Non-Orthogonal Multiple Access Zekun Zhang† , Haijian Sun† , Xianfu Lei‡ †
Utah State University, Logan, UT Southwest Jiaotong University, Chengdu, China Email: † {
[email protected],
[email protected]}, ‡
[email protected] ‡
I. S YNONYMS NOMA, power domain NOMA, PD-NOMA II. D EFINITION
IV. F OUNDATIONS NOMA SIC of UE-2 signal
Power
Non-Orthogonal Multiple Access (NOMA) is a multiple access method proposed by NTT DoCoMo for the next generation (5G) mobile network. NOMA is a type of nonorthogonal multiplexing where users are multiplexed in the power domain, which is not sufficiently utilized by previous wireless mobile systems. Specifically, signals from different users are superposed at transmitter side and separated at receiver by using successive interference cancellation (SIC). Note that in some articles “NOMA” is used as a general name for any scheme that multiplexes users in a non-orthogonal manner, instead of the specific scheme described above. In such articles NOMA proposed by NTT DoCoMo normally is renamed as power domain NOMA (PD-NOMA).
(MUST)”, which is a downlink version of NOMA, is approved (3GPP 2015a). More details on MUST can be found in (3GPP 2015b). Though initially proposed as a downlink scheme, existing works show that NOMA can also be applied in uplink (Zhang Z et al 2017; Takeda T and Higuchi K 2011) and the standardization of uplink NOMA is currently under evaluation (3GPP, NTT-DOCOMO 2016).
UE-1 signal decoding
UE-1
UE-2
UE-2 signal decoding
UE-1
UE-2
Frequency
BS OFDMA UE-1 signal decoding
III. H ISTORICAL BACKGROUND Power
Multiple access (MA) scheme has been regarded as a key technology to distinguish each generation of wireless communication systems in the past decades. In the existing wireless systems, orthogonal multiple access (OMA) has been widely used from first generation (1G) mobile communication system to 4G. Well known frequency division multiple access (FDMA) for 1G, time division multiple access (TDMA) for 2G, code division multiple access (CDMA) for 3G and orthogonal frequency division multiple access (OFDMA) for 4G, are all primarily OMA schemes. In these OMA schemes, users are multiplexed orthogonally in either the frequency, time or code domain so that mixed signal can be separated at receiver with low complexity. While OMA can effectively minimize inter-user interference with a relatively low implementation complexity, its spectral efficiency needs to be further improved in order to meet the requirements in next generation (5G) wireless mobile network such as supporting ultra high data throughout, massive connectivity, low latency, etc. To accommodate such demands, NTT DoCoMo proposed a downlink NOMA scheme (Higuchi K and Kishiyama Y 2012; Saito Y et al 2013; Benjebbour A et al 2015)where multiple users are multiplexed in the power domain at transmitter side and de-multiplexed at receiver side by using SIC (Mazen O.H et al 2003). The standardization of NOMA starts from being proposed to 3GPP LTE Release 13 (3GPP 2014). Then a study item (SI) under the name of “multiuser superposition transmission
UE-1 UE-2 UE-1 UE-2 signal decoding
Frequency
BS
UE-2
High
Low Normalized Channel Gain
Fig. 1. Simple comparison between basic downlink NOMA and OMA (OFDMA)
Fig. 1 illustrates the downlink NOMA with a simple one base station (BS) and 2 user equipments (UEs) case. The transmission bandwidth allocated to two UEs is assumed to be 1 Hz. In NOMA, signals for 2 UEs are superposed by using Superposition Coding (SC) (Cover T 1971) at transmitter side. Denoting the signal for UE-i, i ∈ {1, 2}, as xi where E[|xi |2 ] = 1, the coded signal at transmitter is p p x = P1 x 1 + P2 x 2 . Pi is the transmit power allocated to UE-i and the sum of Pi is restricted to P , i.e., P1 + P2 = P . In this example Pi is set as P1 = 0.2P and P2 = 0.8P . Then the received signal at UE-i is yi = hi x + wi , where wi denotes that Gaussian noise including inter-cell interference and the power density of wi is N0 . hi is the complex channel coefficient from BS to UE-i. In downlink
