PaperlayerView exact revision on arXiv

Source-linked AI summary

Non-Orthogonal Multiple Access in Multi-Cell Networks: Theory, Performance, and Practical Challenges

Wonjae Shin, Mojtaba Vaezi, Byungju Lee, David J. Love, Jungwoo Lee, H. Vincent Poor

arXiv:1611.01607v2cs.IT

TL;DR

Multi-cell NOMA remains less studied than single-cell NOMA, while increasing network density makes inter-cell interference a major obstacle. The paper develops the theory, reviews interference-management literature, and discusses implementation and research challenges. Its numerical results show that interference cancellation is significant in multi-cell NOMA, alongside major practical challenges.

  • Problem

    Multi-cell NOMA is less studied than single-cell NOMA, and increasing network density makes inter-cell interference a major obstacle.

  • Method

    The paper discusses NOMA theory, reviews interference-management techniques for multi-cell networks, and examines practical implementation and research challenges.

  • Results

    Numerical results show the significance of interference cancellation in NOMA.

  • Takeaways & Limitations

    Understanding and addressing interference is central to realizing multi-cell NOMA solutions.

  • Takeaways & Limitations

    The paper highlights major practical issues and challenges in implementing multi-cell NOMA.

Abstract

from arXiv · show

Non-orthogonal multiple access (NOMA) is a potential enabler for the development of 5G and beyond wireless networks. By allowing multiple users to share the same time and frequency, NOMA can scale up the number of served users, increase the spectral efficiency, and improve user-fairness compared to existing orthogonal multiple access (OMA) techniques. While single-cell NOMA has drawn significant attention recently, much less attention has been given to multi-cell NOMA. This article discusses the opportunities and challenges of NOMA in a multi-cell environment. As the density of base stations and devices increases, inter-cell interference becomes a major obstacle in multi-cell networks. As such, identifying techniques that combine interference management approaches with NOMA is of great significance. After discussing the theory behind NOMA, this paper provides an overview of the current literature and discusses key implementation and research challenges, with an emphasis on multi-cell NOMA.

I. WHAT DRIVES NOMA?

NOMA is motivated by massive connectivity, diverse latency requirements, and higher spectral efficiency, but multi-cell deployment introduces inter-cell interference as a central challenge.

  • NOMA Operation: Power-domain NOMA exploits channel-gain differences to multiplex users through power allocation.NOMA may also be realized in code or other domains; code-domain NOMA uses user-specific spreading sequences.
  • Massive Connectivity: NOMA can theoretically serve an arbitrary number of users in each resource block by superimposing their signals.This supports IoT applications involving many devices that sporadically transmit small packets.
  • Low Latency: NOMA supports flexible scheduling for devices with diverse latency and quality-of-service requirements.OMA may require devices to wait for an unoccupied resource block, whereas NOMA accommodates a variable number of devices.
  • High Spectral Efficiency: NOMA gives each user access to the whole bandwidth, while OMA limits users to fractions of spectrum inversely proportional to their number.The paper identifies this as the basis for NOMA’s spectral-efficiency and user-fairness advantages.
  • Multi-Cell Motivation: Much prior NOMA work focuses on single-cell settings, motivating analysis of more realistic multi-cell networks.The article therefore considers multi-cell NOMA and reviews its implementation issues and research challenges.
  • Multi-Cell Motivation: As network density increases, inter-cell interference becomes a major obstacle to realizing NOMA’s benefits.The paper emphasizes combining interference management with NOMA and evaluating multi-cell solutions at the system level.

II. THEORY BEHIND NOMA

The paper frames cellular downlink and uplink as broadcast and multiple access channels, respectively, and examines the rates promised by multi-user information theory and how NOMA achieves them.

  • Channel Models: Cellular downlink is modeled as a broadcast channel, whereas uplink is modeled as a multiple access channel.In downlink the base station transmits to multiple users; in uplink multiple users transmit to the same base station.
  • Information-Theoretic Perspective: The paper asks what the highest achievable throughputs are for multi-user channels and how systems can achieve those rates.This motivates reviewing information-theoretic results in both single-cell and multi-cell settings.
  • Single-Cell Capacity Regions: For the two-user MAC and BC, NOMA achieves the capacity regions by transmitting both users’ signals simultaneously in the same time-frequency band.The figure compares the achievable regions of OMA and NOMA for MAC, BC, and interference channels.
  • Single-Cell Capacity Regions: Except for a few points, OMA is strictly suboptimal in the illustrated MAC and BC comparisons.The paper explains the corresponding regions using time allocation for OMA and simultaneous transmission with decoding strategies for NOMA.

1) Uplink (MAC):

In the MAC, NOMA uses simultaneous transmission and successive interference cancellation to attain the capacity region; in the BC, superposition coding and power allocation achieve its boundary and support fairness.

