Special Topic. 1 Introduction. 3 Non Orthogonal Multiple Access Schemes

Transcription

1 DOI:.3969/j. ssn publshed onlne October 8, 26 (ZTE Corporaton, Shengzhen 5857, Chna) Abstract Multple access scheme s one of the key technques n wreless communcaton systems. Each generaton of wreless communca ton s featured by a new multple access scheme from G to 4G. In ths artcle we revew several non orthogonal multple access schemes for 5G. Ther prncples, advantages and dsadvantages are dscussed, and followed by a comprehensve comparson of these solutons from the perspectve of user overload, recever type, recever complexty and so on. We also dscuss the applca ton challenges of non orthogonal multple access schemes n 5G. Keywords 5G; non orthogonal multple access; mmtc Introducton M ultple access scheme s the key technque of wreless communcatons. In 3rd generaton (3G) code dvson multple access s appled. In 4G orthogonal frequency dvson multplex ng access (OFDMA) s employed. In the comng 5G, non or thogonal multple access schemes are hot topcs because they can acheve hgh system capacty. Moreover, massve machne type communcaton (mmtc) s one of the key scenaros for 5G n whch massve connecton s requred. In ths paper, we manly focus on the non orthogonal multple access schemes supportng mmtc whch has the rapdest growng speed and the urgent deploy demand. Several non orthogonal multple access schemes are pro posed for 5G, whch nclude mult user shared multple access (MUSA) []-[4], resource spread multple access (RSMA) [5], sparse code multple access (SCMA) [6]- [8], pattern dvson multple access (PDMA) [9]-[], nterleaver dvson multple access (IDMA) [2], [3], and non orthogonal multple access (NOMA) by power doman [4]. In ths paper, the prncples, merts and demerts of these schemes are dscussed to let read ers have a full overvew on that. 2 Features of 5G 5G has three man techncal features, ncludng enhanced moble broadband (embb), mmtc and ultra relable and low latency communcaton (URLLC). The embb s the evoluton of MBB targetng for hgh data rate and can support hgh mobl ty The mmtc s characterzed by massve connecton wth low cost termnals. Hgh relablty and ultra low latency are the goals of URLLC. Wth the development of Internet of thngs, a large number of termnals wll have access to the network. Therefore, mmtc needs to support one mllon of connectons per square klome ter. The mmtc, whch has the fastest growng speed and the most urgent deployment demand, wll create new chances n 5G. The non orthogonal multple access should support at least mmtc where hgh user overload s the key requrement. In LTE there are several nteractve processes between base staton and termnal before the data s transmtted from term nal to the base staton. Ths makes sense for long tme and con tnuous data transmsson because sgnalng overhead s small by averagng over a long tme. In mmtc each termnal only transmts small data and massve termnals would sporadcally transmt ther data to the base staton. When the same access procedure lke n LTE A s appled, the sgnalng overhead wll be comparably large and the access effcency s very low, thus grant free for mmtc s needed n whch multple term nals can send ther data on the same resource block wthout mult step negotatons wth base staton. 3 Non Orthogonal Multple Access Schemes for 5G Several non orthogonal multple access schemes have been proposed for 5G. Based on ther propertes, they can be catego rzed to dfferent types. Most non orthogonal multple access schemes use spreadng codes. When such schemes have other October 26 Vol.4 No. 4 ZTE COMMUNICATIONS D:\EMAG\26--53/VOL3\F.VFT 6PPS/P

