Semi-blind Channel Estimation for Massive MIMO LEO Satellite Communications

Abdollah Masoud Darya; Saeed Abdallah

REVIEW 2 major objections 4 minor 19 references

Semi-blind Channel Estimation for Massive MIMO LEO Satellite Communications

T0 review · 2 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper claims that a decision-directed semi-blind estimator can track a LEO satellite channel by periodically re-estimating from recently detected data symbols, outperforming a pilot-based estimator and approaching perfect-channel…

desk verdict The MDD-SB estimator idea is reasonable and the NMSE results are plausible, but the near-genie SER claim does not line up with the algorithm as written and needs an explicit evaluation fix. read the letter →

arxiv 2411.13944 v1 pith:ZCQPVFNU submitted 2024-11-21 eess.SP

Abdollah Masoud Darya , Saeed Abdallah This is my paper · ORCID

classification eess.SP

keywords massiveMIMOLEOsatellitecommunicationssemi-blindchannelestimationdecision-directedagingleast-squaressymbolerrorrate

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This letter argues that a low-Earth-orbit satellite link can keep its channel estimate fresh without frequent pilot retransmissions, by periodically re-estimating the uplink channel from recently detected data symbols. It proposes two semi-blind estimators: a decision-directed estimator that uses pilots plus detected data, and a modified version that drops the outdated pilots and refreshes the estimate using only the most recent detected data symbols. In simulations of a 600 km, 30 GHz, 10-user massive MIMO system, the modified estimator keeps its normalized mean square error below the best pilot-based estimator at all reported blocks after block 0, and its symbol error rate close to a Genie-aided detector that knows the channel perfectly at every block. The practical payoff, if the decision-feedback loop stays stable, is lower pilot overhead and more reliable direct-to-cell satellite service in a fast-changing channel.

What carries the argument

The load-bearing mechanism is Algorithm 1's decision-directed loop. Starting from a pilot-based least-squares estimate at block 0, the receiver performs zero-forcing equalization on the current data block, detects the symbols with a minimum-distance rule, then recomputes a least-squares channel estimate using those detected symbols in place of pilots. The updated estimate is averaged and becomes the equalizer for the next block, so the channel is refreshed every few blocks without new pilot overhead. Each update has per-subcarrier complexity $O(D M K + DK)$, linear in the block length $D$, number of antennas $M$, and number of users $K$; dropping the pilot symbols from the update is what distinguishes MDD-SB from the earlier DD-SB estimator and is what lets it outrun channel aging.

What would settle it

Run Algorithm 1 with the paper's 600 km, 30 GHz, 16-QAM parameters at SNR below 1 dB, where the unmodified DD-SB already falls behind pilot-only least squares, and record NMSE and SER at blocks 10, 20, and 30; if the estimate error grows with block index or the SER gap to the Genie-aided detector widens, the decision-feedback assumption would be falsified.

Watch

Extended reading notes

Core claim

The central claim is that the modified decision-directed semi-blind (MDD-SB) estimator in Algorithm 1 mitigates channel aging by replacing pilot-only re-estimation with periodic least-squares updates computed from the most recent detected data symbols. Under the paper's simulation setup, an MDD-SB update every fifth block keeps the NMSE below the P-bound, a pilot-based estimator with perfect channel knowledge at block 0, for blocks 5 through 50; the same run keeps SER close to the Genie-aided detector across blocks 5, 10, 15, and 20, while the pilot-based bound degrades to SER above $10^{-1}$ by block 20. The paper also finds that the unmodified DD-SB estimator, which retains pilots alongside detected data, improves NMSE over pilot-only least squares for SNR above roughly 1 dB but loses at lower SNR.

Load-bearing premise

The loop's stability rests on the assumption that the symbols detected from the current channel estimate are reliable enough to train the next estimate, with no bound or low-SNR analysis given for the case where detection errors cascade.

