A New SA-PNC Scheme for Uplink HetNets

Transcription

1 217 25th European Signal Proceing Conference (EUSIPCO) A New SA-PNC Scheme for Uplin etnet Syed Saqlain Ali, Daniel Catanheira, Adão Silva, Atilio Gameiro DETI, Intituto de Telecomunicaçõe, Univerity of Aveiro Aveiro, Portugal yedaqlain@avitpt, dcatanheira@avitpt, ailva@avitpt, amg@uapt Abtract obile traffic in cellular baed networ i increaing exponentially, mainly due to the ue of data intenive ervice lie video Networ operator are urged to explore new technologie in order to enhance the capacity, data rate and maximizing the utilization of available pectrum reource One effective way to cope with thee demand i to reduce the cell-ize by deploying mall-cell along the coverage area of the current macro-cell ytem In thi paper, we conider the uplin of heterogeneou networ with a number of mall-cell coexit with a macro-cell under the ame frequency band To deal with intertier/ytem interference, we combine ignal alignment (SA) baed precoding at the mall-cell tranmitter in conjunction with phyical networ coding (PNC) at the macro-receiver The joint deign of SA and PNC provide higher ytem degree of freedom (DoF), than the cae where only PNC or interference alignment (IA) i employed individually The reult how that the propoed cheme i robut to inter-tier/ytem interference while allowing to increae the overall data rate, by erving more uer, a compared with the IA baed method eyword Signal Alignment; Phyical Networ Coding; eterogeneou Networ; Small/acro-cell ytem; Uplin; I INTRODUCTION Due to new generation of wirele uer equipment and ubiquitou connectivity of mobile communication, the correponding networ load are increaing in exponential manner, where thi maive mobile internet acce i expected to exceed wired device acce by 22 [1] Thi increaing demand will bring huge challenge in term of networ operational capabilitie and global tandardization Although, the 4G of cellular networ i already operational and ha achieved maturity both from indutry and academia but the future 5G networ will bring ervice demand that the current infratructure are far from being capable to handle [2] There i group of emerging technologie that promie to olve the technical challenge of current and future wirele networ, however, the two attracting the mot attention are: mall-cell [3][4] and aive multiple-input multiple-output (IO) [5], where the former enhance the capacity and data rate by increaing the pectrum re-ue and the later enhance the lin bandwidth efficiency in multi-uer environment by exploiting multi-uer IO on a maive cale Cell reduction concept ha numerou benefit that include the offloading of traffic from macro cellular ytem and they can be Thi wor wa upported by the Portuguee Fundação para a Ciência e Tecnologia (FCT) under PURE-5GNET project UID/EEA/58/213 and FCT grant for the firt author (SFR/BD/94548/213) operated inide the coverage area of macro-cell ytem creating a heterogeneou networ (etnet) [6] and offer great advantage for operator and uer, who get better coverage and higher data rate and can acce new ervice owever, due to expected maive deployment of mallcell and the cot aociated in acquiring the new pectrum licene, the two ytem are liely to coexit under the ame pectrum Thi condition require the deign of efficient interference management technique otherwie if not carefully deigned the mall-cell ytem may caue inter-tier/ytem interference at the macro-cell ytem (ie, the owner of the pectrum licene) [7] One recent and effective olution, to deal with interference in etnet i the interference alignment (IA) technique [8][9] IA i a precoding technique that achieve the maximum degree of freedom (DoF) in interference channel Recently, the author in [9] ued the concept of IA in order to olve the interference problem in etnet It wa hown that only 1-bit of information exchange i enough to achieve the ame diverity order a in the cae where full information i exchanged between the macro and mall-cell ytem Another promiing approach to deal with interference and to enhance the overall throughput i the direct application of networ coding (NC) at the phyical layer nown a phyical networ coding (PNC) [1] Although the original invetigation on PNC wa in the context of wired networ but due to the broadcat nature of the wirele medium, PNC ha the potential to boot the capacity of wirele networ [11] The concept of PNC provide a new perpective about interference, which i interference utilization By uing thi new