Source-linked AI summary
Non-Orthogonal Multiple Access: Common Myths and Critical Questions
Mojtaba Vaezi, Robert Schober, Zhiguo Ding, H. Vincent Poor
TL;DR
The paper addresses widespread misunderstandings about NOMA and questions surrounding its adoption in 5G and beyond. It develops theoretical downlink and multi-cell foundations, then uses them to examine myths, practical benefits, and open issues. The paper concludes that NOMA’s power allocation is flexible rather than inherently biased toward weaker channels, while practical adoption still requires investigation.
Problem
Widespread myths and unresolved practical questions surround NOMA’s resource allocation, interference, decoding complexity, security, and adoption in 5G and beyond.
Method
The paper develops two-user and arbitrary-user, multi-cell NOMA foundations and uses them to analyze myths, misunderstandings, and critical adoption questions.
Results
NOMA does not inherently allocate more power to the user with the worse channel; power allocation depends on the operating objective and can support fairness.
Takeaways & Limitations
NOMA’s value comes from serving multiple users in one resource block while enabling flexible allocation, but its effective practical adoption requires continued investigation.
Abstract
from arXiv · showhide
Non-orthogonal multiple access (NOMA) has received tremendous attention for the design of radio access techniques for fifth generation (5G) wireless networks and beyond. The basic concept behind NOMA is to serve more than one user in the same resource block, e.g., a time slot, subcarrier, spreading code, or space. With this, NOMA promotes massive connectivity, lowers latency, improves user fairness and spectral efficiency, and increases reliability compared to orthogonal multiple access (OMA) techniques. While NOMA has gained significant attention from the communications community, it has also been subject to several widespread misunderstandings, such as $``\textit{NOMA is based on allocating higher power to users with worse channel conditions. As such, cell-edge users receive more power in NOMA and due to this biased power allocation toward cell-edge users inter-cell interference is more severe in NOMA compared to OMA. NOMA also compromises security for spectral efficiency.}''$ The above statements are actually false, and this paper aims at identifying such common myths about NOMA and clarifying why they are not true. We also pose critical questions that are important for the effective adoption of NOMA in 5G and beyond and identify promising research directions for NOMA, which will require intense investigation in the future.
I. INTRODUCTION AND BACKGROUND
NOMA is presented as a candidate for 5G and beyond because orthogonal allocation struggles with massive, heterogeneous connectivity demands. The paper reviews its power-domain downlink form and uses theory to clarify common misunderstandings and practical adoption questions.
- Motivation: Orthogonal allocation becomes inefficient and infeasible when many sporadically active, low-rate devices each require a separate resource block.One resource block may instead carry data for many such devices.
- NOMA concept: NOMA serves more than one user in the same wireless resource, including a time slot, frequency band, spreading code, or spatial resource.This paper focuses on power-domain NOMA in the downlink.
- Potential benefits: NOMA can improve spectral efficiency and user fairness, while grant-free uplink NOMA can reduce latency, signaling overhead, and terminal power consumption for light traffic.The paper also notes combinations with massive MIMO and millimeter-wave communications for enhanced mobile broadband requirements.
- Motivation: 5G connection density requirements reach 1,000,000 devices per km2, which is 100 times the 4G figure.This scale contributes to the challenge of assigning dedicated resources to every connection.
- Paper scope: The paper illustrates NOMA with two users, extends the analysis to arbitrary users and multi-cell settings, and applies the theoretical basis to myths about allocation and interference.It also poses questions critical to successful practical adoption.
II. DOWNLINK NOMA BASICS: A REVIEW
Downlink NOMA uses superposition coding to transmit users’ signals simultaneously, whereas OMA separates them in time or frequency. The section formulates achievable rates for two-user and K-user orthogonal systems as a basis for comparison.
- Two-user downlink model: In downlink NOMA, superposition coding adds two users’ codewords and transmits the combined signal to both users simultaneously in time and frequency.This is contrasted with TDMA or FDMA, where users occupy separate orthogonal slots.
- OMA rate region: For two-user OMA, allocating fractions τ and 1−τ of time gives rates R1 = τC(γ1) and R2 = (1−τ)C(γ2).The formulation assumes TDMA and defines C(x) as the logarithmic capacity expression.
- Signal model: The received SNR is γi = |hi|2P, where hi is user i’s channel gain, P is base-station transmit power, and noise power is normalized to unity.These definitions specify the parameters used in the rate expressions.
