PaperlayerView exact revision on arXiv

Source-linked AI summary

A Primer on Rate-Splitting Multiple Access: Tutorial, Myths, and Frequently Asked Questions

Bruno Clerckx, Yijie Mao, Eduard A. Jorswieck, Jinhong Yuan, David J. Love, Elza Erkip, Dusit Niyato

arXiv:2209.00491v2cs.IT

TL;DR

The paper addresses how multiple-access design should manage interference in next-generation networks, especially under imperfect CSIT and diverse service requirements. It provides a tutorial that builds RSMA from rate-splitting principles across network architectures and applications. The tutorial concludes that RSMA offers broad benefits and a growing role in next-generation communication systems.

  • Problem

    Conventional designs face imperfect CSIT, residual multi-user interference, and practical signaling and latency constraints, while general capacity results remain open in some settings.

  • Method

    The tutorial develops RSMA from rate-splitting principles, covering downlink, uplink, multi-cell, and MIMO frameworks, then examines applications, myths, and implementation issues.

  • Results

    RSMA is presented as partially decoding and partially treating interference as noise, with reported benefits spanning spectral and energy efficiency, flexibility, robustness, reliability, and latency.

  • Takeaways & Limitations

    The tutorial concludes that RSMA can serve as an underpinning multiple-access and interference-management framework for next-generation communication systems.

Abstract

from arXiv · show

Rate-Splitting Multiple Access (RSMA) has emerged as a powerful multiple access, interference management, and multi-user strategy for next generation communication systems. In this tutorial, we depart from the orthogonal multiple access (OMA) versus non-orthogonal multiple access (NOMA) discussion held in 5G, and the conventional multi-user linear precoding approach used in space-division multiple access (SDMA), multi-user and massive MIMO in 4G and 5G, and show how multi-user communications and multiple access design for 6G and beyond should be intimately related to the fundamental problem of interference management. We start from foundational principles of interference management and rate-splitting, and progressively delineate RSMA frameworks for downlink, uplink, and multi-cell networks. We show that, in contrast to past generations of multiple access techniques (OMA, NOMA, SDMA), RSMA offers numerous benefits. We then discuss how those benefits translate into numerous opportunities for RSMA in over forty different applications and scenarios of 6G. We finally address common myths and answer frequently asked questions, opening the discussions to interesting future research avenues. Supported by the numerous benefits and applications, the tutorial concludes on the underpinning role played by RSMA in next generation networks, which should inspire future research, development, and standardization of RSMA-aided communication for 6G.

I. INTRODUCTION

The introduction reframes multiple-access design around interference management rather than a simple orthogonal/non-orthogonal classification. It presents RSMA as a flexible framework extending across architectures, MIMO settings, and many 6G applications.

  • 6G systems must support high throughput, reliability, heterogeneous QoS, and massive connectivity for diverse service requirements.
  • Beyond Orthogonal versus Non-Orthogonal: The orthogonal-versus-non-orthogonal classification is oversimplified because modern 4G and 5G systems combine time-frequency orthogonality with spatial non-orthogonality.
  • Beyond Orthogonal versus Non-Orthogonal: SDMA treats interference as noise, whereas NOMA fully decodes interference, making them fundamentally different interference-management strategies.
  • Toward Rate-Splitting Multiple Access: RS splits messages so interference can be partially decoded and partially treated as noise, enabling flexible inter-user interference management.
  • Toward Rate-Splitting Multiple Access: Imperfect CSIT creates residual multi-user interference for conventional SDMA and MU-MIMO, motivating robust interference-management strategies.
  • Objectives and Contributions: The tutorial develops downlink, uplink, multi-cell, and general MIMO RSMA frameworks, and discusses benefits across over 40 applications and scenarios.

E. Organization and Notations

The paper organizes its foundations around interference management, beginning with a symmetric two-user interference channel and then constructing rate-splitting architectures. It explains how common and private streams support flexible power allocation and decoding.