  • Uplink (MAC): OMA assigns users separate time fractions, producing rates R1 = αC(γ1) and R2 = ¯αC(γ2).Power control can boost these achievable rates.
  • Uplink (MAC): In uplink NOMA, users transmit concurrently and the base station uses successive interference cancellation to achieve any point in the capacity region.For one decoding order, the base station first decodes one signal while treating the other as noise, then removes it before decoding the remaining signal.
  • Uplink (MAC): The gap between NOMA and OMA regions becomes larger when OMA does not use power control.This comparison is shown for the multiple access channel.
  • Downlink (BC): Downlink NOMA strictly increases the OMA rate region through superposition coding and successive interference cancellation at the stronger-channel user.Varying the power split allows rate pairs on the boundary of the broadcast-channel capacity region.
  • Downlink (BC): Flexible power allocation enables smooth, optimal user-fairness improvement by assigning more power or weighted-sum-rate priority to a disadvantaged user.The allocation parameter β determines the boundary point associated with the selected weighting.

3) K-User Uplink/Downlink:

For K-user uplink and downlink channels, NOMA-based coding schemes achieve the capacity regions in the single-cell MAC and BC models. In multi-cell interference channels, capacity is generally unknown, but achievable regions show advantages for NOMA and combined NOMA-OMA schemes.

  • K-user downlink: Superposition coding with SIC achieves the largest rate region for the K-user broadcast channel.
  • K-user uplink/downlink: OMA is strictly suboptimal for the single-cell MAC and BC capacity regions.
  • Multi-cell interference channels: Capacity-achieving schemes are unknown for general multi-cell interference channels, although achievable regions indicate NOMA superiority over OMA.
  • Interference channel: Combining NOMA and OMA through the Han–Kobayashi scheme with time-sharing gives the largest known rate region for the interference channel.

2) Interfering MAC and BC:

Interfering MAC and BC models provide information-theoretic evidence that NOMA can outperform OMA in multi-cell settings. However, optimal transmission and reception strategies remain elusive, motivating practical interference-management approaches.

  • Interfering MAC: Interfering multiple-access models use Han–Kobayashi coding for interfering transmitters and single-user coding for non-interfering transmitters.
  • Interfering MAC: NOMA-based transmission yields an inner bound within one bit of the capacity region for a mutually interfering two-cell uplink model.
  • Open challenge: Despite intensive research, optimal uplink and downlink transmit/receive strategies for multi-cell networks remain elusive.
  • Interfering MAC and BC: Achievable rate regions for mutually interfering two-cell networks suggest that NOMA-based techniques outperform OMA.
  • Practical approaches: The lack of optimal multi-cell strategies has motivated pragmatic approaches that treat interference as noise.
  • NOMA background: NOMA has been reported to improve system throughput and user fairness over OFDMA, while practical studies address pairing, power allocation, and SIC implementation.

IV. MULTI-CELL NOMA SOLUTIONS

Multi-cell NOMA combines NOMA with interference-management methods because inter-cell interference is the main obstacle, especially for cell-edge users. Coordination can improve cell-edge service, but joint transmission may require substantial CSI-sharing overhead.

  • Interference management: Multi-cell NOMA combines NOMA with coordinated scheduling/beamforming or joint processing to manage inter-cell interference.
  • Motivation: Inter-cell interference is the main issue in multi-cell NOMA because it reduces cell-edge user performance.
  • Solution classes: NOMA-CS/CB keeps each user’s data at and transmitted from one base station, whereas NOMA-JP shares data among multiple base stations.
  • NOMA-JT: NOMA-joint transmission uses multiple base stations to simultaneously serve a user over a shared wireless resource.
  • NOMA-JT trade-offs: Network MIMO can cancel cell-edge inter-cell interference completely, but typically requires global CSI and causes excessive backhaul overhead.
  • Coordinated superposition coding: Coordinated superposition coding serves cell-center users from their corresponding base stations and the cell-edge user from both base stations.
  • NOMA-JT: Coordinating two cells can provide a common cell-edge user a reasonable transmission rate without sacrificing cell-center users’ rates.

2) NOMA-DCS:

NOMA-DCS shares a cell-edge user’s data across multiple base stations but selects one serving BS, while coordinating transmission and beamforming to manage inter-cell interference. Its variants trade data sharing and CSI requirements against coordination overhead and antenna needs.

  • NOMA-DCS: NOMA-DCS dynamically selects one serving BS for the cell-edge user using order statistics.The selected BS can change when the channel ordering changes.
  • NOMA-DCS: Only the selected BS applies NOMA to the cell-edge and cell-center users, while the other BS serves its cell-center user with OMA.This changes the NOMA-JT rate expressions because the nonselected user does not use SIC for NOMA transmission.
  • NOMA-CS/CB: NOMA-CS and NOMA-CB do not share user data among BSs, but cooperating BSs exchange global CSI and scheduling information through X2.The resulting overhead can be non-negligible for high-mobility cell-edge users.
  • NOMA-CB: Interference-alignment-based CB jointly optimizes beamforming vectors to remove ICI and inter-cluster interference.The ICA-based method requires global CSI, whereas the IA-based method uses serving-channel knowledge but needs more antennas.
  • NOMA-CB: When the number of users is sufficiently large, the extra antennas required by the IA-based CB method become negligible.

