Regionalized Interference Alignment in Two-Tiered Cognitive Heterogeneous Networks

Transcription

1 Jounal of Communications Vol 9, No, Novembe egionalized Intefeence Alignment in Two-Tieed Cognitive eteogeneous Netwoks Jingwen Yang, Li Li*, and Xiaoping Zhou College of Infomation, echanical and Electical Engineeing, Shanghai Nomal Univesity, Shanghai 5, China {lilyxuan, Abstact Intefeence alignment in cognitive netwoks is an efficient way to achieve the cognitive system thoughput gain without degading the thoughout of the pimay system In this pape a novel intefeence management stategy is pesented to use intefeence alignment in cognitive heteogeneous netwoks flexibly Chaacteistics of intefeence distibution in twotieed netwoks ae consideed to divide the cell deployed by this stuctue into two egions Cognitive adio system detemines which egion the pimay uses locate in based on channel state infomation, and suppesses intefeence by designing thei pecodes to achieve intefeence alignment, hence is temed egionalized intefeence alignment Theeby, it can manage intefeence in diffeent situations effectively Numeical esults ae given fo the uplink aveage ates of pimay use and second uses, which shows that egionalized intefeence alignment ensues the pefomance of the cognitive heteogeneous netwoks Index Tems eteogeneous netwoks, cognitive adio, intefeence alignment, egionalization I INTODUCTION Femtocell is a type of small cell usually deployed by esidential uses [] It povides bette coveage and high speed data sevice fo indoo uses via intenet backhaul and shaes the same fequency band with the existing macocell netwok [] As fequency is a scace esouce, utilizing femtocell and intefeence management will become a developing tend fo next geneation wieless cellula netwoks Cognitive adio (C) can sense o captue infomation inteested fom its tansmission envionment, and change its paametes to meet the demands of the communication netwok Second uses (SU) oppotunistically access the idle fequency band licensed to pimay uses (PU), impoving spectum efficiency of whole netwok [] In some cognitive scenaios, SUs have chance to communication only when they detect a spectum hole unused by pimay uses which have the licensed ights to the spectum oweve, thee ae also some types of C paadigms, like undelay, which allow the coexistence of PU tansmission and SU tansmission by adjusting thei tansmission powe so that the intefeence eceived by PU is unde a cetain theshold [] C technique can be a valid way to enhance the ability of measuing, sensing and adapting on both femtocell netwok and macocell anuscipt eceived June, ; evised Novembe 5, *Coesponding autho lilyxuan@shnueducn doi:7/jcm99-5 netwok This pape applies C to heteogeneous twotieed (femtocell/macocell) netwoks Femtocells educe the load on macocell netwok in ode to ensue the quality of sevice (QoS) of macocell use (U), so we conside the U as the pimay use The femtocell use () is egaded coespondingly as the second use (SU) As femtocell netwoks and existing macocell netwok shae spectum, it will cause inevitably high level intefeence when and U communicate ove the same bandwidth Intefeence alignment (IA) [5] is an intefeence management method that can effectively decease the intefeence in the same band IA ovelaps intefeence in the same one signal space at eceiving teminal by pecoding so as to thooughly eliminate the influence of intefeence on expected signals, thus making K-use intefeence channel achieve the maximum degee of feedom (DoF) at multiple eceives simultaneously [5]-[] A numbe of pactical IA schemes in single-tieed system have been developed to date, including maximizing signal to intefeence plus noise atio (SIN) [7], minimizing the leakage intefeence [7], [], and minimizing mean squae eo (SE) [9] All of these algoithms ae developed fo single-tieed K-use intefeence channels, whee each eceive has an expected tansmitte, and othe tansmittes ae