  • Interference Channel: The fundamentals section uses a symmetric two-user interference channel as the simplest setting that captures interference-management principles.
  • Interference Channel: Orthogonalization avoids interference but can be suboptimal, while treating interference as noise is suited to weak interference and decoding interference to strong interference.
  • Interference Channel: Rate-splitting divides each transmitter message into common and private parts, encoding them into streams decoded by both receivers or by the intended receiver.
  • Interference Channel: In the two-user interference channel, common streams are decoded first while private streams are decoded after common-stream cancellation and treated interference remains as noise.
  • Interference Channel: The achievable symmetric rate is Rsym = Rp + Rc, with parameter t controlling the power allocation and information assigned to common and private streams.

C. Interference Regimes

The symmetric interference channel exhibits four regimes determined by the relative strengths of direct and cross links. Rate-splitting adapts across these regimes by partially decoding and partially treating interference as noise, approaching capacity across channel parameters.

  • Very weak interference: Very weak interference favors treating interference as noise, with all power assigned to private streams and no message splitting.This occurs when |hc| is much smaller than |hd|.
  • Weak interference: Weak interference requires message splitting and nonzero common-stream power, balancing private rates against interference caused to the other receiver.The private-stream interference remains limited while |hc| ≤ |hd|.
  • Strong interference: Strong interference favors decoding interference, with both receivers decoding desired and interfering signals and no private streams.In this regime, messages are entirely encoded in common streams and decoded by both receivers.
  • Rate-splitting: Within one bit/s/Hz of capacity for all channel parameters, RS achieves near-capacity rates and is information-theoretically optimal at asymptotically high SNR.The symmetric-rate example shows RS outperforming other strategies in the weak-interference regime.
  • Scope: For a two-cell interpretation, inter-cell interference is unlikely to lie in the strong regime because users typically associate with their closest base station.The interference-regime classification is more complicated for asymmetric channels.
  • Rate-splitting: RS bridges all four regimes by partially decoding and partially treating interference as noise, while encompassing both extreme strategies.Treating interference as noise becomes inefficient as interference grows, whereas decoding interference becomes inefficient as interference weakens.

III. TWO-USER RATE-SPLITTING MULTIPLE ACCESS

The tutorial builds two-user downlink RSMA from interference-channel principles using common and private streams with linear or nonlinear precoding. Its flexible stream mapping adapts to propagation conditions and contains conventional downlink multiple-access schemes as special cases.

  • Two-user downlink: The two-user interference channel becomes a MISO broadcast channel when transmit antennas cooperate through shared channel state information and messages.The resulting base station serves two users using multiple transmit antennas.
  • MISO RSMA: RSMA splits each user message into common and private parts, encodes them into corresponding streams, and applies precoding across transmit antennas.Common streams are decoded by both users, while private streams serve individual users.
  • MISO RSMA: The four-stream architecture requires two SIC layers per receiver to decode two common streams and one private stream.Each receiver decodes the common streams before retrieving its private message.
  • MISO RSMA: The 1-layer RS architecture combines common parts into one common stream, reducing the design to three streams and one SIC layer per receiver.Both users decode the common stream while treating private-stream interference as noise.
  • Propagation conditions: Precoded channel strengths determine the effective interference regime: orthogonal channels favor private streams, intermediate channels use all streams, and aligned channels favor decoding interference.These cases correspond respectively to very weak, weak, and strong interference.
  • Relationship to existing schemes: RSMA adapts across propagation conditions, whereas SDMA and NOMA are tailored to particular interference strategies, regimes, and channel conditions.In the two-user case, OMA, SDMA, NOMA, and physical-layer multicasting are instances of RSMA.

3) Unifying OMA, SDMA, NOMA, and Multicasting:

RSMA provides a unified framework in which OMA, SDMA, NOMA, and physical-layer multicasting arise as special cases. It manages interference by splitting messages so receivers partially decode interference and partially treat it as noise.