- K-user extension: For K-user single-cell OMA, the resource is divided into K orthogonal portions and user k achieves Rk = τkC(γk).The fractions satisfy a unit-sum resource constraint.
- Multi-cell setting: With different frequencies in adjacent cells, the multi-cell solution in each cell is similar to the K-user single-cell case.This statement does not cover universal frequency reuse.
B. NOMA
NOMA enlarges OMA’s rate region through superposition coding and successive interference cancellation, while flexible power and scheduling support fairness and multiple-user service. Universal frequency reuse makes multi-cell optimization substantially harder because capacity-achieving schemes are unknown.
- Rate-region gain: NOMA enlarges OMA’s rate region using superposition coding at the base station and successive interference cancellation at receivers.Varying the power split can reach boundary rate pairs in the two-user broadcast channel.
- Successive interference cancellation: The stronger-channel user cancels interference and decodes interference-free, while the weaker-channel user treats the other signal as noise.The corresponding rates are R1 = C(αγ1) and R2 = C(ᾱγ2/(αγ2+1)).
- User service: NOMA can theoretically serve as many users as required in one resource block, although practice limits the number served for complexity reasons.This increases the chance that a user is scheduled and can improve fairness.
- Fairness: Flexible power allocation can improve fairness by increasing power for the weaker user and optimizing a weighted sum rate with a larger weaker-user weight.For the weaker user, µ > 1 improves fairness whereas µ < 1 makes it worse.
- Multi-cell interference: Under universal frequency reuse, multi-cell NOMA becomes much more involved, and capacity-achieving schemes are not known.The best described strategy decodes part of inter-cell interference while treating the remainder as noise, combining NOMA and OMA.
C. MIMO-NOMA
MIMO-NOMA clusters multiple users on each beam using SC-SIC, enabling more users to share spatial dimensions. The paper also shows that common assumptions about NOMA power allocation, SIC order, and spectral-efficiency objectives are not generally valid.
- C. MIMO-NOMA: MIMO-NOMA assigns a cluster of users to each beam and applies SC-SIC within each group, while different beams manage inter-cluster interference.Unlike multi-user MIMO, each spatial dimension serves a user cluster rather than a single user.
- C. MIMO-NOMA: MIMO-NOMA can serve a larger number of users and supports massive connectivity by sharing each spatial dimension among multiple clustered users.
- C. MIMO-NOMA: MIMO-NOMA's described clustering solution is not theoretically optimal for the MIMO broadcast channel.
- III. MYTHS AND MISUNDERSTANDINGS ABOUT NOMA: SIC decoding order depends on users' SNR ordering rather than power allocation; in the two-user case, the stronger user decodes the weaker user's signal first.This ordering remains optimal for any α and extends to K-user NOMA.
- III. MYTHS AND MISUNDERSTANDINGS ABOUT NOMA: NOMA's main motivation is serving more users with limited resource blocks, while sum-rate maximization can favor OMA and NOMA gains can vanish for similar channel gains.NOMA may sacrifice sum rate to increase the number of served users or improve QoS and fairness.
F. Myth 6: NOMA is not compatible with FFR
NOMA and FFR can be combined by pairing users across cell-interior and cell-edge regions, although the pairing strategy affects bandwidth use and spectral efficiency.
- FFR assigns separate frequency bands to cell-interior and cell-edge users within each cell.
- NOMA conventionally pairs cell-interior and cell-edge users on the same frequency band to gain over OMA.
- The apparent conflict between FFR’s orthogonalization and NOMA’s non-orthogonal reuse has led some researchers to question NOMA in multi-cell networks.
- Pairing cell-interior and cell-edge users in the cell-edge band combines NOMA with FFR, but makes more users share that band and requires a larger cell-edge bandwidth fraction.
- Alternatively, pairing interior users together and edge users together preserves separate bands, but similar channel gains reduce NOMA spectral efficiency relative to OMA.
G. Myth 7: Decoding complexity of NOMA is prohibitively high for UEs
The paper argues that NOMA decoding is no longer prohibitively complex for advanced UEs, while simpler IoT devices still require less demanding receiver strategies.
- NOMA’s underlying multi-user decoding concept is longstanding, but limited UE processing power previously hindered practical deployment.
- Recent UE advances make interference cancellation practical, and LTE-A NAICS terminals already use cancellation to mitigate multi-cell interference.