NOMA, SIC is implemented at UEs and the optimal decoding order is in the order of increasing channel gain normalized 2 by noise and inter-cell interference, i.e., |hNi0| (Saito Y et al 2013; Benjebbour A 2015). In the 2-UE case above, assuming |h1 |2 |h2 |2 N0 > N0 , UE-2 does not need to perform SIC since it comes first in the decoding order. Therefore, UE-2 decodes its received signal x2 directly treating x1 as interference and the throughput of UE-2 can be represented as P2 |h2 |2 . R2 = log2 1 + P1 |h2 |2 + N0 Different with UE-2, UE-1 first decodes x2 to subtract its component from the received signal y1 . The throughput at UE-1 decoding x2 can be then represented as P2 |h1 |2 R1→2 = log2 1 + . P1 |h1 |2 + N0 2
2
As |hN10| > |hN20| , R1→2 > R2 is guaranteed which means x2 can be completely decoded at UE-1 assuming SIC process at UE-1 is error free. After subtracting x2 from y1 , the throughput of UE-1 decoding its own signal x1 is P1 |h1 |2 R1 = log2 1 + . N0 In comparison with NOMA, the throughput of each UE in OMA (OFDMA) systems shown in Fig. 1 can be derived as P |h1 |2 o P |h2 |2 1 1 R1o = log2 1 + , R2 = log2 1 + , (1) 2 N0 2 N0 assuming the bandwidth and the transmit power are allocated 2 2 2| = 0 to each UE equally. When |hN10| = 20 dB and P |h N0 dB, (Saito Y et al 2013) numerically compared throughput of each UE between NOMA and OMA, and demonstrated that the corresponding gains of NOMA from OMA are 32% and 48% for UE-1 and UE-2, respectively. Besides the numerical example shown here, the performance gain of NOMA is also theoretically analyzed in (Ding Z et al 2014). The gain of NOMA comes from the multiplexing gain which is translated from the channel gain difference between 2 UEs through using SC at transmitter side and SIC at receiver side. Although in NOMA the transmit power allocated to a single UE can be lower than that in OMA, i.e., the transmit power of UE-1 is 0.2P and 0.5P for NOMA and OMA respectively, both UEs can benefit from being scheduled more bandwidth. While NOMA relies on advanced receiver processing ability such as SIC, the expectation for evolution of processing ability of user devices is reasonable, generally following Moore’s law. The merits of NOMA can be summarized as following (Dai L et al 2015; NTT-DOCOMO 2014). 1) Improved Spectral Efficiency: As shown in the example above, NOMA can offer higher spectral efficiency for both UEs. This advantage of NOMA over OMA can also be explained from the perspective of information theory. As proofed in (Tse D and Viswanath P 2005), NOMA with SC at transmitter side and SIC at receiver side can achieve the optimal capacity region of the downlink
broadcast channel. However, OMA is not able to achieve so. 2) Massive Connectivity: Different with OMA, the number of supported users in NOMA is not strictly limited by the amount of available resources and their scheduling granularity. Therefore, NOMA can be used to address the challenges of massive connectivity. 