2 predomnant propertes, such as SCMA and PDMA use code matrx to llustrate how multple users share the same resource block, and IDMA uses nterleaver for user separaton, we cate gorze them as other knd of schemes. In the followng jont de tecton denotes message passng algorthm (MPA) based schemes. 3. Non Orthogonal Multple Access Schemes Based on Spreadng Sequences 3.. MUSA MUSA s a non orthogonal multple access scheme operat ng n code doman and power doman. Spreadng code wth short length s appled n MUSA to support a large number of users that share the same resource block. When the number of users s large and the length of the spreadng code s small, t s dffcult to desgn large number of spreadng code wth low correlaton when bnary element of the spreadng code s as sumed. For bnary spreadng code the element of the spreadng code belongs to the set {, }. Only two values are employed n the spreadng code. To overcome ths drawback, non bnary and complex value spreadng code s proposed n MUSA. E ther the real or the mage element of the non bnary spreadng code belongs to the set {,, }, there are nne values for se lecton. Ths provdes much more flexblty of spreadng code desgn. Because the real and mage elements of the spreadng code are, or, the multplcaton operaton can be mple mented by addton operaton whch wll reduce the mplemen taton complexty. Fg. shows the basc features of MUSA, where multple users could transmt data on the same resourc es by usng randomly selected non orthogonal complex spread ng codes wth short length (e.g. 4). In ths example 2 users share 4 resource blocks, and the user overload s 3%. MU SA s always modeled by multple spreadng codes superposed on the same resource block. It can also be modeled by a code I - R - Complex spreadng code set Elements of complex spreadng code Each user randomly pcks one code for spreadng S + S2 + + S2 = C C2 C2 SIC: successve nterference cancelaton Fgure. An example of MUSA wth 3% user overload [4]. Codeword level SIC recever matrx. The code matrx of MUSA wth 3% overload s gven by é + G = ê + + ë ù ú + û In 5G, mmtc s one mportant applcaton scenaro. In ths scenaro MUSA s preferred snce grant free transmsson can be readly supported. A devce termnal autonomously access es the communcaton system wthout base staton (BS) sched ulng. Blnd detecton s appled at BS for MUSA n whch ac tve user, user spreadng code and user channel would not be known before hand. Because the spreadng code length s rela tve short and ts elements have lmted values, BS can gener ate numerous local spreadng codes wth low correlaton. By us ng these local spreadng codes and the receved sgnal, we can closely approxmate the optmal performance of MMSE estma tor. Then the user sgnal wth the hghest sgnal to nterference plus nose rato (SINR) can be detected and decoded. After that user s sgnal s successfully decoded, t can be employed for channel estmaton. After nterference cancellaton, the us er sgnal wth the second hghest SINR s detected and decod ed. Durng ths process no plots or preamble are needed for channel estmaton, whch facltates MUSA applcaton n mmtc because most other schemes rely on addtonal over head for channel estmaton. The blnd detecton for MUSA s verfed over flat fadng channel and mult path fadng channel [3], [5]. The man advantages of MUSA are reflected by hgh over loadng factor, robust blnd detecton and true sense of grant free transmsson. Due to frequency dversty gan acheved, 7% user overload can be acheved by MUSA over mult path fadng channel [5]. User detecton can be carred out wthout the knowledge of the spreadng code. User transmtted sgnal can be appled for enhanced channel estmaton once t has been correctly decoded. Users can transmt ther sgnals ac cordng to ther demand. The possblty of collson due to the same spreadng code appled s small snce large number of the spreadng codes can be accommodated. Successve nterference cancelaton (SIC) based recever s appled for MUSA. It works well when there s SINR dfference among the receved sgnals. However, when the dfference s small, there would be certan performance loss due to error propagaton. Whle there s nherent SINR dfferent n mmtc due to free power control, t s not a so serous problem for the sgnal detecton of MUSA. The SINR dfference s small, so t can be solved by usng more advanced recever, such as jont detecton and decodng scheme RSMA In RSMA (Fg. 2), a group of users sgnals are superposed on the same resource blocks, and each user s sgnal s spread over the entre frequency/tme resource blocks. Dfferent users ZTE COMMUNICATIONS October 26 Vol.4 No. 4 D:\EMAG\26--53/VOL3\F.VFT 6PPS/P2

