{"id":"5264639b-8a0f-43e5-8376-34fa36da9b35","arxiv_id":"2501.10694","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"Movable antenna positions and transmit covariance are jointly optimized under statistical CSI via deterministic equivalents and alternating optimization, yielding modest energy efficiency gains over fixed arrays.","lead":"A team from Shanghai Jiao Tong University designed an algorithm that moves antennas and sets transmit power to maximize energy efficiency in a MIMO system using only statistical channel information. Their simulations show about 7-15% energy efficiency gains over fixed antenna arrays, with most gains available from a small movement region.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The central EE gains rest on a large-system DE approximation used at N=M=4 without finite-size Monte Carlo validation; if the DE error is comparable to the 7-15% gains, Figs. 1-2 may not describe the true system EE.","rationale":"I read the paper as making a quantitative claim about real MA-enhanced system EE, and the numerical evidence for that claim is generated entirely from the DE objective. The weakest point is therefore the unverified faithfulness of that objective at N=M=4, exactly the reader's weakest_assumption. The other noted issues (receive surrogate asserted without derivation, no stationarity proof) are real but secondary: they affect the algorithm's convergence guarantee, whereas a MC failure of the DE would invalidate the central numerical conclusion even if the algorithm converged. Conversely, if the proposed MC check passes, the quantitative claims survive and the algorithmic gaps can be addressed as missing proofs rather than fatal flaws. Hence I agree with the reader and do not move the verdict.","tokens_in":10196,"tokens_out":8412,"duration_ms":99800,"concrete_test":"Fix Table I parameters and the optimized solutions from Figs. 1-2 (or rerun the algorithm to obtain them). Compute the true average rate in Eq. (7) by Monte Carlo over at least 10^5 independent realizations of ~Sigma (and over the random AoD/AoA draws if desired) for the MA solution and for the TMA, RMA, and UPA benchmarks at X=2.2λ and at X=0.6λ and 1.4λ. Compare the resulting true EE values with the DE-based EE curves. If the MA-vs-UPA gains (about 7-15%) and the saturation beyond about 1.4λ are not reproduced within Monte Carlo error (e.g., +/-0.01 bps/Hz), the central claim is not supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing identification is treating the DE objective in Eqs. (10)-(15) as the true average achievable rate at the simulated scale. The DE is derived from large-system asymptotics in [16], [17], but all numerical claims in Figs. 1-2 use N=M=4 with L=5 paths. The AO algorithm maximizes and plots this approximate objective, not the expectation in Eq. (7), and the convergence argument in Sec. III.E only establishes monotonicity of the approximate sequence. Because the claimed gains over benchmarks are modest (7.3%, 7.1%, 14.6% at X=2.2λ), a DE error of even a few percent could change the ranking or make the saturation in Fig. 2 an artifact. This is especially consequential for MA optimization: the position search may exploit DE-specific structure rather than a true channel property. No Monte Carlo check, error bars, or finite-size correction is reported.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies a single-user MIMO system with movable antennas at both transmitter and receiver, where only statistical CSI is available. The authors formulate the maximization of the long-term energy efficiency over the transmit covariance matrix and the antenna position vectors, subject to minimum-distance and power constraints. Since the expected rate is difficult to evaluate, they adopt deterministic equivalent (DE) approximations from earlier work to obtain surrogate objectives for the transmit and receive variables, and then propose an alternating optimization (AO) algorithm that updates the transmit covariance and positions via Dinkelbach/water-filling and successive convex approximation (SCA). Numerical results with N=M=4 report energy-efficiency gains of 7.3%, 7.1%, and 14.6% over TMA, RMA, and UPA benchmarks at a movement-region size X=2.2λ, and show saturation of the EE as the region size grows.","tokens_in":10430,"tokens_out":4954,"duration_ms":56399,"significance":"If the numerical claims are correct, the paper offers a tractable S-CSI-based design for MA-enhanced MIMO and demonstrates a useful, if modest, energy-efficiency advantage over fixed arrays, including the practically relevant observation that a finite movement region suffices. The paper builds on established DE results rather than re-deriving them, and the algorithmic structure is reasonable. However, the central numerical evidence currently rests on an unvalidated large-system approximation at small system dimensions and on an incomplete convergence argument. The contribution is therefore promising but not yet convincingly established.","major_comments":[{"comment":"The DE reformulation in Eqs. (10)-(15) is derived under large-system asymptotics in refs. [16] and [17], yet all numerical results use N=M=4 