AMC-based resource allocation in adaptive frequency reused OFDMA-relay networks

Transcription

1 Apil 8(): The Jounal of China Univesities of Posts and Telecounications AMC-based esouce allocation in adaptive fequency eused OFMA-elay netwos CHE Yu ( ) FAG Xu-ing School of Infoation Science and Technology Southwest Jiaotong Univesity Chengdu 63 China Abstact In othogonal fequency division ultiple access (OFMA) elay syste fo suppoting elay tansission the base station (BS)-the elay station (S) lin ust consue an exta pat of esouce which ay esult in seious esouce shotage. In ode to ipove esouce utilization this pape poposes a dynaic esouce allocation schee in adaptive fequency eused OFMA-elay syste based on adaptive odulation and coding (AMC) technology. In this schee elay nodes have two indepent antennas and opeate in decode-and-fowad (F) and full-duplex ode. Then the BS and Ss shae the sae subcaies by spatial ultiplexing by two indepent antennas. The esouce allocation poble is foulated fo syste downlin thoughput axiization. Since the optial solution couldn t be obtained easily a sub-optial algoith is poposed. The adaptive fequency eused algoith with two indepent antennas S ipoves the syste thoughput about 4.3 % copaed with the othogonal fequency allocation with single-antenna odel and inceases the syste thoughput.4 % copaed with adaptive fequency eused algoiths with single-antenna S. It is poved that both of the S with two-antenna odel and adaptive fequency eused schee can ipove the syste thoughput significantly. eywods AMC esouce allocation adaptive fequency euse OFMA elay Intoduction In OFMA syste the use divesity could be ipleented by caefully select the pope fequencies. ecently soe papes have poposed the dynaic subcaie allocation bit allocation and powe distibution solutions to incease the cell thoughput and ipove the spectal efficiency [ 3]. In next geneation counication the elay technology is intoduced in 8.6j 8.6 and LTEadvanced to povide boade signal coveage highe data tansission ate and faste obility. Howeve in elay syste the BS-S lin ust consue an exta pat of esouces and it ay esult in oe esouce shotage. Many pevious eseaches focused on the othogonal esouces (in tie o fequency) allocated to the uses in the cell of the elay syste [4]. They divide the bandwidth into any fixed fequency bands and then diffeent S and BS ae allocated eceived date: -8- Coesponding autho: CHE Yu E-ail: chey593@63.co OI:.6/S5-8885()645-3 diffeent othogonal esouces by a esouce anageent policy. ef. [5] poposed the optial algoith in half-duplex elay syste with the sub-channel allocated to achieve axiu syste thoughput. To save the esouces used in BS-S lin the wo on elay-based netwo assue that S can tansit and eceive siultaneously in the sae subcaies i.e. wo on full-duplex ode [6]. Howeve thee ae any liitations in adio ipleentation peclude the teinals fo this opeation. Theefoe thee lac of sufficient studies about adaptive-euse subcaie allocation between BS and S based AMC elay syste. In this pape suppose BS and Ss shae the sae fequency though AMC and spatial ultiplexing by using the F elay with two indepent antennas. The taget is to axiize the syste thoughput of a boadband cellula OFMA syste with the adaptive fequency euse between BS and Ss. Because the optial algoith is a non-linea optiization poble with intege vaiables this pape poposes a sub-optial algoith. The sub-optial algoith uses the chaacteistic of the AMC-based syste which can

