REVIEW 5 major objections 5 minor 14 references
Joint Phase Shift Optimization and Precoder Selection for RIS-Assisted 5G NR MIMO Systems
T0 review · 5 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash
Pith's one-line read This paper claims that a singular-value-based RIS phase optimization coupled with inner-product codebook matching for precoder selection consistently outperforms conventional mutual-information-based PMI selection in RIS-assisted 5G NR…
desk verdict A practical low-complexity RIS/PMI design, but the central 'consistently outperforms' claim is contradicted by the paper's own figures; the underlying idea is worth a revision, not acceptance as-is. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The key machinery is the singular value decomposition of the cascaded channel matrix $F = G\Phi H$, where $G$ is the RIS-to-UE channel, $H$ is the gNodeB-to-RIS channel, and $\Phi$ is a diagonal matrix of discrete RIS phases. The optimization objective is the normalized sum of the largest $\nu$ singular values $\lambda_1 + \lambda_2$ (for $\nu > 1$) or $\lambda_1$ (for $\nu = 1$), and the maximum cross-swapping algorithm (MCA) maximizes this objective by generating $T$ random phase configurations, identifying the two best, and crossing them to produce offspring configurations. Once the RIS phases are fixed, the dominant and second-dominant right singular vectors are mapped onto the 3GPP Type-I codebook through inner-product maximization in equations (15)–(17), producing the wideband beam pair and the per-subband co-phasing values.
What would settle it
Run the same $\lambda$-based RIS optimization and the two precoder-selection methods with the channel model replaced by $F = H_{\mathrm{direct}} + G\Phi H$, where $H_{\mathrm{direct}}$ follows a Rician distribution with a strong line-of-sight component, and compare achievable rates; if the proposed method does not consistently outperform the conventional mutual-information-based selection, the central claim is falsified.
Extended reading notes
Core claim
The paper's central claim is that for an RIS-assisted 5G NR MIMO link with no direct path, the SVD of the cascaded channel $F = G\Phi H$ should drive both the RIS configuration and the codebook precoder: maximize the top one or two singular values (depending on the number of layers) through the RIS phase pattern, then choose the Type-I precoder by maximizing the inner product between the first half of the corresponding right singular vectors and the IDFT grid-of-beams. The authors assert that this proposed precoder selection consistently outperforms the conventional approach under $\lambda$-based RIS optimization, with the largest gains at higher SNR and for low-to-moderate antenna and RIS counts, and that it reduces the computational complexity of PMI selection by removing the layer-count dependence in the wideband part and eliminating subband mutual information calculations.
Load-bearing premise
The system model assumes there is no direct link between the gNodeB and the UE, so the whole channel is the cascaded RIS path; if a strong line-of-sight path existed, the singular vectors would change and the claimed performance ordering could break.
Editorial extensions
If this is right
- The proposed precoder selection consistently outperforms or matches the conventional PMI selection under $\lambda$-based RIS optimization, with the largest rate gains at higher SNR and with small to moderate antenna and RIS element counts.
- $\lambda$-based RIS optimization delivers higher achievable rate than effective rank-based optimization for both precoder selection methods.
- The selection complexity after SVD becomes $O(N_T(N_B/2 + N_3))$ and is independent of the number of layers in the wideband part, so the savings grow with antenna count, subbands, and layers.
- The method requires only a standard Type-I codebook plus feedback of the optimization parameter, preserving backward compatibility with 5G NR CSI reporting.
Reading between the lines
- The no-direct-link assumption is load-bearing: with a strong line-of-sight path, the effective channel becomes $F = H_{\mathrm{direct}} + G\Phi H$ and the optimal singular vectors change, so the reported performance ordering may not transfer to such deployments.
- The inner-product matching recipe is general and could be tested against other standardized codebooks, such as Type-II, or against non-codebook precoding.
- Because MCA is a heuristic over a discrete phase set with fixed search budget $T$, the reported gap versus the conventional method may shift with different phase quantization bits or search parameters.