Editorial extensions

If this is right

An MDD-SB refresh every five blocks removes the need to retransmit pilots at that cadence, freeing frame resources for data.
The periodically refreshed estimate outperforms the stale pilot-based bound in NMSE, so the estimator is a viable alternative to more frequent pilot transmission under channel aging.
SER stays near the Genie-aided detector even when estimation is run once every five blocks, so the scheme can approach full-CSI detection quality at a fraction of the pilot overhead.
The per-iteration complexity of $O(D M K + DK)$ is linear in block length, number of antennas, and number of users, making frequent updates computationally affordable.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The feedback loop's stability is the unexamined boundary: at low SNR, or if the initial pilot estimate is poor, detected-symbol errors could accumulate. A natural extension is to derive a threshold SNR or maximum refresh interval below which MDD-SB should fall back to pilots.
The same periodic re-estimation mechanism could transfer to other fast-aging links, such as high-speed train or UAV channels, where pilot overhead is similarly costly.
A concrete testable extension is to sweep the refresh interval and the satellite Doppler spread together; the paper fixes one interval (every 5 blocks) and one Doppler profile, so the boundary of the near-Genie regime is not yet mapped.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Quick take: the modified decision-directed semi-blind estimator is a sensible idea, and the NMSE behavior is credible. But the paper's headline claim—SER comparable to a Genie-aided detector—does not match its own Algorithm 1. For the first MDD-SB update at block 5, the loop detects using the initial pilot-based estimate from block 0, then updates. So the SER at block 5 should be no better than the frozen-pilot bound. The only way to get near-genie SER there is to re-detect the block after updating, which is not in the algorithm and would be an in-sample comparison. That is a load-bearing ambiguity that must be resolved before the central claim can be trusted.

What is new is narrow but real: coming from decision-directed semi-blind estimation applied to LEO massive MIMO with channel aging, with the sensible modification of discarding stale pilots and re-estimating from recent detected data only. The least-squares algebra is internally consistent, the complexity analysis is honest, and the footnote about the trade-off with neighboring blocks shows the authors understand the design space. The NMSE result—MDD-SB maintaining lower error than a fixed pilot-based estimate over time—is plausible and the main useful finding.

The soft spots, in order. First, the SER evaluation ambiguity is major and affects the abstract and conclusion. Second, the baseline is weak: the so-called optimal pilot-based estimator is just a frozen initial estimate, never updated. The paper cites Kalman tracking [11] and deep-learning prediction [13] as prior aging-mitigation work but compares against neither, which makes the improvement claim less convincing. Third, error propagation in the decision feedback is acknowledged only in passing; no analysis or low-SNR study is given. Minor point: no code or data are provided, which makes the simulation results harder to verify.

This is a paper for people working on direct-to-cell LEO links who want a simple, statistics-free alternative to Kalman or learned predictors. If the authors clarify the SER evaluation and add one strong baseline, the work would be worth a serious referee. As it stands, the NMSE contribution is worth engaging with, but the SER claim is not supported as written.

Referee Report

2 major / 4 minor

Summary. This letter proposes decision-directed semi-blind channel estimation for massive MIMO LEO satellite uplinks. Two estimators are presented: a decision-directed estimator (DD-SB) that uses pilot and detected data symbols together, and a modified estimator (MDD-SB) that periodically re-estimates the channel using only the most recent detected data symbols to mitigate channel aging. The MDD-SB estimator is evaluated through Monte Carlo simulation against an optimal pilot-based benchmark with perfect channel knowledge at block 0 (P-bound) and a Genie-aided detector with perfect channel knowledge at every block, reporting NMSE and SER results. The paper claims that MDD-SB outperforms the P-bound in NMSE and achieves SER comparable to the Genie-aided detector.

Significance. The proposed approach is attractive for its simplicity: it avoids additional pilot transmissions, does not require channel or noise statistics, and uses closed-form least-squares updates with low per-iteration complexity. The paper provides an explicit algorithm and a complexity analysis. If the reported results are reproducible, the MDD-SB estimator would be a useful low-complexity alternative to Kalman filtering or deep-learning-based channel tracking in LEO mMIMO. However, the SER evaluation in Fig. 4 appears inconsistent with the causal operation of Algorithm 1, and the robustness of the feedback loop to decision errors is not established. These issues affect the central claim and require clarification or additional evidence.

major comments (2)