with L=5 paths (Table I). The paper optimizes and plots the DE objectives R̄_t and R̄_r, not the true expectation R in Eq. (7), and no Monte Carlo validation, finite-size error bound, or sensitivity analysis is reported. Because the claimed improvements over benchmarks are only 7.3%, 7.1%, and 14.6% at X=2.2λ (Sec. IV.B), a DE error of a few percent could change the ranking of schemes or make the saturation in Fig. 2 an artifact of the approximation. The authors should add a finite-size validation of the DE against Monte Carlo rates for the simulated N=M=4 setting and clearly state whether the plotted curves are the DE objectives or the true average rates.","section":"Sec. III-A, Eqs. (10)-(15), and Sec. IV, Table I, Figs. 1-2"},{"comment":"The convergence analysis in Sec. III-E only shows that the AO sequence has a non-decreasing objective value and that the feasible set is compact; this does not imply convergence to a stationary point of the original problem (9) or even of the DE-reformulated problem. In particular, the transmit update uses a finite-difference gradient in Eq. (20) and an Armijo line search in Eq. (24), but the receive surrogate in Eq. (34) is introduced without a proof that it is a valid concave lower bound with the gradient-matching property, and the statement that the receive update is 'similar' to the transmit case is not a substitute. The authors should either provide a formal SCA convergence argument (sufficient decrease and gradient consistency) or explicitly state the weaker convergence guarantee that is actually established.","section":"Sec. III-E, Eqs. (22)-(24), (34)-(36)"},{"comment":"The simulation setup states that distances and AoDs/AoAs are random, but the paper never specifies how many random realizations are averaged or whether confidence intervals are used. Without this information, the smooth curves in Figs. 1-2 could correspond to a single channel realization, in which case the reported gains are not statistically meaningful, or to an unspecified averaging procedure that should be described. The EE values and the comparison at X=2.2λ need error bars or at least an explicit statement of the number of trials and the averaging method.","section":"Sec. IV, Figs. 1-2"}],"minor_comments":[{"comment":"The instruction 'Construct R̄_r(r) by iterative process (12)' is ambiguous because R̄_r in Eq. (14) depends on t and Q through Γ, Θ, and Φ as well as on r; the step should state explicitly that t and Q are fixed at their most recent values before solving problem (17).","section":"Algorithm 1, Step 5"},{"comment":"The finite-difference step ε1=10^-3 is used to compute numerical gradients for the transmit SCA, but no sensitivity study with respect to ε1 is reported; the accuracy of the optimized positions may depend on this parameter, especially because Q(t) in Eq. (18) is itself evaluated by an iterative inner procedure.","section":"Eq. (20) and Table I"},{"comment":"The notation for the auxiliary matrices Φ, ~Φ, Θ, and ~Θ is dense and their dimensions are not stated in the main text; adding a short paragraph that defines the dimensions and the fixed variables in each DE expression would improve readability.","section":"Eqs. (10)-(15)"},{"comment":"The axis label of Fig. 2 shows only 'Energy Efficiency (bps/Hz)'; since the denominator in Eq. (8) includes power, the unit should be bits/s/Hz/W (or per Joule), and this should be stated consistently in the text.","section":"Fig. 2"}],"recommendation":"major_revision","confidential_remarks":"The stress-test concern lands: the DE approximation at N=M=4 is the load-bearing assumption behind the numerical claims, and the paper does not validate it. The authors do not overclaim the DE as their own derivation, which is good, but the revision must add finite-size Monte Carlo validation and tighten the convergence discussion. If those additions are made satisfactorily, the paper could be acceptable for the journal."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a legitimate incremental result—the first EE-maximizing MA-MIMO design under statistical CSI—but the numerical claims lean on a deterministic equivalent (DE) approximation at N=M=4 with no Monte Carlo check. The DE error could be comparable to the 7.3–14.6% gains, so the ranking and the saturation behavior might not survive a finite-size check.\n\nWhat's new: the problem formulation (EE with S-CSI for MA-MIMO) and the AO algorithm combining DE, Dinkelbach, water-filling, and SCA. The transmit-side Dinkelbach and water-filling are standard, but the EE formulation with APV optimization appears new. The paper is well grounded in prior work and doesn't overclaim conceptual novelty.\n\nWhat's done well: the DE-based reformulation is clearly presented, the algorithms are specified enough to reproduce, complexity analysis is included, and the benchmark comparisons are meaningful. As an engineering design procedure, it's coherent.\n\nMain soft spot: the DE is derived from large-system asymptotics, but the simulations use N=M=4, L=5. The objective being maximized is the approximate DE, not the true average rate, and Figs. 1–2 plot that approximation. The gains are small enough that a few percent DE error could flip the ranking. I'd want a Monte Carlo validation of the DE at those sizes, or at least a quantitative statement of agreement, before trusting the numbers.