- The paper does not quantify the feedback overhead for sending the optimization parameter alongside PMI; a standards-compliant design would need to add that field to the CSI report.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper proposes a joint RIS phase-shift optimization and precoder-selection scheme for RIS-assisted 5G NR MIMO downlink transmission. The RIS phases are optimized by a maximum cross-swapping algorithm (MCA) to maximize the dominant singular values of the cascaded channel F = GΦH, and the right singular vectors are then mapped onto 3GPP Type-I codebook precoders, avoiding per-subband mutual-information computations and layer-dependent wideband search. Simulation results over CDL-C channels without a direct link are reported for 4x4 and 12x12 MIMO with 16 and 128 RIS elements. The paper claims that the proposed precoder selection consistently outperforms a conventional approach under the singular-value-based RIS optimization, while reducing time complexity.
Significance. If the performance and complexity claims were fully supported, the paper would offer a practically relevant contribution: a low-complexity PMI selection rule that is compatible with 3GPP Type-I codebooks and with RIS phase optimization, which is a natural step toward standard-compatible RIS-assisted MIMO. The complexity argument is plausible, and the use of a standardized codebook together with a realistic path-loss model is a strength. However, the significance is reduced by the fact that the headline claim of consistent outperformance is contradicted by the paper's own simulation text and figures, and by the absence of any statistical characterization of the reported rate gaps. The contribution is therefore better characterized as a heuristic precoder-mapping method with comparable performance in several regimes, not a demonstrated consistently superior scheme.
major comments (5)
- [Abstract, Section III, Section IV] The central claim that the proposed precoder selection method "consistently outperforms" the conventional approach under λ-based RIS optimization is not supported by the paper's own results. In Section III, the text states that for NRIS = 16 the methods achieve the same performance, and for 12x12 MIMO with NRIS = 128 the distributions of singular values become more uniform, resulting in similar performance. The conclusion itself concedes that higher gains are observed when the number of antennas and RIS elements is small. The abstract and conclusion should be revised to state the actual finding, namely that gains appear in low-to-moderate antenna/RIS configurations, and that the method has lower complexity even when performance is similar.
- [Section III] No confidence intervals, error bars, or number of Monte Carlo runs/channel realizations are reported for any of the achievable-rate curves. Since the claimed advantage is visible only as small separations in some SNR ranges, it is impossible to determine whether the differences are statistically meaningful or just noise. The authors should specify the number of independent channel realizations and provide confidence intervals or a statistical test for the rate differences, especially in the regimes where the curves are close.
- [Section II-C2] The conventional precoder selection method is only cited to reference [9] and is not described in the manuscript. Because the paper's main comparison and complexity reduction claim are against this baseline, the baseline selection rule, its use of SVD information, and its complexity model must be specified. Without this, the reported performance ordering and the claimed complexity advantage cannot be independently assessed or reproduced.
- [Algorithm 2 and Eq. (18)] There is an inconsistency in the definition of the optimization objective. Algorithm 2 sets the OP metric as λ1 for ν=1 and λ1+λ2 for ν>1, whereas Eq. (18) and Algorithm 2 line 16 use the normalized ratio (Σ_{r=1}^ν λ_r)/(Σ_{j=1}^M λ_j). These two objectives can lead to different RIS configurations because the denominator changes with θ. The authors should clarify which objective is actually maximized and align the algorithm with the equation.
- [Section II-D] In Algorithm 2, the final configuration is selected as the best among the offspring configurations Ψson only; the parent configurations θmax and θsubmax are not included in the final comparison. If a parent has a higher OP than every offspring, the algorithm would return a suboptimal configuration. The selection step should also compare against the parents, or the algorithm should justify why this is unnecessary.
minor comments (5)
- [Section I] In the introduction, "RIS phase sift optimization" appears to be a typo for "RIS phase shift optimization."
- [Section II-A, Eq. (4)] The symbol z is used for wavelength, which is unconventional and can be confused with a spatial coordinate; using λ_w or similar would improve readability.