  • Unifying multiple access schemes: OMA, SDMA, NOMA, and physical-layer multicasting are particular instances of RSMA in the two-user case.Each scheme corresponds to a specific message-to-stream mapping or scheduling choice.
  • Unifying multiple access schemes: SDMA forces the common-stream power to zero, so each user decodes its private stream while treating residual interference as noise.This is effective when residual multi-user interference is sufficiently weak.
  • Unifying multiple access schemes: NOMA encodes one user’s message entirely in the common stream, forcing the other user to fully decode that message before decoding its private stream.In the two-user example, W2 is encoded into sc, s2 is turned off, and the resulting model is x = pcsc + p1s1.
  • Interference management: RSMA partially decodes interference and partially treats it as noise, avoiding the requirement that one user fully decode another co-scheduled user’s message.The common-stream contribution is adjusted according to the interference that receivers need to cancel.
  • Rate analysis: In Gaussian SISO downlink with perfect CSIT and CSIR, weighted-sum-rate optimization sets P2 = 0, reducing RSMA to SISO NOMA.With imperfect CSIT, however, the capacity problem is generally open and superposition coding with successive interference cancellation need not achieve capacity.

5) Precoder Design and Power Allocation:

RSMA allocates power and designs precoders according to channel geometry, channel-strength disparity, CSIT quality, and the chosen objective. Its extensions include cooperative relaying and multi-channel transmission for alternating CSIT conditions.

  • Precoder design and power allocation: When private-stream interference saturates private rates, allocating remaining power to the common stream can increase the sum rate beyond SDMA.The private-rate sum remains roughly equivalent to SDMA, while the common rate supplies an additional increase.
  • Precoder design and power allocation: RSMA is information-theoretically optimal in degrees of freedom with imperfect CSIT, whereas SDMA and NOMA are not.Its robustness to imperfect CSIT can also reduce feedback overhead relative to conventional SDMA.
  • Precoder design and power allocation: Common-versus-private power allocation depends on channel-angle separation, channel-strength disparity, the optimization objective, and CSIT quality.These factors motivate either closed-form low-complexity designs or optimization-based precoder and power-allocation methods.
  • Cooperative RSMA: Cooperative RSMA lets a selected user relay the decoded common stream, extending the framework to conventional decode-and-forward and cooperative NOMA instances.Relaying is intended to address propagation conditions involving channel-strength disparity and direction.
  • Space-time / space-frequency RSMA: Space-time or space-frequency RSMA can increase degrees of freedom under alternating user-specific CSIT across time or frequency.It adds a common stream repeated across channel uses; when CSIT is non-alternating, that stream receives zero power and the scheme reduces to separate RSMA.

8) MIMO RSMA:

MIMO RSMA splits each user’s vector messages into common and private parts, transmitting corresponding common and private stream vectors. Both users decode the common vector first, then decode their private vectors while treating the other user’s private vector as noise.

  • MIMO RSMA architecture: In two-user MIMO downlink, each N-dimensional user message is split into common and private parts.The common parts form an N-dimensional common-stream vector, while each user’s private parts form an N-dimensional private-stream vector.
  • MIMO RSMA architecture: Both users decode the common-stream vector first using their receive antennas and successive interference cancellation, then decode their private stream vectors.During private decoding, each user treats the co-scheduled user’s private stream vector as noise.
  • MIMO RSMA architecture: MIMO RSMA extends the message-to-stream mapping to include MU-MIMO, NOMA, OMA, and additional partial-decoding subschemes within one framework.MU-MIMO replaces SDMA by transmitting a vector of private streams to each user.
  • MIMO RSMA performance: RSMA outperforms MU-MIMO and NOMA in MIMO settings and is information-theoretically optimal in degrees of freedom for MIMO broadcast channels with imperfect CSIT.For asymmetric receive-antenna configurations, multi-channel RSMA is needed to achieve optimality.