2 6 The Jounal of China Univesities of Posts and Telecounications toleate soe intefeence. Then the subcaies ae adaptively chosen to the BS o an S o both though channel state infoation (CSI) and the condition of the intefeence. And the subcaies assignents fo BS and S ae sepaated that siplify the non-linea optiization with intege vaiables. Syste odel This pape consides an AMC-based ultiuse dual-hop downlin syste with iniu data ate constaints fo each MS. The syste has active obile stations (MSs) M Ss and a single BS as shown in Fig.. It is denoted that the fist hop is the BS-MS lin o BS-S lin and the second hop is the S-MS lin. It is assued that the wieless channel between each pai of tansitting and eceiving nodes is fequency selective and OFMA is eployed fo data counication to divide the channel into a set of othogonal subcaies with flat channel esponses and additive white Gaussian noises (AWG). Futheoe all MSs and the BS in the syste ae equipped with a single antenna. All Ss use the odel with two antennas an oni-diectional antenna fo the S-MS lins and a diectional antenna fo the BS-S lin i.e. fo eceiving L data fo the BS. This detailed odel is descibed in ef. [7]. Table SI of diffeent coding odulation schees Coding ate and Spectal SI at odulation level efficiency/ (bit s Hz ) cobination BE/dB QPS / QPS 3/ QAM / QAM 3/ QAM / QAM 3/ QAM 7/ We also assue that the two antennas ae indepent of each othe and the diectional antenna s bea is vey naow. Theefoe the S-MS lin doesn t intefee with the BS-S lin when the two lins use the sae esouces siultaneously. The subcaies in BS-S lin can be eused in the S-MS lin. The ate unde AMC is discete and the lin with AMC can toleate soe intefeence. Fo this eason soe subcaies used in BS-MS lin can be also used in S-MS lin if the intefeence is acceptable. Thus the BS and S can use the sae subcaie i.e. the BS and the S euse the adaptive fequency band. Howeve shaing a subcaie by BS o diffeent S is not allowed whee the BS seves as the cental planne fo the cell that contols the esouce allocation of all uses and all elays with all CSI (channel state infoation). Additionally the aveage tansit powe of each subcaie of both BS and S on the othe hand is ept constant. The aveage powe of subcaie tansitted fo BS is P BS and the one tansitted fo S is P S. ext section will focus on the downlin esouce allocation fo the syste which axiizes the syste thoughput. 6 3 Poble foulation Fig. Syste odel AMC can be indepently applied to each lin in the syste because of using F S. The selected odulation and coding schee (MCS) ight be diffeent fo both hops. Theefoe the subcaie can select MCS accoding to SI denoted as γ and the ate of the MCS can be siply expessed as f( γ) = ax{ c c lb( + γ) c { c c... c w }} () whee c c... cn ae the coesponding ates of MCSs which ae all constant. As an exaple w = 7 and diffeent MCSs ae listed in Table [8] the bit eo ate (BE) is 6. In this pape denotes the iniu equied ate of uses. Fo each subcaie of index n ρ n is defined to indicate whethe the nth subcaie is allocated to the BS-the th MS lin o BS-S lin and q n indicates whethe the nth subcaie is allocated to the th S the th MS lin. A n denotes the SI of the lin between BS and the th MS. B denotes the SI of the lin between the th S and n the th MS. C n denotes the SI of the lin between BS and the th S. Since the tansission in elay syste includes both diect tansission and elayed tansission it needs to define subcaie assignent indexes to calculate achievable ate of use fo distinguishing diect tansission fo elayed tansission. Assignent indexes α and β ae defined as

3 Issue CHE Yu et al. / AMC-based esouce allocation in adaptive fequency eused OFMA-elay netwos 6 use choose diect tansission α = othewise use choose elayed tansission β = othewise whee fo each subcaie eithe α = o β = diectly o elayed tansitted. In addition if the use chooses diect d tansission the ate of use is denoted as. If the use chooses elayed tansission and the Ss ae with F odel the ate is denoted as. and can be witten as d ρ nf( A n) M M ρ n n n n = = d = () = in f( C ) q f( B ) whee PBS Hn An = M q β P H B C n n (3) + in i S n = i= i S Hn P = P H + ρin BS n i= BS Hn (4) (5) P = (6) n H n H H n denote the channel gains of BS-MS lin BS-S lin and S-MS lin espectively. The is the powe of white Gaussian noises. The ate of the use can be coputed as d = α + β = α ρ n ( n) + M M β ρ nf Cn q n f B n = = f A in ( ) ( ) (7) A esouce-allocation is consideed as an optial poble which achieves the axiu syste thoughput while eets the subscibes iniu ate equieent. Thus the poble can be foulated