- [Section II-C1] Equation (15) selects a beam by maximizing the inner product with the first half of the dominant right singular vector, but the text does not explain why only the first half is used for a cross-polarized array; a brief justification would help.
- [Section III] The figures use different line styles and colors, but the caption and legend are dense; adding markers or a table of the exact rate differences at selected SNR points would make the claimed gains easier to verify.
- [Section II-D] The definition of the offspring generation "Method 1 in [14]" is not reproduced; since the MCA is central to the paper, a concise description of the cross-swapping operation should be included.
Circularity Check
No significant circularity: the SVD-based RIS optimization and codebook precoder selection are evaluated by independent simulation, with no fitted parameter or self-citation chain forcing the reported result.
full rationale
The paper's derivation chain is not circular. The RIS phase optimization (Eq. 18) maximizes a singular-value objective derived from the SVD of the cascaded channel F = GΦH. The precoder selection (Eqs. 15-17) maps the resulting right-singular vectors onto the 3GPP Type-I codebook via inner-product maximization. The performance metric (Eqs. 8-9) is the ZF achievable rate computed from the effective precoded channel FWν. These are distinct quantities connected by a heuristic approximation: the codebook quantizes the singular vectors, so the selected Wν is not equal to V by construction, and the rate includes ZF noise enhancement (the inverse diagonal terms), which is not equated to the optimization objective. No parameter is fitted to the reported rate; MCA's T and NRIS,new are fixed algorithmic parameters, and the benchmark effective-rank method is adopted from external literature [14]. There are no self-citations by the current authors, so no uniqueness or ansatz is imported from prior author work. The abstract's 'consistently outperforms' is stronger than the figures support (the paper itself notes similar performance for NRIS=16 and for uniform singular-value cases), but that is a correctness/overclaim issue, not circularity. Verdict: self-contained derivation; score 0.
Assumptions & free parameters
free parameters (2)
- T (number of random RIS configurations in MCA) =
200
- NRIS,new (number of crossed RIS elements in MCA) =
6
assumptions (4)
- domain assumption No direct link between gNodeB and UE
- domain assumption Perfect CSI at the UE
- domain assumption Zero Doppler shift and time-invariant channels
- domain assumption Zero-forcing equalizer at the receiver
Cite this review
Pith. "Pith review of Joint Phase Shift Optimization and Precoder Selection for RIS-Assisted 5G NR MIMO Systems." pith.science (2026). https://pith.science/paper/3DHPEYQL
@misc{pith2026250523154,
author = {Pith},
title = {Pith review of: Joint Phase Shift Optimization and Precoder Selection for RIS-Assisted 5G NR MIMO Systems},
year = {2026},
howpublished = {\url{https://pith.science/paper/3DHPEYQL}},
note = {Machine review of arXiv:2505.23154}
}
abstract
By intelligently reconfiguring wireless propagation environment, reconfigurable intelligent surfaces (RISs) can enhance signal quality, suppress interference, and improve channel conditions, thereby serving as a powerful complement to multiple-input multiple-output (MIMO) architectures. However, jointly optimizing the RIS phase shifts and the MIMO transmit precoder in 5G and beyond networks remains largely unexplored. This paper addresses this gap by proposing a singular value ($\lambda$)-based RIS optimization strategy, where the phase shifts are configured to maximize the dominant singular values of the cascaded channel matrix, and the corresponding singular vectors are utilized for MIMO transmit precoding. The proposed precoder selection does not require mutual information computation across subbands, thereby reducing time complexity. To solve the $\lambda$-based optimization problem, maximum cross-swapping algorithm (MCA) is applied while an effective rank-based method is utilized for benchmarking purposes. The simulation results show that the proposed precoder selection method consistently outperforms the conventional approach under $\lambda$-based RIS optimization.