consideed as intefeence souces to that eceive Fo example, minimizing the leakage intefeence algoithm poposed in [7] utilizes the channel ecipocity and iteates between the pecodes of tansmittes and eceives epetitively until the leaked intefeence minimizes ecently, IA has become an impotant tool to study intefeence management on two-tieed heteogeneous netwoks []-[] Like the single-tieed netwok, those IA algoithms also can be used in the two-tieed model oweve, unlike the signal-tieed system, the scheme in two-tieed system should deal with both femtocell and macocell intefeence togethe ef [] intoduces an IA scheme that allows the coexistence of an othogonal fequency division multiplexing (OFD) macocell and a cognitive smallcell The optimal linea pecodes maximize the spectal efficiency of the cognitive link as well as peseve the DoF of the macocell tansmission A patial intefeence alignment in heteogeneous netwoks is pesented in [] It equies only patial channel knowledge and simple computation The poposed scheme in [] povides intefeence-fee communications fo macocell uses which suffe fom Engineeing and Technology Publishing 9

2 Jounal of Communications Vol 9, No, Novembe stong intefeence ef [] poposed an intefeence management temed selective IA It judiciously chooses the set of uses to be aligned at each eceive as a subset of the coss-tie intefees, which could eliminate the destuctive uplink macocell intefeence at the femtocell base station so as to enhance the QoS of A tansceive beamfoming design stategy in IO femtocell netwoks was discussed in [] IA algoithms ae used to mitigate the intefeence at macocell uses caused by femtocell base stations The tansceive design algoithms poposed in [] achieve a lage femtocell thoughput gain with only a small sacifice of macocell use thoughput By and lage, all these above algoithms egad singletieed IA schemes as the foundation, and ty to make vaious degees of optimization so as to meet the demands of solving both femtocell and macocell intefeence in the same time Fo example, [] exploits minimizing SE algoithm, and the schemes in [] is vey simila to maximizing SIN algoithm Simila with the above efeences, single-tieed IA schemes ae also applied in this pape, diffeently fom pevious woks, we pesent a new intefeence management stategy called egionalized IA Chaacteistics of intefeence distibution in two-tieed netwoks ae consideed, and the macocell is divided into two egions U is the tue mobile use which appeas in egion I o egion II andomly Cognitive system detemines the active pimay use in which egion based on channel state infomation (CSI) IA scheme is used in both macocell netwok and femtocell netwok to eliminate thei intefeences We analysis the uplink ates of U and to show the advantage of IA The aveage uplink ate of U in the whole cell is also given to eflect the pefomance gain by egionalized IA The emainde of the pape is oganized as follows: the system model is intoduced in Section II In Section III, the position of PU is estimated egionalized IA is used to suppess intefeence in diffeent egions Numeical simulation esults of aveage uplink ae given in section IV Conclusions ae discussed in Section V U egion I BS Fig A cell employed two-tieed netwoks egion Ⅱ U BS Femtocell Use acocell Use acocell Base Btation Femtocell Base Btation II SYSTE ODEL Fig is a cell scenaio of two-tieed cognitive heteogeneous netwoks with thee s and one U As shown in Fig, the coveage adius of macocell base station is, ie, the whole macocell has adius and thee ae egion I and egion II in it The adius of egion II is, which eflects the coveage of femtocell base station () U moves in egion I o egion II andomly s ae confined to egion II elatively When U moves in egion I and communicates with BS (uplink tansmission), will be intefeed because of the coveage of BS is the whole cell oweve, the communication between and will geneate little intefeence on BS at this moment