as ax (8) αβρ q = s.t. = α ρ f( A ) + n n M M β ρ nf Cn q n f B n = = in ( ) ( ) (9) ρn n () = M qn n () = = α β ρ q {} () α + = (3) n n β whee constaint Eq. (9) ensues that each use ust eet the iniu equied ate. Constaint Eq. ( 3) guaantees that each subcaie is occupied by only one tansitteeceive pai in each hop and BS and S can shae the subcaies in the syste. Constaint Eq. (3) ensues that the use ay choose diect tansission o elayed tansission. Unfotunately this esouce-allocation poble is a cobinatoial ixed intege pogaing poble fo which an exact solution usually involves an exhaustive seach. Thus the poble is coputationally extensive when the nube of vaiables is lage that is -P had. Theefoe to siplify the poble we poposed a suboptial esouce-allocation algoith. We siplify the Eq. (7) and allocate the esouce though two steps by utilizing AMC. 4 Suboptial solution As we now in F elay syste the ate of the elayed use is constained by the lowe one between the BS-S lin and the S-MS lin. And fo the Lea in ef. [5] we also now that the optial solution to Eqs. (8) (3) with continuous elaxation satisfies that the elayed use s ate of the BS-S lin equates the ate of S-MS lin. In the suboptial solution without continuous elaxation the subcaie allocation fo second hop is elated to the esult of the subcaie allocation fo fist hop. Besides adaptive fequency euse is adopted between the two hops in this syste i.e. the subcaies used in S-MS lin euse the subcaies assigned to BS-S o BS-MS lin. Hence thee ae oe available esouces assigned to S-MS lin than the BS-S lin. Howeve the suboptial solution without continuous elaxation satisfies the elayed use s ate of the S-MS lin is slightly geate than o equal to the ate of BS-S lin i.e. the ate of the elayed use is ostly deteined by the ate of the BS-S lin. The Eq. (7) can be siply witten as = α ρ f( A ) + β ρ f( C ) n n n n (4) Fo the Eq. (4) Eq. (5) and Eq. (4) it can be nown that the syste thoughput is affected by the ate of the fist hop and the intefeence to the fist hop fo the second hop. In an attept to siplify the subcaie allocation poble

4 6 The Jounal of China Univesities of Posts and Telecounications we conside the following thee-step distibuted esouce allocation appoach suaized as following steps: ) The use is deteined whethe the tansission is though the diect lin. ) Subcaies ae assigned to the fist hop though the optial algoith without consideing intefeence fo the second hop. 3) Subcaies ae assigned to the second hop with consideing the iniu toleable intefeence to the fist hop. By sepaately pefoing subcaie allocation in each hop this appoach yields a suboptial algoith which is consideably siple to ipleent than the intege pogaing poble. And the suboptial algoith will be pesented in detail in following subsections. 4. Tansission ode selection If the fist hop lin doesn t be intefeed the SI of the use tansitted diectly can be changed as PBS Hn A n = (5) Without consideing the ulti-path fading then A = A n B = Bn C = Cn ae assued with the subcaie assigned. Then the tansission odel can be selected accoding to the etic fo ef. [9] which uses capacity of thee wieless lins in elay syste (diect lin / bachaul lin / access lin) to decide whethe to tansit via S o not. The bandwidth equied fo tansitting X bits to the MS will be X /( T s) (diect tansission) o X /( BTs) + X /( ATs) (elaying tansission) whee T s is duation of subfae B and A ae use capacity of BS-MS lin BS-S lin and S-MS lin espectively. Thus this algoith decides to use elaying tansission when / B + / A < / that eans the bandwidth will be educed by elaying tansission. The optiization solution is as follows: = ax f( B ) (6) * = ag ax f( B ) (7) + < (8) f ( C * ) f( A ) * whee the th elay is selected by use. If Eq. (8) is satisfied the use selects the elay to tansission because of using less esouce i.e. α = β = othewise selects diect tansission i.e. α = β =. And the vaiables of the object function ae educed to two as ax (9) ρ q = 4. Subcaie allocation fo the fist hop We suppose that the subcaie