaels remais a relatively high level at the same time. he data i Figure 2 ad Figure 3 are coverted i followig way so that we ca have a compariso more uderstadable ad reasoable. A /max( A ) RAASA (11) /max( ) R ixcasa A A /max( ) /max( ) (12) he results of R ad AASA RixCASA are show i Figure 4, which i part reflect the differet overhead of regatio whe catig sub-chaels i differet ways. We ca see that the ormalized average regatio overhead of ixcasa is much smaller tha that of AASA. Parameter ame Frequecy rage able 1. Simulatio cofiguratios. otal able Frequecy adwidth Per Sub-carrier Aggregatio Value 100Hz to 600Hz 300Hz 1Hz 40Hz Figure 2. umber of chaels cated of differet algorithms.

5 302 X. Y. LIU E AL. Figure 3. Aggregatio umber of AASA ad ixcasa. Figure 4. Coverted overhead of regatio. 5. Coclusios A ixcasa algorithm has bee proposed i this paper for CR based D-LE system. I this algorithm, spectrum resources are cated cotiguously i best-fit method if there are sufficietly eough cotiguous vacat chaels for a s demad. Otherwise spectrum regatio is applied to guaratee the umber of chaels that ca be cated. Simulatio results have show that whe the sesed spectrum is ot extremely broke, this algorithm ca guaratee a high umber of cated chaels ad reduce the overhead brought by excessive spectrum regatio as well. ecause spectrum efficiecy ad overhead are the mai factors that we try to balace, we do t take the iterferece s brig to PUs ito cosideratio i this paper. Our future work will focus o both spectrum efficiecy ad power optimizatio. 6. Ackowledgemets his work is supported i part by atioal Key echology R&D Program of Chia (2012ZX ), SFC (Grat ), ad atioal 973 Program of Chia (2012C316005). REFERECES [1] LE: he US Log erm Evolutio, ew York: Joh Wiley & Sos, [2] I. F. Akyildiz, W. Y. Lee,. C. Vura, et al., ext Geeratio/Dyamic Spectrum Access/cogitive Radio Wireless etworks: A Survey, Computer etworks, Vol. 50, o. 13, 2006, pp doi: /.comet [3] I. F. Akyildiz, W. Y. Lee,. C. Vura, et al., A Survey o Spectrum aagemet i Cogitive Radio etworks, Commuicatios agazie, IEEE, 2008, pp [4] L. Che, W. Che, X. Zhag, et al., Aalysis ad Simulatio for Spectrum Aggregatio i LE-advaced System, Vehicular echology Coferece Fall (VC 2009-Fall), 2009 IEEE 70th. IEEE, 2009, pp [5] Z. She, A. Papasakellariou, J. otoo, et al., Overview of 3GPP LE-advaced Carrier Aggregatio for 4G Wireless Commuicatios, Commuicatios agazie, IEEE, Vol. 50, o. 2, 2012, pp doi: /co [6]. asmao, I. Ripoll, P. albastre, et al., A Costat-time Dyamic Storage Allocator for Real-time Systems, Real-ime Systems, Vol. 40, o. 2, 2008, pp doi: /s [7]. Hamid, A. ohammed ad Z. Yag, O Spectrum Sharig ad Dyamic Spectrum Allocatio: AC Layer Spectrum Sesig i Cogitive Radio etworks, Commuicatios ad obile Computig (CC), 2010 Iteratioal Coferece o IEEE, 2010, Vol. 2, pp [8]. Lee ad. J. Lee, ultichael AC Protocol with Discotiguous-OFD for Cogitive Radio etworks, Iformatics Egieerig ad Iformatio Sciece, Spriger erli Heidelberg, 2011, pp doi: / _53 [9] Che Y S, Liao S H. Spectrum-aware routig i discotiuous orthogoal frequecy divisio multiplexig-based cogitive radio ad hoc etworks, etworks, IE, Vol. 1, o. 1, 2012, pp doi: /iet-et [10] Posto J D, Hore W D. Discotiguous OFD Cosideratios for Dyamic Spectrum Access i Idle V Chaels, ew Frotiers i Dyamic Spectrum Access etworks, DySPA, First IEEE Iteratioal Symposium o. IEEE, 2005, pp [11] Y. Li, L. Zhao, C. Wag, et al., Aggregatio-based Spectrum Allocatio Algorithm i Cogitive Radio etworks, etwork Operatios ad aagemet Symposium (OS), IEEE, 2012, pp [12] F. Huag, W. Wag, H. Luo, et al., Predictio-ased Spectrum Aggregatio with Hardware Limitatio i Cogitive Radio etworks, Vehicular echology Coferece (VC 2010-Sprig), 2010 IEEE 71st., 2010, pp [13] D. Che, Q. Zhag ad W. Jia, Aggregatio Aware Spectrum Assigmet i Cogitive ad-hoc etworks, Cogitive Radio Orieted Wireless etworks ad Commuicatios, CrowCom rd Iteratioal Coferece o. IEEE, 2008, pp. 1-6.