An Analysis on Rate-Splitting Multiple Access for IRS Aided 6G Communication

I. INTRODUCTION

• IRS and RSMA are presented as technologies attracting interest for next-generation wireless communication, with IRS able to adjust passive-element phases to control propagation.The introduction also surveys prior IRS work on spatial modulation, resource allocation, and beamforming –.Source: §I · p. 1
• To the authors’ knowledge, prior work had not studied applying RS to an IRS-aided network, despite work on IRS-NOMA and RSMA in other settings.The introduction cites IRS-NOMA studies, and RSMA studies –.Source: §I · p. 1
• IRS-RS jointly selects IRS reflection coefficients and designs BS beamformers to maximize weighted sum-rate under BS power, user QoS, and finite-resolution constraints.The comparison with IRS-NOMA considers both perfect and imperfect CSIT.Source: §I · p. 1
• The authors report that IRS-RS achieves better cell-edge-user rates than IRS-NOMA, improving network QoS and user QoE.They describe the proposed design as better than NOMA under both perfect and imperfect CSIT.Source: §I · p. 1

II. NETWORK MODEL

• The model has 2K users—K near and K cell-edge—served by a BS at the center of a hexagonal cell, with K IRSs employing 2-layer HRS.Each cell-edge user is aided by an IRS, and there is no direct BS-to-cell-edge-user link.Source: §II · p. 2, Figure 1 · p. 2
• Each IRS is placed near a cell-edge user, and its diagonal reflection matrix specifies element-wise amplitude coefficients and phase shifts.The BS-to-IRS and IRS-to-user channels are denoted G_k and h_k′, respectively.Source: §II · p. 2
• For tractability, the BS is assumed to know each user's CSI perfectly, while imperfect CSIT is also evaluated in the numerical results.A central control unit assists the BS and IRS integration and estimates the CSIs.Source: §II · p. 2

III. DESIGN OF IRS-RS

• The design applies 2-layer hierarchical RS (HRS) to the IRS-aided network.Source: §III · p. 2

1) Architecture for RSMA:

• The 2K users are organized into K groups, each pairing a near user with a cell-edge user.Source: 1) Architecture for RSMA · p. 2
• Each user's message is split into three parts, with streams including a common stream for all groups, a group-common stream, and user-specific private streams.The private streams are specified separately for near and cell-edge users.Source: 1) Architecture for RSMA · p. 2
• The BS precoder matrix contains precoders for the all-groups common stream, each group's common stream, and each user's private stream.Source: 1) Architecture for RSMA · p. 2
• The design constrains total transmit power, and the authors choose 2-layer HRS as a more general, scalable architecture with 1-layer HRS as a sub-scheme.The total transmit power is constrained by tr(PPH) ≤ Pt.Source: 1) Architecture for RSMA · p. 2

2) Finite resolution beamforming for IRS:

• Each IRS diagonal element is either 1 (on) or 0 (off) under the finite-resolution ON-OFF keying approach.The paper uses this binary design to account for IRS hardware limitations.Source: 2) Finite resolution beamforming for IRS · p. 3
• The IRS design specifies Z = diag{a_p}, where a_p is the pth column of A.Source: 2) Finite resolution beamforming for IRS · p. 3
• P and Q are integers, and the IRS element count is N = PQ.Source: 2) Finite resolution beamforming for IRS · p. 3

IV. PROBLEM FORMULATION

• The WSR objective is subject to base-station power and per-user QoS constraints.Source: §IV · p. 3
• Successive interference cancellation at near and cell-edge users yields SINR expressions for their decoded streams.Source: §IV · p. 3
• The common rate is shared among users as per-user portions that are added to their total rates.Source: §IV · p. 3
• For a given weight vector, the 2-layer HRS WSR problem is difficult in its current form but can be solved using a modified WMMSE approach.Source: §IV · p. 3

V. RSMA BASED BEAMFORMER DESIGN

• At a cell-edge user, SIC decodes the overall common, group-common, and private streams sequentially, removing decoded components before the next stream.Source: §V · p. 3
• With an MMSE receiver, MSE and SINR are related by MSE = 1/(1 + SINR), allowing rate-based design to be expressed using MSE.The paper uses MSE minimization to relate the design to a tight lower bound on rate.Source: §V · p. 3
• The illustrated WMMSE-rate relationships extend across the near and cell-edge user sets, using the corresponding channel matrices.Source: §V · pp. 3, 5
• The alternating algorithm addresses a problem that is not jointly convex by solving convex subproblems while alternating fixed optimization variables.With other variables fixed, the method updates the MMSE equalizer or weight and alternates optimization over P or d until convergence or a fixed iteration limit.Source: §V · p. 5

• Rate-region calculations use a two-user cell, four BS antennas, 20 reflecting elements per IRS, and 20 dB SNR unless otherwise stated.The near-user weight is fixed at 1 while the cell-edge-user weight varies.Source: §VI · p. 5
• The total rate region achieved by IRS-RS is greater than that achieved by IRS-NOMA.Source: §VI · p. 6
• With perfect CSIT, IRS-RS achieves higher sum-rate than IRS-NOMA across the SNR range, and the performance gap grows with SNR.Source: §VI · p. 6, Figure 3 · p. 6
• IRS-RS outperforms IRS-NOMA under both perfect and imperfect CSIT, with significantly better performance reported under imperfect CSIT.Source: §VI · p. 6, Figure 4 · p. 6, Figure 3 · p. 6

• IRS-RS combines BS-side RSMA with IRS assistance for cell-edge users in a multi-user downlink.Source: §VII · p. 6
• The framework jointly selects IRS reflection coefficients and designs BS beamformers to maximize WSR under power, user-QoS, and finite-resolution constraints.Source: §VII · p. 6
• A modified WMMSE-based algorithm tackles the non-convex optimization problem, and numerical results show better performance than the IRS-NOMA baseline.Source: §VII · p. 6