doesn t be assigned to the S-MS lin yet. Then the fist hop doesn t be intefeed by the second hop. And the poble can be foed as ax () ρ = s.t. = ρ [ α f( A ) + β f( C )] () n n n () ρn = n (3) = ρn {} (4) So the poble could be conveted to assignent poble of the intege optiization. We can use Hungaian algoith to solve this poble [] which has a wost-case tie coplexity of O( 3 ) fo exhaustive seach in theoy. Howeve the siulation tie is actually uch shote in ef. []. With the Hungaian algoith we obtain the assigned esults of the fist hop of each use. If both β = and ρ n = it indicates that the subcaie n is assigned to the BS-S lin. Then we can assign the subcaies to the second hop lin accoding to the set of β and ρ. 4.3 Subcaie allocation fo the second hop Let denote the set of the uses which select elayed tansission lin and d denote the set of the uses which select diect tansission lin ϒ denotes the set of the all subcaies Λ is the set of the subcaies assigned to the BS-S lin Ω is the set of the subcaies assigned to the BS-MS lin Θ is the set of the subcaies not to be assigned M is the set of the elays. Because the elays ae fixed fo ease of analysis the thoughput of the BS-S lin can be constant with diffeent subcaie. We assue η = f( Cn ). The detailed algoith is poposed as Algoith. In Line 7- fistly select the unassigned subcaies and allocate esouces to the second hop fo use. Because of using the diectional antenna in the elay fo the BS-S lins the S-MS lin doesn t intefee with the BS-S lin using the sae esouce. If the unassigned subcaie isn t enough subcaies assigned to the BS-S lin ae selected to be

5 Issue CHE Yu et al. / AMC-based esouce allocation in adaptive fequency eused OFMA-elay netwos 63 eused in the second hop fo use (Line 3-). As in AMC-based syste within a cetain ange of the SI the ates ae the sae. So the subcaies with high SI in the ange can toleate oe intefeence. If the equied subcaie in second hop isn t still enough the S-MS lin can euse the subcaies which assigned to the BS-MS lin without o with less effect using AMC (Line 3-33). Algoith Subcaie allocation fo the second hop Input: M α β B n ρ n η * A n H n H * n P B. Output: q * n Begin = { β = } d = { α = } ϒ ={ n} Ω = { n ρ nβ = } Λ = { n ρ α = } Θ = ϒ Ω Λ d = cad ( Ω) n //The nube of eleents in set Ω = Ode ( ) // Let the nube in be in non-inceasing ode with SI. fo i = : cad ( ) = () i o = ρn // the nube of the subcaie in the fist hop fo use. n ρn * ρ * nη n o f (in B ) = f ( C ) = n // the nube of the subcaie in the second hop fo use. if o cad ( Θ ) fo j=: o n = Bestsubcaie ( Θ ) //Find the best subcaies in Θ. q * = Θ = Θ { n} n = {} d = d o bea else //Fistly assign the all est of subcaies in Θ to the use. x= fo j=: cad Θ n = Bestsubcaie ( Θ ) q * = Θ = Θ { n} x=x+ n if cad Λ o x fo j = : o x n = Bestsubcaie ( Λ) q * = Λ = Λ { n} n = {} d = d o bea else // Fistly assign the all est subcaies in Λ to the use. fo j=: cad ( Λ ) n = Bestsubcaie Λ q * = Λ = Λ { n} n fo j=: o x cad Λ PBS H * * n + S * * * n A * * = n P H f A = f A n Ω * * if ( * *) ( * * *) d n n * * = n = n q = * * * Ω = Ω { n } n = {} d = d o if d = φ o d = bea End Once obtaining αβρ q the syste thoughput can be copute as C = α ρ nf( A n) + β ρ nf( Cn ) = (5) Theefoe the tie coplexity of suboptial algoith is 3 O ( + ) in the wost case theoetically. 5 Siulation This section copaes the pefoance of adaptive euse schee in full-duplex elay with the othogonal schee. A elay enhanced OFMA single cellula is assued. The BS is located in the cente of each cell and six fixed Ss ae placed on the lines connecting BS and six cell vetices. Each S is /3 adius away fo the BS so as to achieve the optial syste pefoance. We also assue that all MSs eain stationay and the channel condition doesn t change. The siulation paaetes ae listed in Table which efe ainly to ef. []. In ode to avoid only siulating a single cell while to siulate a eal cellula scenaio we add the inte-cell intefeence and assue it is constant and ae the use in cell edge only choose the lowest odulation and coding schee. Table Siulation paaetes Paaete Assuption Site-to-site distance/.5 Caie fequency/ghz.5 Syste bandwidth/mhz Lognoal shadowing/db 8 ube of points in full FFT 4 # of used subcaies 8 # of the ultipath 3 BS-MS and S-MS path loss Baseline test scenaio BS-S path loss 8.6j EVM Type BS and S Tx powe/ db 46/38 Theal noise density/ (db Hz ) 74 Fig. shows the esult fo the adaptive euse esouce allocation schee with two antennas the adaptive euse