1) Two-User Architectures:

Two-user RSMA architectures split messages into streams and use decoding choices to manage interference across uplink and broader multiple-access settings. In the uplink, RSMA contains OMA and NOMA as subschemes and extends to MIMO configurations.

  • Uplink: Two-user uplink SIMO RSMA splits one user’s messages into two parts, producing three transmitted streams and requiring two SIC layers for recovery.The receiver can alternatively use joint decoding.
  • Uplink: In the two-user MAC, OMA and NOMA are subschemes of RSMA, with their relationships represented through messages-to-streams mappings.NOMA turns off one split stream, whereas OMA schedules only one user.
  • Uplink: Two-user uplink MIMO RSMA extends the architecture to N transmit antennas per user and M receive antennas, while allowing alternative message-splitting choices.MIMO NOMA and OMA remain subschemes of MIMO RSMA.
  • Lessons Learned: OMA, SDMA, and NOMA target orthogonalization, treating interference as noise, and decoding interference, respectively, whereas RSMA uses rate-splitting across interference regimes.RSMA is presented as a superset that can specialize to these schemes through message-to-stream mappings.
  • Scope: RSMA applies across downlink and uplink as well as SISO, MISO, and MIMO settings, with extensions to cooperative relays and space-time/frequency transmission.The K-user setting extends the two-user schemes and also permits architectures with variable SIC-layer counts.

A. Downlink

Downlink MIMO RSMA splits user messages into common and private components, combines common parts into streams decoded by designated users, and supports layered and generalized architectures. Hierarchical RS and generalized RS expand the decoding structure beyond one-layer RS.

  • System model: A symmetric MIMO broadcast channel serves K users from an M-antenna transmitter, with each user receiving a Qk-dimensional message vector and noise variance set to 1.Messages are split, combined, encoded into streams, and linearly or non-linearly precoded before transmission.
  • 1-layer RS: In 1-layer RS, each user message is split into common and private parts, producing one common stream decoded by all users and private streams decoded individually.Each user uses SIC or joint decoding; SIC requires one layer.
  • 1-layer RS: The common achievable rate is the minimum decoding rate across users, and each user’s total rate combines its allocated common portion with its private rate.This allocation is expressed as sum_k Ck = Rc and Rk,tot = Ck + Rk.
  • Hierarchical RS: Hierarchical RS clusters users by channel-covariance similarity and adds group-common streams alongside an inter-group common stream and private streams.Each user decodes three stream vectors using two SIC layers when SIC is employed.
  • Hierarchical RS: HRS reduces to 1-layer RS when all inner-group common streams are turned off, making HRS a more general framework.The HRS inter-group stream then corresponds to the single common stream of 1-layer RS.
  • Generalized RS: Generalized RS splits messages into multiple common streams intended for different user subsets, while encompassing 1-layer RS, HRS, linearly precoded MU-MIMO, and MIMO NOMA as special cases.Its stream structure includes inter-group, inner-group, and private components.

3) Generalized RS:

Generalized RS organizes common streams by the subsets of users that decode them, enabling a flexible multi-layer downlink structure. Its special cases include one-layer RS and hierarchical RS.

  • Generalized RS: Each user message in GRS is split into 2K−1 sub-message vectors, with subset-specific parts combined into streams decoded by the corresponding user subsets.A stream intended for l users is called an l-order stream vector.
  • Generalized RS: The transmitter groups all l-order streams into sl and linearly precodes them, while each user decodes only the subset-specific streams that includes that user.Other users treat those streams as noise.
  • Generalized RS: Each subset stream’s achievable rate is limited by the minimum decoding rate among its intended users, with portions allocated to their corresponding messages.The overall user rate sums the allocated portions and the private rate.
  • Special cases: GRS reduces to 1-layer RS when only the all-user and private streams remain, and to HRS when inter-group, inner-group, and private streams remain active.These reductions establish GRS as a more general framework than both architectures.
  • Example: In the three-user example, each message is split into four sub-message vectors, seven transmit stream vectors are formed, and three SIC layers recover each user’s four intended streams.User 1 sequentially decodes s123, s12, s13, and s1.