3 Varable rate encoder Coder Plot nserton TDM plot nserton CP: cyclc prefx IFFT: nverse fast fourer transform OFDM: orthogonal frequency dvson multplexng Fgure 2. RSMA block dagrams [5]. Spreader/ scrambler (a) Sngle carrer RSMA Spreader/ scrambler (b) OFDM RSMA Seral to parallel Optonal CP IFFT sgnals wthn the resource blocks may be not orthogonal. code rate channel codes are employed to acheve large codng gan. Relatve long spreadng codes wth good correlaton prop erty are appled to reduce the mult user nterference. Scram blers can be employed wth the same purpose as the spreadng codes. Interleaver s optonal for RSMA accordng to the sys tem requrements. Dependng on the applcaton scenaros, t ncludes sngle carrer RSMA and mult carrer RSMA. For the former t s op tmzed for battery power consumpton and coverage extenson for small data transactons by utlzng sngle carrer wave forms, very low peak to average power rato (PAPR) modula tons. It allows grant less transmsson and potentally allow asynchronous access. Whle for the latter t s optmzed for low latency access for rado resource connecton (RRC) con nected users (.e., tmng wth enb already acqured) and al lows for grant less transmsson. The advantage of RSMA s that t supports asynchronous and grant less transmsson, so the sgnalng overhead s re duced. The dsadvantage s that ts user overload s lmted when rake recever s appled. By usng advanced recever, such as SIC based recever, the overload can be enhanced. 3.2 Non Orthogonal Multple Schemes Based on Structured Codng Matrx Parallel to seral PAPR modulaton Cyclc prefx e j 2πfct e j 2πfct PAPR: peak to average power rato RSMA: resource spread multple access TDM: tme dvson multplexng ments from a codebook. Wth 6 sparse codewords transmtted over 4 orthogonal resources, the user overload s 5%. The codng matrx of Fg. 3 s gven by é G = ê ë To reduce the mult user nterference and the de tecton complexty, sparse sgnature sequence s ap pled n SCMA for spreadng. User sgnal s modu lated by a codebook n whch multdmensonal modulaton maps of the nput coded bts to the ponts n the multple complex dmensons [6]. By such operaton shapng gan s acheved, whch s clamed as one major property of SCMA. The man dsadvantage of SCMA s ts hgh detecton and de codng complexty even sparse sgnature sequence s appled. The detecton and decodng complexty s even hgher when large sze constellaton and a large number of users are em ployed. And addtonal plots or preambles are needed for mult user channel estmaton, whch may reduce system spectral ef fcency. Because the sze of the codebook s lmted, f two us ers choose the same codeword, collson wll happen. Collson s a serous problem for SCMA, whch lmts ts overload capa blty. For example, wth 6 users transmtted over 4 unts, the user overload s only 5%. Although the overloadng factor can be enhanced by usng longer spreadng codes, the detec ton complexty wll ncrease sgnfcantly snce the sze of the codebook and the searchng space s enlarged PDMA For PDMA, the code n a code matrx s used to defne map png from data to a group of resources. Each element n the code corresponds to a resource n the resource group. PDMA can be detected by SIC type recever. It also can be detected by MPA based scheme n the recever. PDMA s desgned for SIC based recever orgnally. The dfferent dversty orders of dfferent users by carefully desgn the code matrx facltate the mult user sgnal detecton. The user wth the largest dver ù ú û 3.2. SCMA Sparse codebook s appled at SCMA to reduce the detecton complexty. At the same tme jont detecton s employed for SCMA to acheve excellent performance. The codewords are composed of mult dmensonal com plex symbols, and the codewords n the same codebook have the same sparse pattern. Sparse codeword mappng utlzes low densty spreadng and could be referred to as sparse spreadng. At the recever, teratve mult user detecton based on MPA s used. Fg. 3 shows an exam ple of SCMA, where the coded bts of a data stream are drectly mapped to a codeword wth sparse non zero ele Bt streams are mapped to sparse codewords Codebook Codebook 2 Codebook 3 Codebook 4 Codebook 5 Codebook 6 (,) (,) (,) (,) (,) (,) 6 sparse codewords are transmtted over 4 orthogonal resources MUD: multple user detecton MPA: message passng algorthm Fgure 3. An example of SCMA wth 5% user overload [8]. MUD based on MPA October 26 Vol.4 No. 4 ZTE COMMUNICATIONS 3 D:\EMAG\26--53/VOL3\F.VFT 6PPS/P3

