Source-linked AI summary
Signal Processing for MIMO-NOMA: Present and Future Challenges
Yongming Huang, Cheng Zhang, Jiaheng Wang, Yindi Jing, Luxi Yang, Xiaohu You
TL;DR
MIMO-NOMA offers a way to combine spatial processing with non-orthogonal access, but system design must account for coupled choices in beamforming, clustering, power allocation, and SIC. This paper reviews those designs and their limitations, highlights SIC stability concerns when achievable rates guide practical systems, and discusses massive/mmWave MIMO challenges and research directions.
Problem
The paper addresses how to design MIMO-NOMA systems across beamforming, clustering, power allocation, and SIC while accounting for practical SIC stability concerns associated with achievable-rate-based designs.
Source: §I · p. 2Method
The paper surveys MIMO-NOMA concepts and design problems, reviews single- and multi-cluster approaches, and discusses massive/mmWave MIMO challenges and future directions.
Source: §I · pp. 2–3, §V · p. 13Results
The review identifies SIC stability as an issue for practical systems using achievable rates and highlights massive/mmWave MIMO challenges tied to antenna configuration and path loss.
Source: §I · pp. 2–3Takeaways & Limitations
The paper concludes that MIMO-NOMA research should address SIC stability and the challenges of combining NOMA with massive and mmWave MIMO.
Source: §I · pp. 2–3, §V · p. 13
Abstract
from arXiv · showhide
Non-orthogonal multiple access (NOMA), as the newest member of the multiple access family, is envisioned to be an essential component of 5G mobile networks. The combination of NOMA and multi-antenna multi-input multi-output (MIMO) technologies exhibits a significant potential in improving spectral efficiency and providing better wireless services to more users. In this article, we introduce the basic concepts of MIMO-NOMA and summarize the key technical problems in MIMO-NOMA systems. Then, we explore the problem formulation, beamforming, user clustering, and power allocation of single/multi-cluster MIMO-NOMA in the literature along with their limitations. Furthermore, we point out an important issue of the stability of successive interference cancellation (SIC) that arises using achievable rates as performance metrics in practical NOMA/MIMO-NOMA systems. Finally, we discuss incorporating NOMA with massive/millimeter wave MIMO, and identify the main challenges and possible future research directions in this area.
I. INTRODUCTION
NOMA improves spectral efficiency and fairness by allowing users with different channel conditions to share resources through superposition coding and successive interference cancellation. The paper reviews MIMO-NOMA designs, highlights SIC stability concerns, and discusses challenges and research directions for massive and mmWave MIMO.
I. INTRODUCTION
- NOMA lets users with significantly different channel conditions share a resource block using superposition coding and successive interference cancellation, balancing spectral efficiency and fairness through pairing and power allocation.The introduction contrasts this with OMA’s low spectral efficiency when resources are assigned to users with poor channel conditions.Source: §I · p. 2
- MIMO-NOMA combines spatial-domain beamforming with intra-cluster SIC, while its full benefit calls for jointly designing user clustering, beamforming, power allocation, and SIC.The review covers single-cluster systems, where all users except the weakest perform SIC, and multi-cluster systems, where users are partitioned into clusters.Source: §I · p. 2
- The paper flags SIC stability as a practical concern because achievable-rate-based strategies can harm SIC; simulations show a tradeoff between decoding accuracy for weak and strong users.The introduction notes that equal power allocation can lead to the largest error propagation.Source: §I · pp. 2–3
- For massive and mmWave MIMO-NOMA, large antenna arrays and path loss challenge traditional linear channel estimation and make perfect instantaneous CSI at the base station impractical.The paper identifies channel estimation and limited feedback, hybrid massive MIMO-NOMA, and statistical massive MIMO-NOMA as future research directions.Source: §I · p. 3
II. BASIC CONCEPTS AND PROBLEMS IN MIMO-NOMA
The section introduces NOMA through a two-user SISO example and describes how SIC extends to an ordered set of users. Users share one resource block, with decoding and interference treatment determined by their order.
II. BASIC CONCEPTS AND PROBLEMS IN MIMO-NOMA
- The section uses the two-user SISO scenario to introduce NOMA’s essential principle before presenting the concepts and design problems of MIMO-NOMA.The broader section also introduces performance criteria and design problems.Source: §II · p. 3
A. Principle of NOMA
- In the two-user SISO example, a far user decodes while treating the near user’s message as noise, whereas the near user decodes the far user first and then removes its message before decoding its own.The users share a resource block with transmit powers P1 > P2, and both have access to the whole block.Source: A. Principle of NOMA · p. 3
- For K users in a given order, SIC has each user decode messages sequentially, removing interference from previously decoded messages and treating later users’ messages as noise.The channel coefficient of user Ui is denoted hi.Source: A. Principle of NOMA · p. 3
B. From SISO-NOMA to MIMO-NOMA
MIMO-NOMA adds spatial degrees of freedom to NOMA, making interference management and design more complex. The section outlines a clustered downlink model and the rate and design criteria used for these systems.