[Section IV, Fig. 4; Algorithm 1] The SER evaluation in Fig. 4 appears inconsistent with the causal order of Algorithm 1. At block ı=5, the detection step (line 3) uses the channel estimate inherited from block 0 (the initial P-LS estimate), because the update in line 5 occurs after detection. Since the P-bound benchmark has perfect channel knowledge at block 0, the MDD-SB detector cannot have a lower SER than the P-bound at block 5; yet Fig. 4 reportedly shows MDD-SB performing comparably to the Genie-aided detector. Please clarify whether the reported SER is for the first detection (line 3) or for a re-detection performed after line 5 using the updated estimate. If the latter, the comparison is in-sample because the same received block is used to form the estimate and evaluate the detection, and it cannot support the claim of 'SER performance comparable to that of a Genie-aided detector' in the abstract.
[Section III-C and Section IV] The robustness of the MDD-SB estimator to decision-error propagation is not demonstrated. The text acknowledges 'some error accumulation' in the NMSE curves but provides no analysis, bound, or low-SNR simulation. If the initial pilot estimate is poor or the operating SNR is low, detected symbols fed back in line 5 may be unreliable and could cause the channel estimate to diverge. The paper should at least report the SER/NMSE over the SNR range of Fig. 2 or discuss conditions under which the feedback loop remains stable.

minor comments (4)

[Section III-C, Algorithm 1] The loop start index n is not specified in Algorithm 1 or in the surrounding text. Please state which block index is used for the first update (e.g., n=5) and reconcile this with the statement that updates are performed 'once every 5 blocks'.
[Section IV, Fig. 2 caption] The caption says 'using P pilot and D data symbols' without giving the values; the text later states P=15 and D=15, which should be stated in the caption for clarity.
[Section II, Eqs. (7)-(9)] The simplification from (1) to (7) is not fully derived; in particular, the definitions of hLoS_k and hNLoS_k in (8) and (9) appear to have absorbed some factors from (5) and (6). A short explanation would improve readability.
[Section IV] The acronym 'P-bound' is used without definition; it is first described as 'optimal pilot-based estimator (perfect channel knowledge at block 0)' in Section IV, but the name 'P-bound' should be introduced explicitly.

Circularity Check

1 steps flagged · score 6.0 of 10

Fig. 4's near-Genie SER at block 5 is only reproducible by an in-sample re-detection with the same block's estimate; the central SER claim is not an out-of-sample prediction.

fitted input called prediction [Algorithm 1 (lines 3–5) and Section IV, Fig. 4 discussion]
"ˆX^{D,ı}_c = (Y^{D,ı}_c A^+)^T ⊘ ˆ¯H^{new}_c; Detect ˆX^{D,ı}_c ⇒ ˆ¯X^{D,ı}_c; ˆH^{new}_c = (Y^{D,ı}_c A^+)^T ⊘ ˆ¯X^{D,ı}_c ... These detected data symbols are then used to update the channel estimate ˆ¯H^{new}_c, which is then used to detect the received data symbols in the next block. ... The MDD-SB estimator performs comparably to the GA detector."

At block 5, Algorithm 1 has not yet performed any MDD-SB update: line 4 detects block 5 using the block-0 pilot estimate, so its SER cannot beat the P-bound, which has the exact block-0 channel. The near-Genie block-5 SER shown in Fig. 4 is reproducible only if block 5 is re-detected after line 5, but line 5 constructs the updated estimate from the same received block Y^{D,5} and from symbols detected from that same block. The channel estimate and the detection then share the same noise realization, making the reported SER an in-sample fit of the estimator rather than an independent out-of-sample prediction.

full rationale

The paper's NMSE comparisons are self-contained Monte Carlo simulations: the P-LS, DD-SB, and MDD-SB estimators are defined by explicit least-squares expressions, benchmarks are external (P-bound with perfect block-0 knowledge, Genie-aided detector with perfect per-block knowledge), and no fitted constant is renamed as a prediction. Those NMSE results do not reduce by construction to their inputs. However, the strongest advertised result—SER comparable to a Genie-aided detector—depends on the evaluation protocol of Fig. 4, which is not specified in Algorithm 1. Algorithm 1 is causal: detection of block i uses the currently available estimate, and the updated estimate is used for the next block. Consequently, at block 5 the detector uses only the block-0 pilot estimate, so a causal evaluation cannot give near-Genie SER at block 5. Reproducing Fig. 4 requires re-detecting the same block with the estimate formed from that same block's received data, i.e., the estimation and detection use the same noise realization. That is an in-sample evaluation, known to be optimistically biased, and it makes the central SER claim circular rather than predictive. Because the NMSE results retain independent content but the strongest SER claim is affected, the appropriate circularity score is 6.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central claim rests on known system-model assumptions (perfect array response, perfect satellite Doppler compensation, Rician channel model) and on an unanalyzed decision-feedback stability assumption. The hand-chosen update interval and block size are tuning parameters that affect the reported performance.