because of the limited coveage of Thus the intefeence of to BS can be viewed as backgound noise When U moves into egion II, the SIN of U deceases apidly, the coss intefeence between and U exists as well, ie, will eceive the intefeence fom U and BS will also eceive the intefeences fom s Specific system models ae as follows A Case :U in egion I U Fig Uplink tansmission model fo case BS(BS) BS() Fig is an uplink tansmission model fo case with one U and seveal s U communicates with BS and communicate with In the case scenaio, the numbe of is denoted by, and each mobile use has N t tansmitting antennas Thee ae N f eceive antennas in both BS and Assuming the communication between and has no intefeence to BS because of the shot coveage of the, but the communication between U and BS will have intefeence on The eceived signal at BS ae given as Eq() y V s n () The eceived signal at is shown as Eq(), y V s V s F i V s n i i i whee ij is the matix with N t ows and N f columns, which epesent the channel matix fom the j th use () Engineeing and Technology Publishing

3 Jounal of Communications Vol 9, No, Novembe (j {,,, }) to the ith base station (i {, } ) as shown in Fig We assume that system has got a pefect CSI s i d denotes the tansmission signal of the ith use, the numbe of message bits tansmitted fom each mobile use is denoted by d, should satisfy d min ( N t, N f ) V i N f d epesents the pecoding matix of the ith use The noise vecto at the base station is denoted by n i N t d, which consists of independent zeo mean Gaussian andom vaiables with E ( n i n i ) = σ I N t I d is the d d identity matix E( ) stands fo the expectation opeato and ( ) efes to the emitian tanspose i i V i s i is the desied signal of eceive, i e, tansmittes and eceives ae connected by solid line in Fig V s epesents the intefeence eceived fom U by, ie, coss-tie intefeence i V i s i is the intefeence eceived fom othe s by, ie, inne-tie intefeence U BS BS BS neaby Assuming that system has got a pefect CSI, we define a theshold value S to eflect the channel envionment of egion II, whee S t (( ) ( )) t( ) is the tace of matix SU is efeed as the channel envionment between U to, SU can be expessed as Eq(5) SU t (( ) ( )) (5) I f S U γs, wh e e γ (, ] is a constant which is detemined by the specific tansmitting powe of, eflects the coveage of, t he n the U is efeed to causing highe intefeence at, which eflect that is closed to, so that U is consideed in egion II If S U < γs, the U is thought as in egion I accodingly B IA in Case Since the intefeence of s to BS can be viewed as backgound noise in egion I, we just conside the intefeence alignment in, Vi is used to align the intefeence in the same subspace at eceive, Vi should satisfy the constaint as Eq() () i Vi j Vj whee i, j=,,,, and i j U N t d is designed to eliminate the intefeence at eceive, U should satisfy the constaint shown as Eq(7) Fig Uplink tansmission model fo case U B Case U in egion II Uplink tansmission model fo case with one U and seveal s is depicted in Fig As U communicates in egion II, SIN of U deceases shaply fo case The coss intefeence between and U exists as well Since the distance among s is vey shot compaed with the distance fom s to BS, the channels between s to BS can be simply viewed as the same The eceived signal at is the same as in case which has been denoted by Eq () The eceived signal at BS shown in Fig, given as Eq () i Vi (7) ank (U V ) d whee ank( ) denotes the ank of matix Afte the intefeence alignment, the eceived signal at in case can be witten as Eq() ~ y F U y F U Vs U V s U i Vi si U n () i In ode to veify what benefits the IA technique eally bings to both U and, we intoduce the uplink tansmission ate in [] to descibe the system pefomance The uplink tansmission ate of U in case is efeed to p and descibed by Eq(9), () i Like case, V s epesents the desied signal and V i s i is the intefeence eceived fom s In section