4 sty order s detected frst, and then the user wth the largest d versty order among the remanng users s detected; n ths way, all users sgnals wll be detected. To further mprove the performance of PDMA, jont detec ton based scheme s proposed. In ths case the unbalance weght of each column s nterpreted as the rregular code weght. As we know rregular low densty party check (LDPC) code has better performance than that of the regular one. By carefully desgnng the code matrx wth jont detecton, even better performance can be obtaned by PDMA compared wth regular code matrx (for example non orthogonal multple ac cess wth low densty sgnatures can be regards as regular code). The man dsadvantage of PDMA s ts low user overload (us er overload s defned by the number of user over the resource block that all users share). It s dffcult to acheve overload of 4% wth the 4 row code matrx (when the row of the code ma trx s K, the largest user number t supported s 2 K []).The complexty s hgh for hgh order modulaton when jont detec ton scheme s appled. Addtonal plots or preamble are need ed for channel estmaton. Because the number of patterns s lmted, there s hgh probablty of collson when users are al lowed to randomly select the patterns. 3.3 Non Orthogonal Multple Schemes Based on Interleaver IDMA was proposed by [2], [3], n whch users are sepa rated by dfferent nterleavers. rate channel decodng s appled and the coded bts are repeated multple tmes to n crease the SINR after accumulatng the receved sgnals. After channel codng and repetton, nterleaver s employed to make the transmsson bts randomly dstrbuted. A block dagram of IDMA s shown n Fg. 4 where C represents channel encod ng, S denotes repetton and π s the nterleaver. The strategy of user separaton for IDMA s dfferent from other non orthogo nal multple access schemes. Interleaver s used for user sepa raton and the length of the nterleaver s very large (the length of the nterleaver equals to the number of the bts after channel codng and repetton), thus ths provdes good base for a large number of users access by usng IDMA. It s reported that 64 users can be supported by IDMA whch share the same re source block [2]. Ths goal can never be acheved by other non orthogonal multple access schemes at present. Transmtter for user d C Transmtter for user K d K C Fgure 4. IDMA block dagram [3]. S S π x π K x K Multple access channel At the recever sde each user s sgnal s detected, demodu lated and de nterleaved accordng to ts own nterleaver pat terns. The soft nformaton of decoded bts s nput to elementa ry sgnal estmator (ESE) for soft nformaton updatng. After soft nformaton updatng new soft nformaton s nput to the decoder for channel decodng agan. Several teratve detec tons between ESE and channel decoder are needed to acheve the best performance. The detecton and decodng complexty does not ncrease exponentally wth the user number and total spectral effcency. The complexty ncreases lnearly, whch s also dfferent from other non orthogonal multple access schemes whch use jont detecton and decodng scheme. The man advantages of IDMA are ts hgh user overload and excellent performance. And hgh spectral effcency can be acheved by IDMA (as hgh as 8 b/s/hz). The performance gap between IDMA smulaton result and the system capacty bound s almost the same from the spectral effcency b/s/hz to 8 b/s/hz (ths means the detecton and decodng scheme s very robustness) [2]. These two merts are seldom acheved by other non orthogonal multple schemes smultaneously. The man dsadvantage of IDMA may be ts large decodng complexty and decodng latency, especally when a large num ber of users are supported. The reason s that when large num ber of teratve detecton and decodng are needed wth the n creasng of user number. For example, tens of channel decoder procedures are needed n the sgnal detecton and tens of nter actve actons between channel decoder and ESE detector are requred. Thus hgh convergence algorthm s needed n the sgnal detecton for IDMA n future. To solve the problem of large decodng complexty and decodng latency, nterleaver patterns can be pre allocate to small number of users,.e., the relatvely small pool sze, so that the complexty of blnd decod ng and channel decodng latency can be mantaned below cer tan level. Another dsadvantage s that addtonal plots or long preamble s needed to estmate the users channels. 