free parameters (2)

MDD-SB update interval = every 5 blocks
The channel estimate is refreshed once every 5 data blocks in the simulations. The paper states the optimal interval depends on Doppler, computational capability, and accuracy requirements and is to be found offline, but no selection criterion or sensitivity analysis is given.
Data symbols per MDD-SB update = 15
The modified estimator uses the most recent 15 detected data symbols per update, matching the 15 pilot symbols used at block 0. This choice is not justified beyond matching the pilot count.

assumptions (4)

domain assumption The array response matrix A in Eq. (3) is perfectly known and time-invariant over the estimation horizon.
A+ is computed once and used in P-LS, DD-SB, and MDD-SB (Section III-A, Eqs. (14) to (17)). The paper notes A is time-invariant and can be computed offline, but does not address array response estimation error.
domain assumption Satellite-induced Doppler is perfectly compensated at the UT side, leaving only user Doppler and multipath as sources of channel aging.
The received signal model in Eq. (11) includes a satellite Doppler compensation term V_SAT, and the introduction states ephemeris and GNSS data are used. The aging effect simulated is therefore only the residual user-side variation.
domain assumption The channel follows the Rician model in Eqs. (1) to (9) with known K-factor and number of paths, and the reference channel in Eq. (18) is the exact true channel for NMSE.
The NMSE and SNR definitions in Eqs. (18) and (19) use this reference. If the model does not match a real LEO channel, the reported performance numbers do not transfer.
domain assumption Detected data symbols after ZF equalization and minimum-distance detection are reliable enough to serve as training signals in the decision-directed loop.
This is the load-bearing assumption of the MDD-SB estimator (Algorithm 1, steps 3 to 5). The paper gives no error-propagation analysis; it only notes empirically that 'some error accumulation can be seen' in Section IV.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Semi-blind Channel Estimation for Massive MIMO LEO Satellite Communications." pith.science (2026). https://pith.science/paper/ZCQPVFNU

@misc{pith2026241113944,
  author       = {Pith},
  title        = {Pith review of: Semi-blind Channel Estimation for Massive MIMO LEO Satellite Communications},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZCQPVFNU}},
  note         = {Machine review of arXiv:2411.13944}
}

read the original abstract

This letter proposes decision-directed semi-blind channel estimation for massive multiple-input multiple-output low-Earth-orbit satellite communications. Two semi-blind estimators are proposed. The first utilizes detected data symbols in addition to pilot symbols. The second, a modified semi-blind estimator, is specially designed to mitigate the channel-aging effect caused by the highly dynamic nature of low-Earth-orbit satellite communication channels -- an issue that adversely impacts the performance of pilot-based estimators. Consequently, this modified estimator outperforms an optimal pilot-based estimator in terms of normalized mean square error and achieves symbol error rate performance comparable to that of a Genie-aided (perfectly known channel) detector. The trade-offs between the proposed estimators are also examined.

Figures

Figures reproduced from arXiv: 2411.13944 by the authors.