III, IA technique is used to suppess the intefeence at base stations both in case and case p log I ( p V ( V ) )(nn )- (9) and the uplink tansmission ate of is defined as s, shown as Eq(), III EGIONALIZED INTEFEENCE ALIGNENT s log I U QU (U (nn Q )U )- A Estimate the Position of U Fist we descibe how to estimate which egion the U locates in System model consides cognitive netwoks have the ability to sense and monito the U Engineeing and Technology Publishing i, i y ' Vs Vi si n () () whee ( ) - epesent the invese of matix and means the deteminant of a matix Q epesents the powe of desied signal at the eceive, is descibed as Eq()

4 Jounal of Communications Vol 9, No, Novembe Q p V ( V ) () Q epesents the powe of intefeence and noise at the eceive, shown as Eq() Q p V ( V ) () i i i i i i, i whee p is the tansmitting powe of U and p i is the tansmitting powe of the i th (i {,,}) Fig and Fig 5 ae the simulations of uplink ate in case with U and s In Fig, U does not change tansmitting powe and s ise thei powe with the same slope We could see that with the inceasing powe of, the uplink tansmission ate of is inceasing yet the uplink tansmission ate of U stay the same In Fig 5, the powe of s ae fixed and U ise its tansmitting powe, the uplink tansmission ate of is constant while the uplink tansmission ate of U is inceasing Fom Fig and Fig 5 we could see neithe U no is influenced by the inceasing powe of the othe when we employ IA, the simulation esults demonstate that IA successfully emoves the intefeences on BS and in case tansmit ate(bps/z) p(tansmit ate of U) s(tansmit ate of ) tansmit powe of (dbm) Fig Incease powe of in case tansmit ate(bps/z) p(tansmit ate of U) s(tansmit ate of ) tansmit powe of U(dbm) Fig 5 Incease powe of U in case C IA in Case The SIN of U deceases apidly when U is in egion II The coss intefeence between and U will geneate in this scenaio as stated in section II The intefeence geneated by s on BS should be consideed U N t d is designed to eliminate the intefeence at BS, U should satisfy the constaints as Eq() i i i U ( V ) U ( V ) ank( U V ) d i () Afte the intefeence alignment, the eceived signal at BS in case can be witten as Eq() ~ ' ' y U y U Vs i i i U V s U n the uplink tansmission ate of U [] in case is ' - p () log I U Q U ( U ( n n Q ) U ) (5) Q epesents the powe of desied signal at the eceive, is descibed as Eq() Q p V ( V ) () whee Q epesents the powe of intefeence and noise at the eceive, can be witten as Eq(7) Q p V ( V ) (7) i i i i whee p is the tansmitting powe of U and p i is the tansmitting powe of the i th (i {,,}) The uplink tansmission ate of in case is the same as it in case We also do a simulation with U and s like case, p, p, p is inceasing while p is constant in Fig and p is inceasing while p, p, p is fixed in Fig 7 Compaing Fig, Fig 5, Fig and Fig 7, the ate of and U is stable and neithe s no U could intefee each othe eceives All the esults show that IA technique has a good effect on intefeence management both in case and case tansmit ate(bps/z) p(tansmit ate of U) s(tansmit ate of ) tansmit powe of (dbm) Fig Incease powe of in case Engineeing and Technology Publishing

5 Jounal of Communications Vol 9, No, Novembe tansmit ate(bps/z) p(tansmit ate of U) s(tansmit ate of ) tansmit powe of U(dbm) Fig 7 Incease powe of U in case IV AVEAGE TANSITTING ATE Simulations in section III eveal that IA technique ceates a zeo intefeence cicumstance fo uses in both case and case In this section, we put effots on the impovement of system pefomance by applying egionalized IA Fig and Fig 9 show the changes of uplink tansmission ates with the gowth of SN in case and case espectively It is obvious that the ates of both and U incease as the SN inceases, which eflects IA guaantees the pefomance in case and case tansmit ate(bps/z) p(tansmit ate of U) s(tansmit ate of ) SN(dB) Fig Tansmit ate inceases as SN inceases in case tansmit