6 64 The Jounal of China Univesities of Posts and Telecounications esouce allocation schee with single antenna and the othogonal esouce allocation schee with single antenna. With the single antenna odel the BS-S lin and the S-MS lin use the sae subcaies but in diffeent phases i.e. the elays opeate in the half-duplex. Hence the othogonal esouce allocation schee with single antenna as the subcaies allocation schee in ef. [5] and the adaptive euse esouce allocation schee with single antenna using the euse schee poposed in this pape ae unde the half-duplex odel elay. The iniu equied ate of each use is 7 bit/s. Fo the figues we can see that with the nube of the uses the adaptive euse esouce allocation schee with single antenna can ipove.3 % syste thoughput copaed to the othogonal one and the adaptive euse esouce allocation schee with two antennas can ipove.44 % the syste thoughput copaed to the adaptive euse esouce allocation schee with single antenna i.e. the inceent in syste thoughput is caused by the two-antenna odel. Since thee ae an oni-diectional antenna fo the S-MS lins and a diectional antenna fo the BS-S lin in the two-antenna odel S the Ss eceive the signals diectionally and tansit oni-diectionally in downlin. This two-antenna odel can ae the S opeate in full duplex odel while the single-antennal can t. And the S eceives the signals diectionally the BS-S lin can t be intefeed by S-MS lin in full duplex odel. It can obtain spatial divesity theeby the BS-S and S-MS can euse fequency at the sae tie to obtain oe fequency esouces. Mentioned above ae the easons why the adaptive euse esouce allocation schee with two antennas has highe syste thoughput copaed to the adaptive euse esouce allocation schee with single antenna. Howeve it is woth to ention that the blue line inceases fistly and then deceases to the sae as the othe lines. The eason is shown in Fig. 3. As the nube of uses inceases the nube of the diect tansission uses inceases. These uses have bette path gain than elayed tansission ones. Using Hungaian algoith it assigns oe subcaies to diect uses and less to elayed uses. Then the elayed uses affect the syste thoughput less by the adaptive euse esouce allocation schee with two antennas. If the nube of uses inceases lage enough the subcaies ae all assigned to the diect tansission use the thee line will coincide in Fig.. The Fig. 4 is the esult of the thoughput fo the adaptive euse esouce allocation schee using two antennas unde the situation in which both and incease. It can be shown that both and can ipact on the syste thoughput yet affect geate. Fig. The syste thoughput of the thee schees Fig. 3 The thoughput analysis fo the adaptive euse & two-antenna schees Fig. 4 The syste thoughput of the adaptive euse & two-antenna schee 6 Conclusions In this pape we pesent an AMC-based dynaic esouce allocation schee in adaptive fequency eused OFMAelay netwos by using the F elay with two indepent antennas. With the spatial ultiplexing the esouce in BS-S lin can be eused by S-MS lin well. Then with the help of AMC BS and S can euse the sae subcaies. The esults show that ou poposed suboptial algoith can solve the -P had optial algoith and ipove the syste thoughput copaed with othe schees. To p. 7