B. From SISO-NOMA to MIMO-NOMA
- MIMO-NOMA can use beamforming in both power and spatial domains, but its effective channel gains depend on beamforming, coupling beamforming design with SIC ordering.The system may have multi-antenna base stations and users with one or multiple antennas.Source: B. From SISO-NOMA to MIMO-NOMA · p. 4
- In the general multi-cluster downlink model, a base station with M antennas serves K users with N antennas each, partitioned into C clusters with K̄ users per cluster.Within a cluster, users are ordered from weakest to strongest for SIC.Source: B. From SISO-NOMA to MIMO-NOMA · p. 4
- The model uses SIC within clusters and beamforming across clusters to suppress interference; with sufficient base-station antennas, beamforming can also suppress or eliminate intra-cluster interference.Source: B. From SISO-NOMA to MIMO-NOMA · p. 4
- Achievable rate is the common performance metric, with a user’s rate determined by the minimum SINR for its message across the users required to decode it.Outage probability and symbol error rate are also identified as performance metrics; design problems include beamforming, power allocation, user clustering, and SIC ordering.Source: B. From SISO-NOMA to MIMO-NOMA · pp. 4–5
III. OVERVIEW OF EXISTING WORKS
The review groups existing MIMO-NOMA works into single-cluster and multi-cluster approaches, and notes that they generally assume perfect instantaneous CSI at the base station.
III. OVERVIEW OF EXISTING WORKS
- The review classifies representative MIMO-NOMA works into single-cluster and multi-cluster systems and discusses their limitations.Source: §III · p. 5
- Unless otherwise stated, the reviewed works assume perfect instantaneous CSI at the base station.Source: §III · p. 5
A. Single-Cluster MIMO-NOMA
Single-cluster MIMO-NOMA designs optimize beamforming and power allocation under a given SIC order, but their optimality and sensitivity to channel conditions remain open.
A. Single-Cluster MIMO-NOMA
- The sum-rate design is nonconvex and uses successive convex approximation, with constraints on total power and users’ effective message powers.The effective-power constraint protects weak-user performance.Source: A. Single-Cluster MIMO-NOMA · p. 5
- The gap between the suboptimal design and optimal DPC performance is unknown, and channel norm and correlation effects need study, especially for users at similar BS distances.The stated results concern single-cluster MIMO-NOMA, where all K users share one resource block.Source: A. Single-Cluster MIMO-NOMA · pp. 5–6
- For quasi-degradation channels, gives a closed-form power-minimizing beamformer under SINR constraints; ZFBF is preferable for non-quasi-degradation channels, especially at large channel angles.Hybrid NOMA switches dynamically between these designs, but its optimal beamformer is known only for two users.Source: A. Single-Cluster MIMO-NOMA · p. 6
B. Multi-Cluster MIMO-NOMA
Multi-cluster MIMO-NOMA combines SDMA and NOMA to address limits of large single clusters, but inter-cluster interference complicates beamforming. With insufficient BS antennas, NOMA's power dimension complements limited spatial degrees of freedom.
B. Multi-Cluster MIMO-NOMA
- Multi-cluster MIMO-NOMA combines SDMA with NOMA because large single clusters limit spatial-degree-of-freedom gains and increase beamforming and SIC complexity.Inter-cluster interference further complicates beamforming.Source: B. Multi-Cluster MIMO-NOMA · p. 6
1) Insufficient BS Antennas:
- When users’ total receiver-antenna count exceeds the BS antenna count, spatial degrees of freedom cannot support complete SDMA, motivating NOMA’s additional power dimension.In and, two single-antenna users share each cluster’s beamformer; ZFBF and user clustering are used to reduce inter-cluster interference.Source: 1) Insufficient BS Antennas · p. 6
- The and designs determine beamforming, clustering, and power allocation separately, whereas jointly optimizes intra-cluster power under rate and SIC-reliability constraints.The approach assumes distinguishable near-far effects, which may not hold for randomly deployed users.Source: 1) Insufficient BS Antennas · p. 7
2) Sufficient BS Antennas:
Multi-cluster NOMA is also considered when BS antennas exceed users’ total antenna count; reviewed approaches use staged beamforming and clustering, with limits on optimality and cluster size.
2) Sufficient BS Antennas:
- [10] uses two-stage SVD-based and intra-cluster beamforming to minimize transmit power under SINR constraints, then clusters users using channel correlation or angle.With sufficient BS antennas, multiple beams can serve each cluster, unlike the single-beam ZFBF-like designs in –; the second-stage result is not guaranteed optimal.Source: 2) Sufficient BS Antennas · p. 7
- [11] proposes two-stage PH-NOMA beamforming without a near-far distribution condition, but it applies only to two-user clusters and lacks a generalization to larger clusters.Its clustering aims to form more pairs satisfying the quasi-degradation condition.Source: 2) Sufficient BS Antennas · p. 8
- The discussion focuses on narrow-band systems and spatial- and power-domain joint design; extending wide-band NOMA to MIMO remains worth studying for practical wide-band 5G systems.Source: 2) Sufficient BS Antennas · p. 8
IV. FUTURE DIRECTIONS
The paper flags SIC stability as a concern when achievable rates guide practical NOMA designs. It reports a power-allocation tradeoff in decoding accuracy and identifies related quantities for further study.