\n\nSecond: the convergence analysis only shows objective-value non-decreasing, not stationarity of the limit point. The contribution section claims a stationary point is \"efficiently developed,\" but that isn't proven. This is a weaker claim than stated.\n\nThird: the receive surrogate (Eq. 34) is asserted without derivation; there is no proof it's a valid concave lower bound. SCA convergence depends on that property. The transmit side is clearer.\n\nFourth: the simulation setup draws random AoDs/AoAs, but the text doesn't say whether the plotted curves are single realizations or averaged over many. No error bars, no trial count. That should be explicit.\n\nOverall: the core idea is sensible and the paper is a careful incremental contribution. The gaps are fixable. I'd send it to review, but the referees should insist on a finite-size DE check and a proper convergence statement. If the DE error is small, this is a solid engineering result; if not, the main numerical conclusions lack evidence.","headline":"Solid incremental MA-EE design under statistical CSI, but the 7-15% gains rest on an unvalidated large-system approximation at N=M=4; needs a Monte Carlo check before I'd trust the ranking.","tokens_in":10974,"tokens_out":2766,"would_cite":true,"duration_ms":27636,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Movable antennas under statistical CSI can raise MIMO energy efficiency by about 7–15% over fixed arrays, with a finite movement region sufficient for near-optimal performance.","keywords":["movable antennas","fluid antennas","energy efficiency","statistical CSI","MIMO","deterministic equivalent","alternating optimization","successive convex approximation"],"falsifier":"Run a Monte Carlo estimate of the true average rate in Eq. (7) at $N=M=4$ for the same channels and compare the resulting EE landscape with the DE-based objective used in Figs. 1–2; a systematic gap or a shift of the saturation point would show that the reported gains are artifacts of the approximation.","tokens_in":9957,"feed_emoji":"📡","tokens_out":4362,"duration_ms":44099,"temperature":0.7,"pith_summary":"This paper tries to show that a MIMO link whose antennas can be repositioned inside small regions can be made more energy-efficient than a fixed-array link even when the transmitter knows only statistical channel state information, not the instantaneous channel. The authors formulate energy efficiency as the ratio of the average achievable rate to an affine power-consumption model, and propose an alternating-optimization algorithm that updates the transmit covariance matrix and the antenna position vectors in turn. Because the expectation in the rate is intractable, they approximate it with a deterministic equivalent derived for large antenna arrays. Their numerical study reports gains of 7.3%, 7.1%, and 14.6% in energy efficiency over one-sided movable-antenna and fixed-uniform-array benchmarks, and shows that the gain saturates once the antenna movement region reaches about 1.4 wavelengths.","feed_headline":"Movable antennas add up to 14.6% energy-efficiency gain","feed_subtitle":"Statistical-CSI antenna-position optimization beats fixed arrays, and a finite movement region suffices for near-optimal EE.","key_machinery":"The load-bearing object is the deterministic equivalent of the average achievable rate, Eqs. (10)–(15), which replaces the expectation over the random scattered path matrix with fixed-point equations for $\\Phi$, $\\Theta$, $\\tilde{\\Phi}$, $\\tilde{\\Theta}$ and the matrix-valued functions $\\eta$, $\\tilde{\\eta}$. This converts the stochastic rate into a closed-form function of the transmit covariance matrix and of the antenna position vectors separately, making the EE ratio tractable inside an alternating-optimization loop; the inner problems are then handled by Dinkelbach's method, water-filling, and successive convex approximation with numerical gradients.","core_discovery":"The central claim is that energy-efficiency maximization for a movable-antenna MIMO system under statistical CSI can be reformulated into deterministic subproblems and solved by an alternating-optimization algorithm that generates a non-decreasing sequence of EE values. On the transmit side, the algorithm alternates between a Dinkelbach/water-filling update of the covariance matrix and an SCA update of the transmit antenna positions; on the receive side, it runs an SCA update of the receive positions against a quadratic surrogate. With $N=M=4$ antennas and a Rician channel, the optimized system is reported to outperform the benchmarks at every power level and region size, reaching 7.3%, 7.1%, and 14.6% higher EE than the TMA, RMA, and UPA schemes at region size $X=2.2\\lambda$, while EE saturates for region sizes above roughly $1.4\\lambda$.","pith_inferences":["If the DE approximation is tight only asymptotically, the exact small-system gains may differ from the reported percentages; a Monte Carlo check at $N=M=4$ would settle which part of the claimed 