7 Issue CHAG Ling-jun et al. / istibuted intefeence suppession schee fo ulti-cell uplin OFMA 7 adjustable paaetes. The poposed algoith can ipove the pefoance of the taditional distibuted algoith in both spectal efficiency and noalized fequency efficiency. In addition due to the poposed schee is ipleented in distibuted anne it is easy to cobine with fainess citeions. Acnowledgeents This wo was suppoted by the Sino-Swedish IMT-Advanced Coopeation Poject (8FA78) the Canada-China Scientific and Technological Coopeation (FA3) the ational atual Science Foundation of China ( ) the Hi-Tech eseach and evelopent Poga of China (8AAZ) the Fundaental eseach Funds fo the Cental Univesities (9C38). efeences. Saayda C U Mandaya B Goodan J. Picing and powe contol in a ulticell wieless data netwo. IEEE Jounal on Selected Aeas in Counications 9(): Zhu H Zhu J Liu J. on-coopeative esouce copetition gae by vitual efeee in ulti-cell OFMA netwos. IEEE Jounal on Selected Aeas in Counications 7 5( 6): Yu W Ginis G Cioffi J M. istibuted ultiuse powe contol fo digital subscibe lines. IEEE Jounal on Selected Aeas in Counications (5): Su H J Geaniotis E. A distibuted powe allocation algoith with adaptive odulation fo ulti-cell OFM systes. Poceedings of the 5th Intenational Syposiu on Spead Spectu Techniques and Applications (ISSSTA 98): Vol Sep Sun City South Afica. Piscataway J USA: IEEE 998: Zhu H Zhu J Liu J. Powe iniization fo ulti-cell OFM netwos using distibuted non-coopeative gae appoach. Poceedings of the IEEE Global Telecounications Confeenc (GLOBECOM 4): Vol 6 ov 9 ec 3 4 allas TX USA. ew Yo Y USA: IEEE 4 : Zhang T Feng C Y Su G et al. Uplin ulti-cell non-coopeative powe allocation gae algoith fo OFMA cellula netwos. Poceedings of the 5th Intenational Confeence on Wieless Counications etwoing and Mobile Coputing (WiCOM 9) Sep Beijing China. Piscataway J USA: IEEE 9 7. won H Lee B G. istibuted esouce allocation though noncoopeative gae appoach in ulti-cell OFMA systes. Poceedings of the IEEE Intenational Confeence on Counications (ICC 6) Vol 9 Jun 5 6 Istanbul Tuey. Piscataway J USA: IEEE 6: Goldsith A J Chua S G. Vaiable-ate vaiable-powe MQAM fo fading channels. IEEE Tansactions on Counications (): Fugenbeg Tiole J. Gae theoy. Cabidge MA USA: The MIT Pess 99. Boyd S Vandenbeghe L. Convex optiization. Cabidge U: Cabidge Univesity Pess 3. Yates. A faewo fo uplin powe contol in cellula adio systes. IEEE Jounal on Selected Aeas in Counications 995 3(7): Salo J el Galdo G Sali J et al. MATLAB ipleentation of the 3GPP spatial channel odel.3gpp. T (Edito: WAG Xu-ying) Fo p. 64 Acnowledgeents This wo was suppoted by the ational atual Science Foundation of China (678) and the Cental Univesities Basic Scientific eseach Special Fund (SWJTU9ZT4). efeences. Wong C Y Cheng S Letaief B et al. Multiuse OFM with adaptive subcaie bit and powe allocation. IEEE Jounal on Selected Aeas in Counications 999 7(): hee W A Cioffi J M. Incease in capacity of ultiuse OFM syste using dynaic subchannel allocation. Poceedings of the 5th Vehicula Technology Confeence (VTC-Sping ) Vol May 5 8 Toyo Japan. Piscataway J USA: IEEE : i I Pa I S Lee Y H. Use of linea pogaing fo dynaic subcaie and bit allocation in ultiuse OFM. IEEE Tansactions on Vehicula Technology 6 55(4): Pa W H Bah S. esouce anageent policies fo fixed elays in cellula netwos. Poceedings of the IEEE Global Telecounications Confeence (GLOBECOM 6) ov 7 ec 6 San Fancisco CA USA. Piscataway J USA: IEEE 6: 5p 5. a W Chang W Chung S Y et al. Tansit Optiization fo elay-based Cellula OFMA Systes. Poceedings of the IEEE Intenational Confeence on Counications (ICC 7) Jun Glasgow U. Piscataway J USA: IEEE 7: ae G Gastpa M Gupta P. Coopeative stategies and capacity theoes fo elay netwos. IEEE Tansactions on Infoation Theoy 5 5(9): Motoola Miyazai. elay syste pefoance fo downlin with latest lin Models. 3GPP Ahed M H Yanioeoglu H. Thoughput fainess and efficiency of lin adaptation techniques in wieless netwos. IET Counications 9 3(7): Miyazai. Capacity-based elay UE selection fo Type II elay. 3GPP Hu Y Q Guo Y H. A Couse in opeational eseach. Beijing China: Tsinghua Univesity Pess. 7 (in Chinese). Guan Z Z Liu Y M. Selection of indepent zeo eleents. Jounal of East China oal Univesity: atual Science 9 (5): 6 (in Chinese). IEEE 86-8_45. IEEE 8.6 Evaluation Methodology ocuent (EM). 9 (Edito: ZHAG Ying)