Figures
Reference graph
Works this paper leans on
-
[14]
Rank optimization for MIMO channel with RIS: simulation and measurement,
S. Meng and et al. , “Rank optimization for MIMO channel with RIS: simulation and measurement,” IEEE Wireless Commun. Lett. , vol. 13, no. 2, pp. 437–441, Feb. 2024
work page 2024
-
[9]
A tutorial on downlink precoder selection strategies for 3GPP MIMO codebooks,
X. Fu and et al., “A tutorial on downlink precoder selection strategies for 3GPP MIMO codebooks,” IEEE Access , vol. 11, pp. 138 897–138 922, 2023
work page 2023
-
[1]
S. A. H. Mohsan and Y . Li, “A contemporary survey on 6G wireless networks: potentials, recent advances, technical challenges and future trends,” 2023. [Online]. Available: https://arxiv.org/abs/2306.08265
work page Pith review arXiv 2023
-
[2]
Intelligent re- flecting surface-aided wireless communications: a tutorial,
Q. Wu, S. Zhang, B. Zheng, C. You, and R. Zhang, “Intelligent re- flecting surface-aided wireless communications: a tutorial,” IEEE Trans. Commun., vol. 69, no. 5, pp. 3313–3351, 2021
work page 2021
-
[3]
C. Pan and et al. , “Reconfigurable intelligent surfaces for 6G systems: principles, applications, and research directions,” IEEE Commun. Mag. , vol. 59, no. 6, pp. 14–20, 2021
work page 2021
-
[4]
Deep learning-based CSI feedback for RIS-assisted multi-user systems,
J. Guo, X. Yang, C.-K. Wen, S. Jin, and G. Y . Li, “Deep learning-based CSI feedback for RIS-assisted multi-user systems,” 2024. [Online]. Available: https://arxiv.org/abs/2003.03303
arXiv 2024
-
[5]
Intelligent reflecting surface enhanced wireless network via joint active and passive beamforming,
Q. Wu and R. Zhang, “Intelligent reflecting surface enhanced wireless network via joint active and passive beamforming,”IEEE Trans. Wireless Commun., vol. 18, no. 11, pp. 5394–5409, 2019
2019
-
[6]
Weighted sum- rate maximization for reconfigurable intelligent surface aided wireless networks,
H. Guo, Y .-C. Liang, J. Chen, and E. G. Larsson, “Weighted sum- rate maximization for reconfigurable intelligent surface aided wireless networks,” IEEE Trans. Wireless Commun. , vol. 19, no. 5, pp. 3064– 3076, 2020
work page 2020
Show all 14 references
-
[7]
Physical layer procedures for data,
ETSI, “Physical layer procedures for data,” ETSI TS 138 214 V16.2.0, July 2020
2020
-
[8]
Massive MIMO evolution toward 3GPP release 18,
H. Jin and et al. , “Massive MIMO evolution toward 3GPP release 18,” IEEE J. Sel. Areas Commun. , vol. 41, no. 6, pp. 1635–1654, 2023
2023
-
[10]
A review of codebooks for CSI feedback in 5G new radio and beyond,
Q. Ziao and Y . Haifan, “A review of codebooks for CSI feedback in 5G new radio and beyond,” China Communications , vol. 22, no. 2, pp. 112–127, 2025
2025
-
[11]
Study on channel model for frequencies from 0.5 to 100 GHz,
ETSI, “Study on channel model for frequencies from 0.5 to 100 GHz,” ETSI TR 138 901 V16.1.0, November 2020
2020
-
[12]
Reconfigurable intelligent surfaces (RIS): communication models, channel models, channel estimation and evaluation methodol- ogy,
ETSI GR, “Reconfigurable intelligent surfaces (RIS): communication models, channel models, channel estimation and evaluation methodol- ogy,” ETSI GR RIS 003 V1.1.1, June 2023
2023
-
[13]
Wireless communications with reconfigurable intelligent surface: path loss modeling and experimental measurement,
W. Tang and et al. , “Wireless communications with reconfigurable intelligent surface: path loss modeling and experimental measurement,” IEEE Trans. Wireless Commun. , vol. 20, no. 1, pp. 421–439, 2021
2021
Reviewed August 7, 2026 · model on record in the stance chip above.
Discussion (0). Sign in to comment.