2) NOMA-CS:

NOMA-CS coordinates scheduling across geographically separated BSs so cell-edge users receive less interference while preserving single-BS data availability. Its scheduling design improves practical interference management but involves an NP-hard combinatorial optimization.

  • NOMA-CS: NOMA-CS coordinates geographically separated BS scheduling to serve NOMA users with less inter-cell interference.An adjacent BS may transmit only to a cell-center user instead of superimposing a message for additional NOMA users.
  • NOMA-CS: NOMA-CS scheduling is formulated as a combinatorial optimization problem that is NP-hard.A simple scheduling algorithm is therefore needed to select NOMA users at each BS within a scheduling interval.
  • Comparison: With K two-user clusters per cell, single-cell NOMA can support 2K users, whereas single-cell OMA can serve only K users.The OMA count is limited by the number of BS antennas.
  • Practical issues: SIC is central to NOMA, but decoding complexity scales with the number of users because each user decodes signals intended for users earlier in the SIC order.Clustering users and applying encoding and decoding within clusters can reduce this complexity.
  • Practical issues: Network-assisted interference cancellation and suppression, or NAICS, is a relatively complex user-terminal category included in 3GPP LTE-A.

2) Error propagation:

Multi-user NOMA performance depends on SIC reliability, clustering, power allocation, and interference management. The paper highlights error propagation, imperfect CSI, unknown optimal multi-cell designs, and trade-offs introduced by FFR-based coordination.

  • Error propagation: An SIC decoding error affects all later users in the decoding order, making their signals likely to be decoded incorrectly.Stronger codes can compensate for this effect when the number of users is not large.
  • Imperfect CSI: Imperfect CSI prevents complete removal of other users’ signals and leaves joint precoders without a known guarantee of zero ICI.The paper calls for beamforming designs robust to CSI errors in multi-cell NOMA.
  • Multi-User Power Allocation and Clustering: Power allocation controls system throughput and user-fairness, while one cluster maximizes system throughput in theory for more than two users.In practice, one cluster can cause serious degradation from SIC errors, motivating multiple clusters.
  • Multi-User Power Allocation and Clustering: Optimal clustering and power allocation remain unknown in multi-cell networks even when SIC errors are ignored.Practical systems therefore require algorithms with reasonable complexity and good performance.
  • FFR and NOMA: FFR offers simple ICI management without CSI but can reduce cell-edge rates and deteriorate user-fairness.Combining FFR and NOMA may also yield limited NOMA gains when paired users have similar channel conditions.

E. Security

NOMA’s stronger-channel user can decode another user’s signal, creating security concerns, while multi-cell performance depends heavily on interference management. Simulations compare schemes using spectral efficiency and throughput distributions under realistic two-cell conditions.

  • Security: Because a better-channel user can decode another user’s signal, NOMA introduces additional security concerns.Upper-layer cryptographic security remains relevant because only the legitimate user has the key, while physical-layer security is harder to apply.
  • Performance evaluation: The numerical analysis evaluates multi-cell NOMA in a realistic setting by treating inter-cell interference as noise.The comparison includes OMA, OMA-FFR, NOMA, NOMA-TDM, NOMA-JT, and NOMA-CB in a two-cell downlink network.
  • Performance evaluation: Fig. 4(a) plots spectral efficiency against cell-edge-user location, while Fig. 4(b) plots the individual-throughput CDF for random deployments.
  • Performance comparison: OMA and NOMA decline as cell-edge users move toward unfavorable locations because they do not include ICI mitigation.NOMA-TDM and OMA-FFR improve cell-edge rates by dividing resources for multi-cell operation.
  • Performance comparison: NOMA-CB fully exploits available resources and achieves twice the performance of NOMA-TDM, while NOMA-JT performs almost similarly to NOMA-CB.NOMA-JT’s gain increases near the cell border because data sharing lets the edge user benefit from the neighboring-BS link.

SIMULATION PARAMETERS

The article reviews multi-cell NOMA and shows that interference cancellation and coordinated transmission can improve throughput, especially under inter-cell interference. It also identifies practical implementation issues and challenges.

  • NOMA-CB and NOMA-JT achieve the best user-throughput performance across the examined throughput distribution.The comparison includes the 5%-tile CDF point and average user throughput.
  • Their advantage is attributed to effective inter-cell-interference control through multi-cell NOMA transmissions.
  • OMA and NOMA benefit cell-center throughput through full resource-block use but are limited by severe interference for cell-edge users.
  • NOMA-TDM and OMA-FFR mitigate inter-cell interference by splitting each resource block between two cells, at the expense of cell-center throughput.
  • The article combines single-cell and multi-cell NOMA theory with a literature review of interference-management techniques and discussion of implementation challenges.
  • Numerical results demonstrate the significance of interference cancellation in NOMA.
Loading 1611.01607v2…