ate(bps/z) p(tansmit ate of U) s(tansmit ate of ) SN(dB) Fig 9 Tansmit ate inceases as SN inceases in case In ode to veify the pefomance enhanced on the whole netwok system moe diectly when egionalized IA is employed, the aveage ates of U and in the whole cell has been discussed As the ate of is the same in case and case, the aveage ate of is equal to s in Eq(), then only the aveage ate of U should be concened about Combining the uplink tansmission ate of U in case and case, aveage uplink tansmission ate of U in the whole cell is defined in this pape as Eq() ' p P p P p P log I U Q U ( U ( n n Q ) U ) - P I U Q U U n n Q U - + log ( ( ) ) () whee P denotes the pobability of U in egion I, is descibed as Eq(9) - P (9) is the coveage adius of BS, is the coveage adius of P epesents the pobability of U in egion II,can be witten as Eq() P () Fist we simulate the Scenaio with one U and thee s in the whole cell In Fig we compae the aveage tansmission ate of U and with diffeent case, including egionalized IA poposed in this pape, IA algoithm poposed in [] and algoithm without IA The s aveage tansmission ate employing egionalized IA and IA ae almost unanimous when SN is smalle than db When SN is moe than db, it can be seen that the s aveage tansmission ate employing egionalized IA is highe than that of IA Compaing the aveage tansmission ate of U, egionalized IA povides about bps/z benefits ove IA As expected, both the aveage tansmission ate of U and employing egionalized IA and IA have significant advantages ove that of without IA Fom Fig we can see that the poposed egionalized IA technique eally guaantees the tansmitting pefomance of both and U aveage tansmit ate(bps/z) U employ egionalized IA employ egionalized IA U employ IA algoithm poposed in [] employ IA algoithm poposed in [] U without IA without IA SN(dB) Fig Aveage ates of both and U Engineeing and Technology Publishing

6 Jounal of Communications Vol 9, No, Novembe inceases in the macocell, the intefeence between femtocells will also be consideed ow to optimize the aveage ate and allocate the powe of s in a easonable way is also the next phase in ou study ACKNOWLEDGENT This wok was suppoted by the Innovation Pogam of Shanghai unicipal Education, NoZZ; the Pogam of Shanghai Nomal Univesity (DZL) The authos would like to thank the eviewes fo thei detailed eviews and constuctive comments, which have helped to impove the quality of this pape BS U EFEENCES [] Fig A macocell coexisting with seveal femtocells Next, we simulate the U s aveage tansmission ate change with the numbe of femtocells Fig shows a macocell coexisting with seveal femtocells The pape assume that all these femtocells keep fom each othes in a easonable distance, so we ignoe the intefeence between these femtocells Thee ae still thee s in evey femtocell in ou simulation scenaio Fig eflects the impact of femtocell s numbe change on the aveage ate of U When the numbe of femtcell incease fom to, the ate of U failed by just unde bps/z in Fig, which indicated egionalized IA has an excellent effect on intefeence management again [] [] [] [5] 5 [] aveage tansmit U(p) aveage tansmit ate(bps/z) 9 [7] 7 5 [] [9] the numbe of femtocell [] Fig U s aveage ate changed by the numbe of femtocell V CONCLUSION [] In this pape, we have developed a egionalized IA that is applicable to two-tieed heteogeneous cognitive netwoks in which femtocell and macocell coexist The macocell is divided into two egions based on the chaacteistics of its intefeence distibution We employ IA technique to suppess the intefeence at eceives both in BS and The simulations show that the poposed intefeence management stategy ensues the pefomance of both U and Futue wok includes consideing moe Us With the numbe of femtocells Engineeing and Technology Publishing [] [] [] V Chandasekha, J Andews, and A Gathee, Femtocell netwoks: A suvey, IEEE Communications agazine, vol, no 9, pp 59-7, Sep W Wang, G Yu, and A