4) Dirty paper coded RS:

Dirty paper coded RS combines rate-splitting with dirty-paper-coded private streams in downlink MIMO, while uplink MIMO RSMA splits user messages and decodes the resulting streams successively. The uplink framework generalizes earlier SISO and SIMO schemes but remains relatively underexplored.

  • Dirty paper coded RS: Dirty paper coded RS is a nonlinearly precoded framework that combines a common stream with dirty-paper-coded private streams under an encoding order.The 1-layer model extends 1-layer RS by applying DPC to private streams.
  • Dirty paper coded RS: Each user decodes the common stream and its own private stream using SIC or joint decoding, with rates defined for both stream types.Overall achievable rates follow the 1-layer RS construction.
  • Dirty paper coded RS: Multi-layer DPCRS combines DPC with GRS, replacing GRS’s linearly precoded private streams with dirty-paper-coded private streams.This provides a further DPCRS architecture for MIMO broadcast channels.
  • Uplink MIMO RSMA: Uplink MIMO RSMA includes SISO and SIMO uplink RSMA as subschemes, while its spectral-efficiency, energy-efficiency, and service applications require further investigation.The stated services include eMBB, URLLC, and mMTC or hybrid services.
  • Uplink MIMO RSMA: Generic uplink MIMO RSMA splits user messages into two sub-message vectors, independently encodes them, precodes them, and superposes them at each transmitter.The receiver detects all streams in a chosen order using SIC.

C. Multi-cell

In multi-cell networks, RSMA is applied to coordinated and cooperative transmission, managing both inter-cell and intra-cell interference. Numerical comparisons report gains in spectral and energy efficiency across multi-user MIMO settings and CSIT conditions.

  • Coordinated transmission: RSMA supports coordinated transmission, where each user’s message is sent from its serving cell while resource allocation and scheduling are coordinated across cells.Each transmitter splits messages into common and private parts, encodes corresponding streams, and applies separate precoders.
  • Transmission and decoding: Each user decodes all common streams and its intended private stream using SIC or joint decoding in coordinated MIMO RSMA.Common and private-stream decoding rates are defined under a selected decoding order.
  • Transmission and decoding: Splitting each user’s message into more than two parts can further boost coordinated MIMO RSMA by assigning parts to user groups with matching CSIT quality.Different common-stream parts can therefore target different groups of users.
  • Cooperative transmission: In cooperative MIMO, all transmitters form a virtual giant transmitter, subject to per-cell power and fronthaul constraints.The downlink RSMA frameworks developed for single-cell MIMO BC can be applied in this setting.
  • Spectral efficiency: RSMA achieves larger ergodic rate regions than MU-MIMO and NOMA, approaching DPC in perfect-CSIT MIMO BC simulations.With equal channel variances, NOMA performs worst because no channel-strength disparity is available for interference management.
  • Spectral efficiency: Under imperfect CSIT, linearly precoded RSMA achieves a larger ergodic rate region than DPC while requiring much lower transceiver complexity.DPC’s rate region drops significantly because it is sensitive to CSIT uncertainty.
  • Energy efficiency and uplink: RSMA achieves higher energy efficiency than NOMA and SDMA, while uplink RSMA reaches capacity-region boundary points without time sharing.The reported downlink comparisons include perfect-CSIT settings and a 1 bit/s/Hz minimum rate threshold per user.
  • Unified framework: RSMA’s operational region and reported SE/EE gains support a unified interference-management framework across multi-cell and other deployments.The tutorial presents RSMA as encompassing OMA, SDMA, NOMA, and physical-layer multicasting as special cases.