3.4 Non Orthogonal Multple Access (NOMA) Scheme Based on Power Doman Dvson Mult user sgnals can be superposed together n NOMA. In NOMA, capacty or throughput mprovement can be expected by sharng the same rado resources among multple user equpments (UEs) as shown n Fg. 5a and Fg. 5b. A typcal applcaton scenaro of NOMA s that a cell center user and a cell edge user are servced by NOMA. Due to small path loss of cell center user, n the sgnal detecton t s detected frst and the sgnal of cell edge user s treated as nterference. In the sgnal detecton of cell edge user, the sgnal of cell center user s detected and decoded frst. Then the sgnal of the cell center user s cancelled from the receved sgnal and sgnal of cell edge user s detected and decoded. The man advantage of NOMA s that excellent performance can be acheved when a cell center user and cell edge user are scheduled wth moderate computatonal complexty (SIC detec 4 ZTE COMMUNICATIONS October 26 Vol.4 No. 4 D:\EMAG\26--53/VOL3\F.VFT 6PPS/P4

5 Table. Summary of dfferent non orthogonal multple access schemes Base staton Cell center user Cell edge user (a) NOMA transmsson Multplexng doman MUSA Spreadng RSMA Spreadng/ scramble SCMA Codebooks PDMA Pattern IDMA Interleaver NOMA Power Fgure 5. NOMA block dagram. Strength of cell edge user sgnal Strength of cell center user sgnal (b) Sgnal strength for NOMA NOMA: non orthogonal multple access tor s always appled). And a user overload of 2% s easly acheved. The man dsadvantage of NOMA s that there s re strcton on the scheduled users. Usually a cell center user and a cell edge user should be scheduled on the same resource block. When two cell center users or two cell edge users are scheduled and SIC type recever s appled, there s perfor mance loss because one user always has low SINR due to nter ference from another user s sgnal. The NOMA s desgned for embb orgnally. Thus when t s appled for mmtc, the re ceved SINR would not be hgh and the number of supported users s very lmted (two or three users are supported on the same resource block, whch s much smaller than other non or thogonal multple access schemes). And addtonal plots or long preamble s needed to estmate the users channels. A summary of these non orthogonal multple schemes are shown on Table. They are compared n terms of multplexng doman, user overload, recever type, recever complexty and so on. Among these schemes MUSA acheves a good balance between performance and complexty, such as hgh user over load, low mplementaton complexty and flexble n grant free transmsson. 4 Applcaton Challenges of Non Orthogonal Multple Access Schemes n 5G Followngs are the requrements for the non orthogonal mul tple access schemes. These factors should be consdered when we desgn the non orthogonal multple access schemes. 4. Coverage Coverage s an mportant ssue for mmtc snce termnals may dstrbute over a large area, thus t s crucal for non or thogonal multple access schemes to support termnals wth df ferent receved power due to path loss. And the non orthogonal multple access schemes should have the ablty of robustness to the hgh nterference. To ncrease the coverage, low code rate channel codng or large spreadng factor could be consd ered. Hgh effcency power amplfer s appealng for coverage User overload Recever type Recever complexty Grant free transmsson Hgh SIC Users can randomly pck up spreadng sequence Raker or SIC Power control needed MUSA: mult user shared multple access RSMA: resource spread multple access SCMA: sparse code multple access PDMA: pattern dvson multple access Mddle Jont detecton Hgh Codeword for each user s predefned and known at BS. Codeword collson s a problem due to lmted number of codewords Mddle SIC or jont detecton for SIC Hgh for jont detecton Pattern