concept, PNC can ignificantly improve the tranmiion rate a well a tacle interference in wirele networ [12] The ey idea in PNC i to jointly detect and decode the um of independent receive ignal at a given node/relay and then forward the demodulated linear combination to multiple detination imultaneouly PNC ha alo been conidered to deal with interference problem in etnet [13] A new wirele communication technique, referred a ignal alignment (SA), ha been deigned to enable PNC in IO baed networ [14][15] Recalling the concept of IA, SA can be conidered a a pecial cae of IA which can achieve the exce DoF in interfering channel A compared to IA, in SA ignal are uperimpoed intead of interference The main idea behind SA i to enhance the networ capacity by enabling imultaneou tranmiion from multiple IO tranmitter SA baed precoding i performed at the tranmitter in uch a way that the number of dimenion panned by received ignal at a given node/receiver i ISBN EURASIP

2 217 25th European Signal Proceing Conference (EUSIPCO) reduced to the number of antenna ued at the receiver At the end, the relay node can decode the linear combination of the tranmitted data by uing IO detection technique uch a, maximum lielihood (L) criterion or zero-forcing (ZF) followed by PNC mapping [15] In thi paper, we apply SA enabled PNC in order to tacle the interference problem in the uplin of etnet To deal with the inter-tier/ytem interference, we combine SA precoding at the mall-cell tranmitter with PNC at the macro receiver The tranmitter and receiver deign i performed by conidering both full-coordination and limited information exchange between the two ytem Namely, we propoed the joint SA-PNC 2n-Bit approach under the limited information exchange requirement We compared our propoed method to the IA baed method propoed in [16][17], where the joint SA-PNC deign allow the ytem to tae advantage of SA and PNC to utilize the interference a a ueful ignal in order to increae the DoF of the overall ytem by erving more mall-cell uer a compared to IA baed method The ret of the paper i organized a follow: Section II preent the ytem In Section III, the decription of joint SA- PNC cheme i preented in detail In Section IV, we preent the numerical reult of the preented and propoed method with exiting one Finally, concluion are provided in Section V Notation: Bold upper cae letter denote matrice; bold lower cae letter denote vector The operation () tand for the ermitian tranpoe of a matrix, repectively null( A ) denote a matrix whoe column pan the null-pace of matrix ( A ) II SYSTE ODEL We conider a cenario with N mall-cell overlaid within the coverage area of a macro-cell, a hown in Fig 1 The two ytem hare the ame pectrum Namely, we conider N mall-cell each with one acce point (AP), and mallcell uer terminal (UT) The mall-cell AP are connected through a bachaul networ (eg Radio over Fiber) to a central unit (CU) that allow joint proceing of the received ignal at the CU In thi wor we conider the uplin cae, ie the cae where the UT tranmit information to the correponding BS and AP (UT end information to AP n) oreover, an Ethernet lin connect the two ytem in order to enable limited coordination between them Only digital pacet are exchanged in thi lin intead of analog ignal, a hown in Fig1 In the following, the macro-cell BS erve one macro uer terminal denominated by UT and the mallcell UT (AP) by UT (AP n ), {, 1, }, n {, 1, N} Thi convention i alo conidered for all variable, a zero repreent to the macro-cell terminal (BS and UT) and an index higher than zero refer to mall-cell ytem a hown in Fig 1 The tranmit power of the macro BS and of each mallcell AP i contraint to P m and P, repectively At the macro-cell ytem, we aume that the macro-cell BS ha antenna and the macro UT ha a ingle antenna (1/ 2) The UT tranmit ignal i given by x = Pm d where d denote the UT tranmit ymbol A dicued above, the CU jointly proce the received ignal of all N mall-cell, o we may conider the N mall cell a a big ingle mall-cell with mall-cell UT The received ignal y at the BS may be mathematically expreed a y = g x + Gx + n, Deired ignal = 1 Interference where x, 1 g, G and n, denote the tranmitted ignal at UT, the channel between macro UT and macro-bs, the channel between UT and macro BS, and the zero mean white Gauian noie with 2 variance σ, repectively For the mall-cell ytem, we aume that each UT ha and the CU ha N antenna ( for each AP) The mallcell UT 1 data ymbol d i multiplied by the ( 1) precoder