3) Flexible:

RSMA is presented as flexible across network loads, channel conditions, CSIT quality, SNR regimes, and cellular settings. Its common stream enables adaptable interference management, while reported results cover robustness, reliability, latency, and broad applicability.

  • 3) Flexible:: With six users and imperfect CSIT, 1-layer RS achieves the best MMF rate in both overloaded and underloaded MISO BC settings.It outperforms SDMA and grouped or ungrouped MISO NOMA across the shown antenna configurations.
  • 3) Flexible:: RSMA adapts to diverse channel directions and strengths, network loads, CSIT conditions, and low, medium, and high SNR regimes.The common stream and power allocation provide the underlying interference-management flexibility.
  • 4) Robust and Resilient:: RSMA is robust to imperfect CSIT and is reported to achieve optimal DoF in MISO BC under imperfect CSIT.Studies address pilot contamination, channel-estimation errors, and user mobility.
  • 6) Reliable and Low Latency:: Under finite blocklength, RSMA achieves a given MMF rate at lower blocklength than SDMA and NOMA, enabling lower latency communications.In one overloaded setting, NOMA requires seven SIC layers while the displayed 1-layer RS scheme requires one.
  • 5) Unifying OMA, SDMA, NOMA, Multicasting:: RSMA is a superset framework that can specialize to OMA, SDMA, and NOMA through message-to-stream mappings across uplink, downlink, and MIMO settings.Not every RSMA instance is a superset of NOMA, but 1-layer RS can outperform NOMA with one SIC layer.
  • 3) Flexible:: RSMA applies to downlink, uplink, and multicell networks with SISO, SIMO, MISO, and MIMO deployments.It can use linear or nonlinear precoding.
  • V. NUMEROUS APPLICATIONS FOR RSMA: The tutorial describes over forty RSMA scenarios and applications spanning modern multi-user systems and 6G services.Reported benefits are supported by stochastic analyses and realistic link-level simulations using 5G-compliant channel models.
  • 4) Robust and Resilient:: RSMA is robust with statistical CSIT and quantized feedback, with fewer feedback bits reported for a given sum-rate performance than conventional SDMA/MU-MIMO.Statistical CSIT can reduce feedback overhead, while quantization-error robustness supports additional overhead reduction.

9) Frequency Division Duplex (FDD) Massive MIMO:

RSMA is presented across massive-MIMO, multi-cell, relaying, security, energy-efficiency, RIS, C-RAN, and resource-management settings as a flexible approach to interference and deployment impairments. The section also identifies cross-layer and long-term resource management as an open direction.

  • FDD Massive MIMO: RSMA mitigates FDD massive-MIMO degradation when covariance spaces overlap or channel statistics and CSIT are imperfect.
  • Massive-MIMO impairments: RSMA is robust to pilot contamination, channel aging, phase noise, and finite-resolution DAC/ADC quantization errors.
  • High-frequency systems: At high frequencies, RSMA combined with hybrid analog-digital precoding can outperform conventional SDMA, while cooperation further combats path loss and blocking.
  • Applications: RSMA has been applied to relaying, full-duplex, physical-layer security, energy efficiency, RIS, C-RAN, caching, and heterogeneous IoT scenarios.
  • Multi-Cell Networks: RSMA supports multi-cell deployments by bridging and outperforming SDMA and NOMA across varying inter-user and inter-cell channel disparities.
  • Open challenges: Current RSMA resource-management work often focuses on short-term optimization, without long-term operational constraints or random traffic arrivals.Cross-layer adaptive source encoding and resource management are recommended for heterogeneous traffic and long-term constraints.

27) Non-Orthogonal Unicast and Multicast (NOUM):

The section describes RSMA for mixed unicast-multicast traffic and diverse 6G scenarios, emphasizing message splitting as a way to manage interference and reuse SIC. It also positions RSMA as a general framework that includes or outperforms conventional schemes in several settings.