3) Robust Performance Gain: NOMA transmitter does not rely much on instantaneous channel state information (CSI), which requires feedback signaling from UEs, to perform multiplexing. Therefore the robust performance gain can be expected irrespective UE mobility and signaling latency. V. K EY A PPLICATIONS There have been many standardization activities on the implementation of NOMA in 5G wireless mobile networks. Particularly, in (3GPP 2015b) the performance of MUST, the downlink version of NOMA, is comprehensively studied and the conclusions are drawn as • MUST can increase system capacity as well as improve user experience in certain scenarios. • MUST is generally more beneficial when the network experiences higher traffic load. • MUST is generally more beneficial in user perceived throughput for wideband scheduling case, compared to subband scheduling case. • MUST is generally more beneficial in user perceived throughput for cell-edge UEs, compared to other UEs. • MUST-far UEs can be legacy UEs when QPSK is applied to MUST-far UEs or the most two significant bits in the modulation symbol are assigned to far UE. Based on the advantages of MUST listed above, the implementation of MUST in some scenarios of 5G networks can be envisaged such as machine-to-machine (M2M) communications, ultra-dense networks (UDN) and mMTC (Recommendation ITU-R M.2083: IMT Vision 2015; Ding Z et al 2017) where massive connections and the IoT functionality of 5G are required. R EFERENCES [1] Higuchi K, Kishiyama Y (2012) Non-orthogonal access with successive interference cancellation for future radio access. Paper presented at the 9th IEEE vehicular technology society asia pacific wireless communications symposium, Kyoto University, Kyoto, Japan, 23-24 Aug 2012 [2] Saito Y, Kishiyama Y, Benjebbour A et al (2013) Non-orthogonal multiple access (NOMA) for cellular future radio access. Paper presented at IEEE 77th vehicular technology conference: VTC 2013-Spring, Dresden, Germany, 2-5 June 2013 [3] Benjebbour A, Li A, Saito K et al (2015) NOMA: from concept to standardization. Paper presented at IEEE conference on standards for communications and networking, University of Tokyo, Tokyo, Japan, 28-30 Oct 2015 [4] Mazen O.H, Alouini M-S, Bastami A et al (2003) Performance analysis of mobile cellular systems with successive co-channel interference cancellation. IEEE Trans Wireless Commun 2(1):29-40 [5] 3GPP (2014) Justification for NOMA in new study on enhanced multiuser transmission and network assisted interference cancellation for LTE. RP-141936, Dec 2014. [6] 3GPP (2015a) New SI proposal: study on downlink multiuser superposition transmission for LTE. RP150496, Mar 2015.
[7] 3GPP (2015b) Study on downlink multiuser superposition transmission (MUST) for LTE (release 13),” TR36.859, Dec 2015. [8] Zhang Z, Sun H, Hu H (2017) Downlink and uplink non-orthogonal multiple access in a dense wireless network. IEEE J Sel Areas Commun. doi:10.1109/jsac.2017.2724646 [9] Takeda T, Higuchi K (2011) Enhanced user fairness using nonorthogonal access with SIC in cellular uplink. Paper presented at IEEE 74th vehicular technology conference: VTC2011-Fall, San Francisco, 5-8 Sept 2011 [10] 3GPP, NTT-DOCOMO (2016) Initial views and evaluation result on non-orthogonal multiple access for NR uplink. R1-163111, Apr 2016. [11] Cover T (1971) Broadcast channels. IEEE Trans Inf Theory 18(1):2-14 [12] Ding Z, Yang Z, Fan P et al (2014) On the performance of nonorthogonal multiple access in 5G systems with randomly deployed users. IEEE Signal Process Lett 21(12):1501-1505 [13] Dai L, Wang B, Yuan Y et al (2015) Non-orthogonal multiple access for 5G: solutions, challenges, opportunities, and future research trends. IEEE Commun Mag 53(9):74-81 [14] NTT-DOCOMO (2014) DOCOMO 5G white paper. https://www.nttdocomo.co.jp/english/binary/pdf/corporate/technology/ whitepaper 5g/DOCOMO 5G White Paper.pdf. Accessed Jul 2014 [15] Tse D, Viswanath P (2005) Fundamentals of Wireless Communication. Cambridge Univ. Press [16] Recommendation ITU-R M.2083: IMT Vision (2015) Framework and overall objectives of the future development of IMT for 2020 and beyond. Sept 2015 [17] Ding Z, Lei X, Karagiannidis GK et al (2017) A survey on nonorthogonal multiple access for 5G networks: research challenges and future trends. IEEE J Sel Areas Commun. doi:10.1109/jsac.2017.2725519