IV. FUTURE DIRECTIONS
- The section also discusses incorporating NOMA into massive/millimeter-wave MIMO and proposing future research directions.Source: §IV · p. 8
A. SIC-Stability
- Achievable-rate optimization can produce equal power allocation, which is worst for SIC and causes the largest error propagation in the two-user SISO-NOMA example.Finite coding length and practical modulation make decoding errors inevitable, while the ideal achievable-rate model assumes successful decoding below the rate bound.Source: A. SIC-Stability · p. 8
- At high SNR, increasing weak-user power first lowers and then raises strong-user SER22; at low SNR, it has no positive effect on SER22.The example uses 4-QAM for strong user U2 and BPSK for weak user U1.Source: A. SIC-Stability · p. 9
- The authors identify the SER22–SER21 relationship under practical modulation and finite coding length, and quantitative definitions of SIC residue or SIC-based achievable rate, as open design questions.Source: A. SIC-Stability · p. 10
B. Massive/Millimeter-Wave MIMO-NOMA
Massive and mmWave MIMO-NOMA raise channel-acquisition challenges that can make traditional linear estimation and perfect instantaneous CSI assumptions impractical. The section discusses channel estimation, limited feedback, and statistical or reduced-dimensional CSI as research directions.
B. Massive/Millimeter-Wave MIMO-NOMA
- Massive antenna arrays and large path loss challenge traditional linear channel estimation and make perfect instantaneous CSI at the BS impractical.The section identifies channel estimation and limited feedback, hybrid massive MIMO-NOMA, and statistical massive MIMO-NOMA as directions for addressing these challenges.Source: B. Massive/Millimeter-Wave MIMO-NOMA · p. 10
1) Channel Estimation and Limited Feedback for Massive MIMO-NOMA:
- For massive/mmWave MIMO-NOMA, channel training and feedback overhead grows with antenna and user counts, motivating redesigned compressive-sensing and beam-alignment methods that account for NOMA’s near-far effect.Existing mmWave beam-alignment algorithms often serve one user at a time, a constraint the passage says must be improved for large user populations.Source: 1) Channel Estimation and Limited… · p. 10
- Limited-feedback studies cover user ordering, but feedback of large-dimensional CSI could offer better performance and warrants further study.The cited studies use no instantaneous CSI for beamforming or power allocation; one-bit feedback matches perfect-ordering diversity in two-user clusters, while carefully quantized channel-quality feedback outperforms distance feedback.Source: 1) Channel Estimation and Limited… · p. 11
2) Hybrid Massive MIMO-NOMA:
The hybrid scheme combines statistical channel information with reduced-dimensional instantaneous CSI for single-antenna users. In the described 32-antenna, 64-user simulation, it matches MIMO-NOMA performance with lower training overhead, while practical grouping and clustering methods remain a need.
2) Hybrid Massive MIMO-NOMA:
- Because user-side ZFBF is infeasible for single-antenna users, reduced-dimensional instantaneous-CSI beamforming can suppress inter-cluster interference.The scheme combines statistical and reduced instantaneous CSI; its statistical-CSI beamforming can be implemented in the analog domain and reduced-dimensional beamforming in the baseband domain.Source: 2) Hybrid Massive MIMO-NOMA · pp. 11–12
- The simulation assumes perfect statistical-CSI grouping and instantaneous-CSI clustering, while practical low-complexity dynamic grouping and clustering approaches remain important.Source: 2) Hybrid Massive MIMO-NOMA · p. 11
- For 32 BS antennas and 64 single-antenna users, the hybrid scheme using statistical and reduced instantaneous CSI achieves the same performance as MIMO-NOMA with much lower training overhead.Source: 2) Hybrid Massive MIMO-NOMA · p. 11
3) Statistical Massive MIMO-NOMA:
For full instantaneous CSI, PH-NOMA needs less transmit power than linear ZFBF to meet users’ rate/SINR constraints, with a larger advantage as antenna count increases or M−K decreases. The paper concludes by reviewing MIMO-NOMA designs and identifying challenges and future directions, including statistical-CSI design.
3) Statistical Massive MIMO-NOMA:
- Statistical-CSI-only beamforming is desirable when short channel coherence makes reduced instantaneous-CSI estimation unaffordable, but its optimization is more challenging.The passage notes a complicated relationship between metrics such as sum-rate and beamforming vectors.Source: 3) Statistical Massive MIMO-NOMA · p. 12
- PH-NOMA needs less transmit power than linear ZFBF to guarantee users’ rate/SINR constraints, with a larger advantage as M increases or M−K decreases.This comparison evaluates the full instantaneous-CSI case; dense or high-mobility deployments can make reduced instantaneous-CSI estimation overhead unaffordable.Source: 3) Statistical Massive MIMO-NOMA · p. 12, Figure 4 · p. 13
V. CONCLUSION
- The conclusion reviews single- and multi-cluster MIMO-NOMA, existing design contributions and limitations, SIC stability, and challenges and future directions for massive and mmWave MIMO.Source: §V · p. 13