  • Non-Orthogonal Unicast and Multicast: In NOUM, RSMA jointly encodes the multicast message and common unicast parts, using SIC to separate multicast from unicast while managing unicast interference.
  • Multigroup multicast: RSMA efficiently handles severe multigroup interference in overloaded multicast networks and significantly outperforms SDMA and NOMA.
  • UAV communications: RSMA can reduce UAV communication energy consumption while allowing an aerial base station to serve multiple ground users simultaneously.
  • Integrated networks: RSMA is used in satellite-terrestrial integrated networks to address severe interference caused by aggressive frequency reuse.
  • Reliability and latency: RSMA’s efficiency, robustness, and flexibility benefits over SDMA and NOMA extend from infinite-blocklength to finite-blocklength communications.
  • Mobile edge computing: RSMA has been reported as more efficient than NOMA for MEC because it can achieve the full MAC rate boundary, whereas NOMA reaches only several separated points.
  • Myths and comparisons: NOMA is described as a particular RSMA instance, while RSMA is a superset of MU-MIMO and can achieve at least MU-MIMO’s performance.

VII. FREQUENTLY ASKED QUESTIONS

The tutorial organizes frequently asked questions about RSMA into principles and benefits, standardization and implementation, and applications and interactions with other technologies.

  • Frequently asked questions are classified into principles and benefits, standardization and implementation, and applications and interplay with other technologies.

A. Principles and Benefits of RSMA

RSMA builds on SDMA to manage interference through flexible rate-splitting, unifying existing multiple-access techniques and adapting across channel, load, and interference conditions. The tutorial reports gains in robustness, efficiency, fairness, complexity, latency, and achievable rates across downlink, uplink, and multi-antenna settings.

  • RSMA unifies and generalizes OMA, SDMA, NOMA, and physical-layer multicasting, with each appearing as a particular RSMA instance.
  • RSMA flexibly adapts to user deployments, network loads, and interference levels, reducing to SDMA or NOMA under corresponding channel conditions.
  • RSMA remains DoF-optimal under imperfect CSIT, whereas OMA, NOMA, and SDMA incur DoF loss.
  • RSMA spectral efficiency and energy efficiency are each always larger than or equal to those of existing multiple-access techniques.
  • RSMA improves QoS and fairness, especially with user rate constraints, weaker-channel weighting, overloaded systems, or degraded CSIT.
  • RSMA can improve performance while reducing receiver and transmitter complexity, latency, and required blocklength compared with multi-antenna NOMA and related schemes.

C. Applications and Interplay between RSMA and other Wireless Technologies

The tutorial positions RSMA as compatible with existing multiple-access, waveform, machine-learning, and integrated-network technologies. It highlights applications spanning OFDMA, delay-Doppler waveforms, federated learning, UAV deployment, and millimeter-wave systems, while noting implementation and signaling constraints.

  • Interplay with multiple access: RSMA can be combined with OFDMA by pairing users and serving each group with RSMA on a resource block or subband.
  • Interplay with multiple access: SDMA is already part of RSMA through private-stream transmission, while power-domain NOMA is part of generalized RSMA; code-domain NOMA remains an open research direction.
  • Interplay with waveforms: RSMA can extend across time or frequency and combine with OTFS or ODDM, supporting flexible allocation and potential spectral-efficiency and sensing benefits.
  • Interplay with machine learning: Machine learning supports RSMA receiver design, resource allocation, power allocation, task offloading, and beamforming, with reported gains over conventional SIC or SDMA.
  • Applications: RSMA enables simultaneous uplink model uploads in federated learning, while ML-assisted UAV deployment has been reported to require less power than other multiple-access schemes.
  • Applications: Millimeter-wave and terahertz systems face blockage, feedback-scaling, beam broadening, and CSI mismatch challenges that motivate robust interference-management approaches.
Loading 2209.00491v2…