- Experimental NOMA trials report complexity within current user-terminal capabilities, and advanced UEs can decode NOMA.
- Interference cancellation remains challenging for simple devices such as low-cost IoT devices.
- IoT devices can be paired with advanced UEs performing SIC, or grouped with interference treated as noise, increasing supported users at the expense of spectral efficiency and decoding simplicity.
- The reduced-efficiency approach may remain acceptable because IoT users usually have very low data-rate requirements.
H. Myth 8: SIC error propagation makes NOMA inviable
SIC error propagation does not make NOMA infeasible in general: multi-user receivers are already used in cellular systems, and experiments support NOMA feasibility under suitable conditions.
- Cellular systems have long handled multiple interfering signals through technologies including CDMA receivers and NAICS UEs.
- Multi-user receivers and NOMA have credible prospects for widespread use, at least under certain conditions.
- Current research and commercial deployments show that SIC implementation is possible with today’s technology.
- SIC is especially beneficial with very low-rate codes and in environments where path loss is high.
- Successful decoding depends on channel disparity between near and far users and on parameters including channel disparity, modulation type, and power allocation.
- USRP experiments confirm the feasibility of SC-SIC in the two-user case.
I. Myth 9: NOMA users must have different channel gains
NOMA does not require users to have different channel gains, although its spectral-efficiency advantage over OMA diminishes when gains are similar. MIMO-NOMA precoding and advanced power allocation can address similar-gain users while meeting QoS requirements.
- NOMA users can have exactly the same channel gains.
- When channel gains are similar, NOMA’s spectral-efficiency benefits compared to OMA diminish.
- For |h1| = |h2|, the NOMA rate region becomes the same as the OMA rate region.
- Spectral efficiency is not the main reason for using NOMA.
- In MIMO-NOMA, BS precoding can degrade one user’s effective channel gain while enhancing another’s, even when their channel gains are similar.
- Cognitive-radio power allocation can strictly guarantee users’ QoS requirements despite similar channel gains.
J. Myth 10: NOMA compromises security and privacy
NOMA’s stronger-user decoding capability does not by itself compromise weaker-user security or privacy. Broadcast-channel exposure also exists in OMA, while upper-layer encryption and physical-layer security can protect NOMA communications.
- Weaker-user security concerns are not unique to NOMA because OMA also has the wireless channel’s broadcast nature.
- Decoding a user’s signal at the physical layer does not imply decoding its message.
- Upper-layer measures such as scrambling bits with a UE-specific C-RNTI can prevent unauthorized message access.
- Other encryption-based solutions can address security and privacy issues when C-RNTI sharing is required.
- Physical-layer security can guarantee security for NOMA broadcast channels even at the PHY layer.
IV. CRITICAL QUESTIONS AND FUTURE OF NOMA
The key remaining NOMA challenges concern system design and implementation under realistic conditions rather than theory alone. Critical questions involve imperfect SIC and CSI, practical clustering, experimental validation, and performance limits in large-scale networks.
- The important remaining NOMA challenges are related to system design and implementation.
- Most NOMA studies assume perfect SIC, making the effect of imperfections on clustered multi-user performance a critical question.
- The central practical question is whether NOMA can work efficiently in cellular networks.
- A 2 × 2 MIMO-NOMA experiment assessed link-level performance with different receivers indoors and outdoors.
- At a block error rate of 10^-1, experimental and simulation-based results differed by no more than 0.8 dB in SNR.
- Imperfect SIC can significantly degrade performance, but well-designed codes can still let SC provide higher rates than OMA.
- Further studies are needed to understand imperfect CSI effects and NOMA limits in terms of CSI, network size, and clusterable users.
B. Can NOMA benefit from machine learning and deep learning?
Machine learning and deep learning are being applied to NOMA tasks including clustering, power allocation, beamforming, and encoding or decoding. Their effectiveness remains an open question because wireless channels and network topologies change rapidly.
- Deep learning has been applied to beamforming, power allocation, and NOMA encoding or decoding in uplink and downlink.
- ML-based approaches are being considered for user clustering, power allocation, and beamforming in MIMO-NOMA systems.
- The critical question is whether learning-based approaches can work effectively in dynamic networks.
- Wireless channels can change within a few milliseconds, while mobile network topology is naturally dynamic.
- Learning network behavior and resource allocation with deep neural networks is challenging but remains an interesting research field.
- The applications of DL and ML in NOMA-based systems, particularly downlink systems, remain in their infancy.