vector V, before tranmiion (1) x = γ Vd, (2) where γ i a normalization contant to enforce the UT tranmit power contraint The element of the data vector d, {, 1, } are drawn from a -QA contellation The received ignal at the mall-cell AP n can be expreed a y = g x + G x + n, (3) n n n = 1 N 1 N N where gn, G n and n denote the channel between the UT and AP n, the channel between the UT and the AP n and zero mean white Gauian noie with 2 variance σ, repectively The propoed method require the nowledge of channel G at the mall-cell UT, ee ection III Thi channel may be acquired by litening to the pilot ignal broadcated by the macro BS, ince the macrocell mode of operation i time-diviion duplex (TDD) and mall-cell UT are low mobility terminal 1 and then the channel G can be conidered a quai-tatic which reduce the overhead required for it etimation 1 The terminal aociated to the mall-cell are mainly indoor/pedetrian uer ISBN EURASIP

3 217 25th European Signal Proceing Conference (EUSIPCO) AP 1 UT (1,1) UT (,1) CU AP N UT (,N) Digital Lin UT (1,N) Deired ignal Interfering ignal Bachaul networ Digital Lin acro-cell Small-cell Fig 1 Sytem odel: Small-cell within coverage area of macro-cell III JOINT SA AND PNC DESIGN A detailed SA-PNC deign that can wor for the IO uplin cenario in Fig 1 include two component: a precoding cheme at the mall-cell UT, and a decoding approach at the macro BS Firtly, we preent the deign of the precoder baed on SA at the mall-cell UT Then we decribe the decoding proce done at the macro BS and how we incorporate the PNC baed feedbac ignal at the CU before performing the decoding proce Finally, the deign of ubpace matrix i preented conidering limited information exchange approach The ey idea i to tae advantage of joint SA and PNC to utilize the interference a a ueful ignal for mall-cell ytem A Precoding at Small-cell UT To perform SA baed precoding at the mall-cell UT, let ( 1) S be the ( 1) dimenional matrix, that denote a target ubpace at the macro BS, and let the mallcell precoder be given by V = G (4) 1 S, The ey idea behind the ue of SA i to align all mall-cell ignal at the macro BS and then intead of removing the mall-cell ignal, by projecting the received ignal into the orthogonal complement of the interference, a linear combination of the aggregated ignal i decoded After performing SA baed precoding mentioned in equation (4), the macro BS received ignal i given by y = g x, + A γ d + n = 1 Deired Signal Interference A dicued previouly, we want to recover the linear combination of mall-cell ymbol γ1d1+ γ2d2 + + γd at the BS Without lo of generality we will conider in the following that γ 1 γ2 γ To enforce the power contraint x x P, for all = 1,,, we mut have (5) γ γ = tr P ( V V ) In the following, we conider the cae that γ, {, 1, } are all different, we call thi cae - Group where the normalization contant i ettled a γ = γ, {,, }, (7) 1 B ZF Detection at acro BS To decode the deired ymbol d and a linear combination of mall-cell data ymbol γ1d1 + γ2d2 + + γd, the macro BS applie a equalizer W to the received ignal Thee information bit can be decoded by uing appropriate ignal proceing technique uch a L criterion, ZF and minimum mean quare error (SE) In thi context, in order to recover the digital pacet d and γ1d1 + γ2d2 + + γd the BS firt applie the equalizer ( ) 1 W = A A A, (8) to the received ignal y, where A = [ g, S ], which reult in the recovered ignal 1/2 = d γ + = 1, (6) Wy, d W n, (9) From equation (9), it i hown that the macro BS can eaily recover it deired ymbol ( d ) and a linear combination d = γ1d1 + γ2d2 + + γ d of the mall-cell ytem ymbol The ignal d i the uperpoition of -QA ignal ( d, {, 1, } ), whoe dimenionality depend on the parameter γ, 1,,, which are elected by taing into account two objective Firt, i to minimize the bit error rate (BER) and econd i to minimize the data rate requirement for the Ethernet lin For the -Group cae, where all the parameter γ, 1,, are different and then the dimenionality will be on the order of C Deign of the ubpace matrix S (2n-Bit ethod) ere, we preent the deign of ubpace matrix for our propoed joint SA-PNC 2n-bit cheme under minor feedbac requirement The optimization of the alignment ubpace will improve the etimate of the data ymbol and conequently of the feedbac digital ignal The optimum i to et matrix S equal to the null pace of the macro-channel, ie S = null( g ) owever, thi require the exchange of 2( 1) real between the two ytem In order to achieve a commitment between performance and inter-ytem information-exchange requirement, we propoe a joint SA-PNC 2n-bit method where only 2( 1) bit ISBN EURASIP