uang, Cognitive adio enhanced intefeence coodination fo femtocell netwoks, IEEE Communications agazine, vol 5, pp 7-, Jun V Gadellin, S K Das, and L Lenzini, Self-coexistence in cellula cognitive adio netwoks based on the IEEE standad, IEEE Wieless Communications, vol, pp 5-59, Ap Z Li, J Gu, and Q Liu, Dynamic spectum shaing and use scheduling fo C-IO system based on intefeence alignment, in Poc IEEE th Telecommunications (ICT), Casablanca,, pp -5 V Cadambe and S A Jafa, Intefeence alignment and degees of feedom of the K-use intefeence channel, Infomation Theoy, IEEE Tansactions on, vol 5, pp 5, Aug Z Shakei, A F Chaghooshi, imohseni, et al, Degees of feedom in a thee-use cognitive intefeence channel, IEEE Communication and Infomation Theoy, Tehan,, pp - K Gomadam, V Cadambe, and S A Jafa, Appoaching the capacity of wieless netwoks though distibuted intefeence alignment, in Poc IEEE Global Telecommunications Confeence, New Oleans,, pp - S W Petes and W ealth J Intefeence alignment via altenating minimization, Acoust, Speech, Signal Pocess, Taipei, 9, pp 5- D A Schmdt, C Shi, A Bey, L onig, and W Utschick, inimum mean squaed eo intefeence alignment, in Poc Asiloma Conf Signals, Syst, Comput, Pacific Gove, 9, pp - aso, L S Cadoso, and Debbah, et al, Cognitive intefeence alignment fo OFD two-tieed netwoks, in Poc IEEE th Signal Pocessing Advances in Wieless Communications, Cesme,, pp - S K Shama, S Chatzinotas, and B Ottesten, Intefeence alignment fo spectal coexistence of heteogeneous netwoks, EUASIP Jounal on Wieless Communications and Netwoking, vol, pp -, Feb B Gule and A Yene, Selective intefeence alignment fo IO cognitive femtocell netwoks, IEEE Selected Aeas in Communications, vol, pp 9-5, a Q Z Li, X Gu, Q Li, and Jia, Intefeence alignment fo IO downlink femtocell netwoks, in Poc IEEE 9th WeidangLu Netwoks, Singapoe,, pp - O Seo, C Yang, G Pak, and J Chung, Patial intefeence alignment in heteogeneous netwoks, IEEE ilitay Communications Confeence, San Diego,, pp

7 Jounal of Communications Vol 9, No, Novembe [5] W Noh, K Jang, and Choi, ieachical intefeence alignment fo downlink heteogeneous netwoks, IEEE Tansactions on Wieless Communications, vol, pp , Dec [] X B ao, L Z uan, and Lau, Limited feedback design fo intefeence alignment on IO intefeence netwoks with heteogeneous path loss and spatial coelations, IEEE Tansactions on Signal Pocessing, vol, pp 59-7, ay [7] L uang, G X Zhu, and X J Du, Cognitive femtocell netwoks: an oppotunistic spectum access fo futue indoo wieless coveage, IEEE Wieless Communications, vol, pp -5, Ap [] N Letwiam, P Popovski, and K Sakaguchi, A study of tadeoff between oppotunistic esouce allocation and intefeence alignment in femtocell scenaios, IEEE Wieless Communications Lettes, vol, pp 5-59, Aug [9] S A Jafa, Blind intefeence alignment, IEEE the Jounal of Selected Topics in Signal Pocessing, vol, pp -7, Jun [] L Li and J Zhang, A two-level pecoding disign in cognitive IO system, China Patent, CN75A, ay 7, [] C Zhou, T atnaajah, and Y C Liang, On seconday netwok intefeence alignment in cognitive adio, IEEE New Fonties in Dynamic Spectum Access Netwoks, Aachen,, pp 7- Jingwen Yang obtained the Bachelo of Science degee in and is cuently pusuing S degee pogam in Shanghai Nomal Univesity is eseach inteest is cuently at intefeence alignment algoithms in heteogeneous cognitive adio netwok Li Li eceived the BSEE degee fom Tsinghua Univesity in 95, the SEE degee fom China eseach Institute of adiowave Popagation in 9, and PhD degee in science fom Peking Univesity in 997 Pof Li s eseach inteest is cuently at the wideband spectum sensing, spectum allocation and intefeence alignment algoithms in heteogeneous cognitive adio netwok Xiaoping Zhou eceived the PhD degee fom Shanghai Univesity in e is an associate pofesso in Shanghai Nomal Univesity is eseach inteest is mainly boadband wieless communications Engineeing and Technology Publishing 5