**Figure 2.** The NMSE vs SNR trend of the P-LS and DD-SB [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 3.** The NMSE versus block index of the proposed MDD [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Figures from the paper (1 more)

**Figure 4.** Figure 4: The SER versus SNR trends of the proposed MDD [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

19 extracted references · 19 canonical work pages

[11]

Block-based Kalman channe l tracking for LEO satellite communication with massive MIMO,

T. Y ue, A. Liu, and X. Liang, “Block-based Kalman channe l tracking for LEO satellite communication with massive MIMO,” IEEE Commu- nications Letters , vol. 27, no. 2, pp. 645–649, Feb. 2023

work page 2023
[13]

Deep learning-based channel prediction for LEO satellite massive MIMO communication system,

Y . Zhang et al. , “Deep learning-based channel prediction for LEO satellite massive MIMO communication system,” IEEE Wireless Com- munications Letters , vol. 10, no. 8, pp. 1835–1839, Aug. 2021

work page 2021
[1]

No more “no service

L. Laursen, “No more “no service”: Cellphones will incre asingly text via satellite,” IEEE Spectrum , vol. 60, no. 1, pp. 52–55, Jan. 2023

work page 2023
[2]

Satellite swarm-ba sed antenna arrays for 6G direct-to-cell connectivity,

D. Tuzi, T. Delamotte, and A. Knopp, “Satellite swarm-ba sed antenna arrays for 6G direct-to-cell connectivity,” IEEE Access , vol. 11, pp. 36 907–36 928, 2023

work page 2023
[3]

Study on New Radio (NR) to support non-terrestria l networks,

3GPP , “Study on New Radio (NR) to support non-terrestria l networks,” 3rd Generation Partnership Project (3GPP), Tech. Rep. 38.8 11, 09 2020, version 15.4.0

work page 2020
[4]

Massive MIMO transmission for LEO satellite commu- nications,

L. Y ou et al. , “Massive MIMO transmission for LEO satellite commu- nications,” IEEE Journal on Selected Areas in Communications , vol. 38, no. 8, pp. 1851–1865, Aug. 2020

work page 2020
[5]

Random access with massive MIMO-OTFS in LEO satellite communications,

B. Shen et al. , “Random access with massive MIMO-OTFS in LEO satellite communications,” IEEE Journal on Selected Areas in Commu- nications, vol. 40, no. 10, pp. 2865–2881, Oct. 2022

work page 2022
[6]

Dist ributed massive MIMO for LEO satellite networks,

M. Y . Abdelsadek, G. K. Kurt, and H. Y anikomeroglu, “Dist ributed massive MIMO for LEO satellite networks,” IEEE Open Journal of the Communications Society , vol. 3, pp. 2162–2177, 2022

work page 2022

Show all 19 references

[7]

Quasi-synchronous random access for massive MIMO- based LEO satellite constellations,

K. Ying et al. , “Quasi-synchronous random access for massive MIMO- based LEO satellite constellations,” IEEE Journal on Selected Areas in Communications, vol. 41, no. 6, pp. 1702–1722, Jun. 2023

2023
[8]

Channel estimation for L EO satellite massive MIMO OFDM communications,

K.-X. Li, X. Gao, and X.-G. Xia, “Channel estimation for L EO satellite massive MIMO OFDM communications,” IEEE Transactions on Wire- less Communications , vol. 22, no. 11, pp. 7537–7550, Nov. 2023

2023
[9]

Joint Bayesian channel estimation and data detection for OTFS systems in LEO satellite communications,

X. Wang et al. , “Joint Bayesian channel estimation and data detection for OTFS systems in LEO satellite communications,” IEEE Transactions on Communications , vol. 70, no. 7, pp. 4386–4399, Jul. 2022

2022
[10]

Performance analysis of FDD massive MIMO systems under channel aging,

R. Chopra et al., “Performance analysis of FDD massive MIMO systems under channel aging,” IEEE Transactions on Wireless Communications , vol. 17, no. 2, pp. 1094–1108, Feb. 2018

2018
[12]

Channel tracking sche me using block-based Kalman algorithm for MIMO wireless syste ms,

J. Zhu, M. F. Siyau, and K. K. Loo, “Channel tracking sche me using block-based Kalman algorithm for MIMO wireless syste ms,” in 2007 IEEE 8th W orkshop on Signal Processing Advances in Wire less Communications. IEEE, 2007, pp. 1–5

2007
[14]

Deep learning for channel estimation: Interpretation, performance, and comparison,

Q. Hu et al. , “Deep learning for channel estimation: Interpretation, performance, and comparison,” IEEE Transactions on Wireless Com- munications, vol. 20, no. 4, pp. 2398–2412, Apr. 2021