4 217 25th European Signal Proceing Conference (EUSIPCO) are hared The quantized verion of the ubpace matrix i exchanged between them Therefore, the optimum target ubpace matrix with 2n-bit ( n bit for the real and n bit for the complex part, where n = 1, 2, 3, ) i given by S q Q Q = f ( Re{( S )}) + jf ( Im {( S )}), (1) where f Q () denote a quantization function, the Re {} and Im {} are the real and imaginary part of ubpace matrix S In thi manucript, for the ae of implicity, we conider only uniform quantizer Notice that for thi cae the macro-cell channel g i alo quantized, by taing into account Sq = null( g, q ), where g, q i a quantized verion of g For thi cae the interference and the UT ignal are cloe to orthogonal D PNC apping and Decoding at the CU After decoding the mall-cell ymbol, the macro BS perform a hard deciion to map the analog ignal d = γ1d1+ γ2d2 + + γ d to the correponding contellation point Afterward, the macro BS perform the mapping of thi contellation point to a binary repreentation and finally relay it to the CU via Ethernet lin between macro BS and the CU The number of bit required to repreent the contellation point will be log 2 ( ) After receiving the digital vector ent by the macro BS, via the Ethernet lin, the CU ha acce to a ytem of N + 1 equation, N from the received ignal y n and 1 from the digital ignal ent through the Ethernet lin A the CU receive a data tream from the UT and ( 1) from the mall-cell terminal it mut be able to recover the ( 1) + 1 data tream from the N + 1 available equation which ettle the requirement N + 1 > ( 1) + 1 (11) For the IA cheme propoed in [16] [17] the requirement i N > ( 1) + 1 (12) Therefore, in comparion to the IA method of [16] [17], the propoed joint SA-PNC method may erve one additional mall-cell uer Thi gain i a conequence of the cooperation between the CU and the BS through the bachaul digital lin between them IV SIULATION RESULTS AND DISCUSSION Thi ection provide the performance aement of our propoed and exiting method The propoed Joint SA-PNC 2n-Bit method i compared to the joint SA-PNC fullcoordinated, joint SA-PNC tatic and IA baed method propoed in [16], [17], where only IA wa conidered to remove the inter-tier interference oreover, a it can be verified from the numerical reult that the propoed joint SA- PNC 2-bit method almot achieve the optimal performance uing n = 1, therefore by uing n > 1 the performance improvement will be marginal We conider a flat fading Rayleigh channel without patial correlation and independent channel realization A cenario with 2 mall-cell (ie N = 2 ) i conidered We aume that the number of antenna at the BS i 2, each mall-cell AP and UT ha 2 antenna, and a ingle antenna at the UT Furthermore, for the previouly propoed IA baed method propoed in [16] [17], we aume that one mall-cell AP i erving two mall-cell UT while the other AP erve jut a ingle mall-cell UT (a total of 3 UT can be erved for IA-baed method), contrarily to the propoed approach, where each mall-cell AP i erving 2 UT (a total of 4 UT can be erved uing our propoed joint SA-PNC 2-bit method) Therefore, we conider that the CU jointly procee the ignal of 2 AP for the propoed and IA-baed approache of [16] [17] The numerical reult are preented for QPS modulation for all the cheme preented in thi paper Fig 2 BER Performance at the macro-cell Sytem uing QPS modulation Fig 3 BER Performance at the Small-cell Sytem uing QPS modulation Fig2 preent the reult for macro-cell ytem A hown in Fig 2, the BER performance of IA-baed method (tatic, 2-bit and full-coordinated) overlap with the BER performance of the joint SA-PNC cheme, ince the method ued to mitigate the interference for IA baed and joint SA- PNC method are imilar Furthermore, it can be noticed the ISBN EURASIP