s predefned and known at BS. User collson s a problem due to lmted number of patterns IDMA: nterleaver dvson multple access NOMA: non orthogonal multple access SIC: successve nterference cancelaton BS: base staton extenson, whch requres transmt sgnals wth low PAPR. 4.2 PAPR When the non orthogonal multple access scheme s appled for uplnk, PAPR should be consdered to ncrease the trans msson effcency and reduce the transmsson power thus save the battery lfe. The battery lfe s desred to be years for mmtc, so t puts a bg challenge on the non orthogonal multple access scheme. The sgnal of the non orthogonal mul tple access schemes whch have low PAPR wll be preferred n practcal mplementaton. Fltered π/2 bnary phase shft keyng (BPSK) and Gaussan fltered mnmum shft keyng (GMSK) have good property of low PAPR and are employed for PAPR reducton n RSMA [6]. 4.3 Implementaton Complexty The mplementaton complexty ncludes two parts: transmt ter mplementaton complexty and recever mplementaton complexty. Because mult user detecton s carred out at re cever sde, whch has the hghest complexty over the entre sgnal processng chan, the man mplementaton complexty s at the recever sde. Two types of recevers are always ap pled for non orthogonal multple access schemes: SIC based recever and jont detecton based recever. The former can acheve a good balance between performance and complexty. As the number of user ncreases, the complexty only ncreases lnearly. Whle t suffers performance loss n some cases, such as the path losses among dfferent users are the same. Jont de tecton based recever acheves excellent performance at the Hgh Iteratve detecton and decodng Hgh * Interleaver patterns are known at BS SIC Grant based * Unlke jont detecton scheme whose complexty ncreases exponentally as the number of the users and spectral effcency ncreases, the complexty of IDMA only lnear n creases wth the number of users and the spectral effcency. The hgh complexty s due to large number of teratve detecton and decodng. October 26 Vol.4 No. 4 ZTE COMMUNICATIONS 5 D:\EMAG\26--53/VOL3\F.VFT 6PPS/P5

6 cost of hgh computatonal complexty. Although by some de sgns, such as sparse codng matrx, the decodng complexty s reduced sgnfcantly, however, as the constellaton sze and the number of users ncrease, the decodng complexty grows exponentally. Ths bottleneck should be solved before such scheme s employed n practcal systems. 4.4 Combnaton wth Multple Input Multple Output (MIMO) By applyng MIMO technque large system capacty or hgh transmsson/recever relablty can be acheved. It had been proved that MIMO s a very effectve technque n wreless communcaton systems. The non orthogonal multple access schemes should be amable for MIMO. As the frst step, SISO s assumed n the research of the new non orthogonal multple access schemes. However, compatblty wth MIMO should be consdered n the next research step. 4.5 Flexblty The non orthogonal multple access schemes should have flexblty. It can change ts parameters to support dfferent use scenaros. For example, n some cases hgh user overload s the system desgn target, whle n other cases coverage s the most mportant factor. Ths mposes requrements on the non orthogonal multple access scheme desgn. By changng the pa rameter of the non orthogonal multple access schemes, dffer ent targets can be acheved. Another example s that non or thogonal multple access schemes should support both mult carrer system and sngle carrer systems to facltate ts appl caton scenaros. 5 Concluson Ths artcle revews the man non orthogonal multple ac cess schemes for 5G. Ther prncples and unque propertes are dscussed. MUSA can support hgh user overload wth low mplementaton complexty and s more sutable for grant free transmsson. RSMA s sutable for sngle carrer system and mult carrer system. It has good property of large coverage. SCMA can acheve addtonal shapng gan and PDMA has the flexblty n the patterns desgn. IDMA can accommodate very hgh user overload and support hgh spectral effcency at the cost of large decodng complexty and decodng latency. NO MA works well for large SINR dfference among the non or thogonal multple users. At the same tme they