2021
[15]

Angle of arrival estimation in multi-user mas- sive MIMO LEO satellite networks,

I. Boukhedimi et al. , “Angle of arrival estimation in multi-user mas- sive MIMO LEO satellite networks,” in ICC 2023-IEEE International Conference on Communications . IEEE, Oct. 2023, pp. 6287–6293

2023
[16]

Deep learning (DL)-based channel prediction and hybrid beamforming for LEO satellite massive MIMO system,

Y . Zhang et al. , “Deep learning (DL)-based channel prediction and hybrid beamforming for LEO satellite massive MIMO system,” IEEE Internet of Things Journal , vol. 9, no. 23, pp. 23 705–23 715, Dec. 2022

2022
[17]

Channel estimation techniques based on pilot arrange- ment in OFDM systems,

S. Coleri et al. , “Channel estimation techniques based on pilot arrange- ment in OFDM systems,” IEEE Transactions on broadcasting , vol. 48, no. 3, pp. 223–229, Sep. 2002

2002
[18]

Semi-blind channel estimation for multiuser massive MIMO systems,

E. Nayebi and B. D. Rao, “Semi-blind channel estimation for multiuser massive MIMO systems,” IEEE Transactions on Signal Processing , vol. 66, no. 2, pp. 540–553, Jan. 2018

2018
[19]

Optimal and mismatched detection of QAM signals in fast fading channels with imperf ect channel estimation,

Y . V . Zakharov, V . M. Baronkin, and J. Zhang, “Optimal and mismatched detection of QAM signals in fast fading channels with imperf ect channel estimation,” IEEE Transactions on Wireless Communications , vol. 8, no. 2, pp. 617–621, Feb. 2009

2009

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [11]

Block-based Kalman channe l tracking for LEO satellite communication with massive MIMO,

T. Y ue, A. Liu, and X. Liang, “Block-based Kalman channe l tracking for LEO satellite communication with massive MIMO,” IEEE Commu- nications Letters , vol. 27, no. 2, pp. 645–649, Feb. 2023

work page 2023

[2] [13]

Deep learning-based channel prediction for LEO satellite massive MIMO communication system,

Y . Zhang et al. , “Deep learning-based channel prediction for LEO satellite massive MIMO communication system,” IEEE Wireless Com- munications Letters , vol. 10, no. 8, pp. 1835–1839, Aug. 2021

work page 2021

[3] [1]

No more “no service

L. Laursen, “No more “no service”: Cellphones will incre asingly text via satellite,” IEEE Spectrum , vol. 60, no. 1, pp. 52–55, Jan. 2023

work page 2023

[4] [2]

Satellite swarm-ba sed antenna arrays for 6G direct-to-cell connectivity,

D. Tuzi, T. Delamotte, and A. Knopp, “Satellite swarm-ba sed antenna arrays for 6G direct-to-cell connectivity,” IEEE Access , vol. 11, pp. 36 907–36 928, 2023

work page 2023

[5] [3]

Study on New Radio (NR) to support non-terrestria l networks,

3GPP , “Study on New Radio (NR) to support non-terrestria l networks,” 3rd Generation Partnership Project (3GPP), Tech. Rep. 38.8 11, 09 2020, version 15.4.0

work page 2020

[6] [4]

Massive MIMO transmission for LEO satellite commu- nications,

L. Y ou et al. , “Massive MIMO transmission for LEO satellite commu- nications,” IEEE Journal on Selected Areas in Communications , vol. 38, no. 8, pp. 1851–1865, Aug. 2020

work page 2020

[7] [5]

Random access with massive MIMO-OTFS in LEO satellite communications,

B. Shen et al. , “Random access with massive MIMO-OTFS in LEO satellite communications,” IEEE Journal on Selected Areas in Commu- nications, vol. 40, no. 10, pp. 2865–2881, Oct. 2022

work page 2022

[8] [6]

Dist ributed massive MIMO for LEO satellite networks,

M. Y . Abdelsadek, G. K. Kurt, and H. Y anikomeroglu, “Dist ributed massive MIMO for LEO satellite networks,” IEEE Open Journal of the Communications Society , vol. 3, pp. 2162–2177, 2022

work page 2022