5 217 25th European Signal Proceing Conference (EUSIPCO) number of mall-cell UT are not affecting the performance of macro-cell ytem, ie, we get the ame performance for each method irrepective of the number of mall-cell UT Numerical reult for mall-cell ytem are preented in Fig 3, where it can be noticed that the propoed joint SA-PNC 2- bit method provide the ame performace a compared to the joint SA-PNC full coordinated and IA full-coordinated method while allowing to increae the data rate by erving additional uer without any performance degradation Furthermore, Fig 4 preent the reult for the BER veru number of mall-cell UT for our propoed joint SA-PNC cheme For Fig 4, we conider 5 different cenario, where the CU jointly procee the ignal from 1, 2, 3, 4 and 5 AP erving 2, 4, 6, 8 and 1 UT, repectively It can be noticed that the BER curve for the joint SA-PNC coordinated and 2- bit method are flat, it i becaue the fed bac ignal from the Ethernet lin ha much lower probability of error a compare to the direct received ignal at the CU On the other hand, for the joint SA-PNC tatic the BER decreae with the number of UT becaue for a large number of UT the error introduced by the fed bac ignal from the Ethernet lin will be averaged by the large number of direct received ignal at the CU and it impact will be marginal Fig 4 BER Veru Number of mall-cell UT uing QPS modulation V CONCLUSIONS In thi paper, we conidered the uplin of etnet with a et of mall-cell overlaid within the coverage area of a macro-cell The macro-cell equalizer baed on PNC and mall-cell Precoder baed on SA are jointly optimized to achieve the minimum mean quare error It i demontrated that the full-coordinated approach achieve the bet performance but with the highet feedbac requirement To reduce the amount of inter-ytem information exchange, we propoed to quantize the macro-cell channel to align the mall-cell ignal along the quantized macro-cell channel null-pace intead The propoed method i able to achieve cloe to the optimum performance with reduced information exchange between the two ytem (2-bit are enough to achieve cloe to full-coordinated performance) oreover, it i hown that the joint deign of SA and PNC ha the potential to increae the overall ytem throughput by utilizing interference a a ueful ignal and achieve higher data rate when compared with the IA-only approache, ie more mallcell UT can be erved imultaneouly uing our propoed joint method REFERENCES [1] Cico Viual Networing Index, 214 White paper at cicocom [2] QC Li, N uaning, AT Papathaniou and W Geng, 5G Networ Capacity- ey Element and Technologie, IEEE Vehicular Technology agazine, vol 9, no 1, p 71-78, 214 [3] V Chandraehar, J Andrew and A Gatherer, Femtocell Networ: A urvey, IEEE Communication agazine, vol 46 no 9, pp 59 67, 28 [4] T Naamura, S Nagata, A Benjebbour, Y ihiyama, T ai, S Xiaodong, Y Ning and L Nan, Trend in mall cell enhancement in LTE advanced, IEEE Communication agazine, vol 51, no 2, p 98 15, 213 [5] Anqi and L Wang, Uplin Interference anagement in aive IO Enabled eterogeneou Cellular Networ IEEE Wirele Communication Letter, vol 5, no 5, p , 216 [6] A Damnjanovic, A urvey on 3GPP heterogeneou networ, IEEE IEEE Tran on Wirele Communication, vol 18, no 3, pp 1 21, June 211 [7] T Zahir, Arhad, A Naata and oener, Interference management in femtocell, IEEE Communication Survey and Tutorial, vol 15, no 1, pp , 213 [8] C Suh, o and D Te, Downlin interference alignment, IEEE Tranaction on Communication, vol 59, no9, pp , 211 [9] D Catanheira, A Silva, A Gameiro, "Set optimization for efficient interference alignment in heterogeneou networ", IEEE Tran on Wirele Communication, vol 13, no 1, pp , Oct 214 [1] S Zhang, S C Liew, and P P Lam, Phyical-layer networ coding, in Proc AC obicom, 26 [11] F Zhaoxi, W Zunyi, Z Shaozhong and S Jiong, A low complexity phyical-layer networ coding cheme for cellular two-way relaying ytem in Proc IEEE WOCC, 214 [12] Y Bo, Y ongyi, L Ling and F Qiang, Spectral efficient cooperative relaying technique via Phyical Networ Coding, in Proc IEEE Int conference on ignal proceing (ICSP), 28 [13] W Shuai, W Zhigang and Zibo, Coordinated Tranceiver in IO eterogeneou Networ with Phyical-Layer Networ Coding in Proeeding of the IEEE 24th International Sympoium on Peronal Indoor and obile Radio Communication (PIRC), Sept, 213 [14] L Lee, B L Jong-Bu Lim, and C Joohwan, Degree of freedom of the IO Y channel: ignal pace alignment for networ coding, IEEE Tran on Information Theory, vol 56, no 7, pp , July 21 [15] Z Ruiting, L Zongpeng, W Chuan and C Williamon Signal alignment: enabling phyical layer networ coding for IO networing, IEEE Tran on Wirele Communication, vol 12, no 6, pp , June 213 [16] S Sharma, S Chatzinota, and B Otterten, Interference alignment for pectral coexitence of heterogeneou networ, EURASIP Journal on Wirele Communication and Networing, vol 213, no 1, p 46, 213 [17] D Catanheira, A Silva, and A Gameiro, "Null-pace cognitive precoding for heterogeneou networ" IET Communication, vol 8, no 5, pp , 214 ISBN EURASIP