have ther own dsadvantages. It s mportant to ntegrate the advantages of df ferent schemes to make the fnal desgned scheme fulfll the challengng requrements of comng 5G. References [] Dscusson on Multple Access for New Rado Interface, 3GPP R 62226, Apr. 26. [2] Z. Yuan, G. Yu, W. L, Y. Yuan, and X. Wang, Mult user shared access for n ternet of thngs, n IEEE Vehcular Technology Conference, Nanjng, Chna, May 26, pp -5. do:.9/vtcsprng [3] Recever Implementaton for MUSA, 3GPP R 6427, May 26. [4] Contenton Based Non Orthogonal Multple Access for UL mmtc, 3GPP R 64269, May 26. [5] Resource Spread Multple Access, 3GPP R 64688, May 26. [6] M. Taherzadeh, H. Nkopour, A. Bayesteh, H. Balgh, SCMA codebook de sgn, n IEEE Vehcular Technology Conference, Vancouver, Canada, Sept. 24, pp.-5, do:.9/vtcfall [7] H. Nkopour and H. Balgh, Sparse code multple access, n IEEE Internaton al Symposum On Personal, Indoor And Moble Rado Communcatons, London, UK, Sept. 23, pp do:.9/pimrc [8] Future Moble Communcaton Forum. (26, Jul. 7). 5G whte paper v2., part d alternatve multple access v [Onlne]. Avalable: forum. org/dl/56/whtepaper.rar [9] Canddate Soluton for New Multple Access, 3GPP R 63383, Apr. 26. [] X. Da, S. Chen, S. Sun, et al., Successve nterference cancelaton amenable multple access (SAMA) for future wreless communcatons, n Proc. IEEE In ternatonal Conference on Communcaton Systems, Macau, Chna, Nov. 24, pp do:.9/iccs [] X. Da, Successve nterference cancellaton amenable space tme codes wth good multplexng dversty tradeoff, Wreless Personal Communcatons, vol. 55, no. 4, pp , Dec. 2. do:.7/s [2] P. L, L. Lu, K. Wu, and W. K. Leung, On nterleave dvson multple ac cess, n IEEE Internatonal Conference on Communcatons, Pars, France, Jun. 24, pp do:.9/icc [3] P. L, L. Lu, K. Wu, and W. K. Leung, Interleave dvson multple access, IEEE Transactons on Wreless Communcatons, vol. 5, no. 4, pp , Apr. 26. do:.9/twc [4] Y. Sato, Y. Kshyama, A. Benjebbour, et al., Non orthogonal multple access (NOMA) for cellular future rado access, n IEEE Vehcular Technology Con ference, Dresden, Germany, Jun. 23, pp. -5. do:.9/vtc Sprng [5] Recever Detals and Lnk Performance for MUSA, 3GPP R 6644, Aug. 26. [6] Resource Spread Multple Access, 3GPP R 66359, Aug. 26. Manuscrpt receved: Bographes YAN Chunln (yan.chunln@zte.com.cn) receved hs PhD degree from Unversty of Electronc Scence and Technology of Chna (UESTC), Chna n 24. He worked at DOCOMO Bejng communcatons lab from 25 to 26. Snce 26 he has been wth ZTE Corporaton. He has publshed tens of papers n IEEE ICC, Globecom, VTC, PIMRC and other nternatonal conferences. Hs man research nterests n clude synchronzaton, bnary and non bnary channel codng, MIMO detecton and non orthogonal multple access technque for 5G. YUAN Zhfeng (yuan.zhfeng@zte.com.cn) receved hs MS degree n sgnal and n formaton processng from Nanjng Unversty of Post and Telecommuncatons (NUPT), Chna n 25. He has been worked wth the Wreless Technology Ad vance Research Department of ZTE Corporaton snce 26 and the leader of the team for new mult access (NMA) for 5G wreless systems snce 22. Hs research nterests nclude wreless communcaton, MIMO systems, nformaton theory, mult ple access, error control codng, adaptve algorthm, and hgh speed VLSI desgn. LI Wemn (l.wemn6@zte.com.cn) receved hs master degree from NUPT, Chna. He joned n ZTE Corporaton n 2, and s responsble for technology research of power control and nterference control n wreless communcatons. Hs current re search focuses on multple access technology for 5G system. YUAN Yfe (yuan.yfe@zte.com.cn) receved hs master degree from Tsnghua Un versty, Chna, and PhD from Carnege Mellon Unversty, USA. He was wth Alca tel Lucent from 2 to 28, workng on 3G/4G key technologes. Snce 28, he has been wth ZTE as the techncal drector of standards research on LTE advanced physcal layer and 5G new rado. Hs research nterests nclude MIMO, channel cod ng, resource schedulng, multple access, and NB IoT. He was admtted to Thou sand Talent Plan Program of Chna n 2. He has extensve publcatons, nclud ng two books on LTE Advanced. 6 ZTE COMMUNICATIONS October 26 Vol.4 No. 4 D:\EMAG\26--53/VOL3\F.VFT 6PPS/P6