7–15% is real.","The same AO/DE machinery should extend to multiuser or secure MIMO, where the average rate is replaced by a weighted sum rate or secrecy rate, and the saturation-with-region-size behavior would suggest how large movement regions need to be in those settings.","As $N$ and $M$ grow, the finite-region saturation hints at a diminishing-returns law in which the optimal region diameter stays bounded in wavelengths rather than growing with aperture; this is a testable prediction the paper does not make."],"forward_implications":["A practical MA transceiver can reap EE gains using slowly varying statistical CSI, avoiding the delay and power cost of instantaneous feedback and antenna movement per coherence block.","The saturation of EE with region size means the antennas do not need large travel ranges; a movement region of roughly $1.4\\lambda$ captures nearly all of the gain.","Because the transmit covariance update has closed-form Dinkelbach/water-filling structure, the AO algorithm scales to larger antenna counts at polynomial cost, with the dominant term $O(L_{\\mathrm{ao}}\\hat{L}_t N^4)$ coming from the SCA position update.","Deploying MAs on only one link side already beats fixed uniform planar arrays, so one-sided retrofits are a viable intermediate step."],"supporting_citations":[{"why":"Supplies the deterministic-equivalent result for the expected MIMO rate that Eqs. (10)–(15) build on.","marker":"[17]"},{"why":"Provides the DE technique used to establish deterministic objective functions in MA-enhanced MIMO systems.","marker":"[16]"},{"why":"Gives the field-response channel model and MIMO capacity characterization for movable antennas that underpin the rate expression and simulation settings.","marker":"[4]"},{"why":"Provides the fractional programming (Dinkelbach) theory used to solve the transmit covariance subproblem.","marker":"[18]"},{"why":"Supplies the affine power-consumption model for the total power in the EE ratio.","marker":"[15]"},{"why":"Motivates the statistical-CSI design by jointly optimizing APVs and transmit covariance for average achievable rate, the problem the paper extends to energy efficiency.","marker":"[9]"}],"fun_headline_variants":["Movable antennas lift energy efficiency by 14.6%","Statistical CSI drives 14.6% EE gain with movable antennas","Finite movement region suffices for optimal MA energy efficiency","Movable-antenna MIMO maximises EE under statistical CSI"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The assumption that the deterministic-equivalent reformulation accurately matches the true average achievable rate at the small antenna count simulated ($N=M=4$), where the technique's large-system justification is not guaranteed to hold.","fun_headline_variants_meta":{"raw":{"variants":["Movable antennas lift energy efficiency by 14.6%","Statistical CSI drives 14.6% EE gain with movable antennas","Finite movement region suffices for optimal MA energy efficiency","Movable-antenna MIMO maximises EE under statistical CSI"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000208,"raw_usage":{"total_tokens":1385,"prompt_tokens":905,"completion_tokens":480,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":521,"completion_tokens_details":{"reasoning_tokens":408}},"tokens_in":521,"tokens_out":480,"duration_ms":5040,"temperature":1.0,"reasoning_tokens":408,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T19:01:07.254749+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run a Monte Carlo estimate of the true average rate in Eq. (7) at $N=M=4$ for the same channels and compare the resulting EE landscape with the DE-based objective used in Figs. 1–2; a systematic gap or a shift of the saturation point would show that the reported gains are artifacts of the approximation.","supporting_citations":[{"cited_title":"Free deterministic equ ivalents for the analysis of MIMO multiple access channel,","cited_arxiv_id":null,"evidence_quote":"Supplies the deterministic-equivalent result for the expected MIMO rate that Eqs. (10)–(15) build on."},{"cited_title":"Sum-rate-optimal statistical precoding for FDD Massive MIMO downlink With deterministic equiv- alents,","cited_arxiv_id":null,"evidence_quote":"Provides the DE technique used to establish deterministic objective functions in MA-enhanced MIMO systems."},{"cited_title":"MIMO capacity characterizat ion for movable antenna systems,","cited_arxiv_id":null,"evidence_quote":"Gives the field-response channel model and MIMO capacity characterization for movable antennas that underpin the rate expression and simulation settings."},{"cited_title":"Energy efficiency in wirel ess networks via fractional programming theory,","cited_arxiv_id":null,"evidence_quote":"Provides the fractional programming (Dinkelbach) theory used to solve the transmit covariance subproblem."},{"cited_title":"Energy efficiency optimization for MIM O broadcast channels,","cited_arxiv_id":null,"evidence_quote":"Supplies the affine power-consumption model for the total power in the EE ratio."},{"cited_title":"Joint be am- forming and antenna movement design for moveable antenna sy stems based on Statistical CSI,","cited_arxiv_id":null,"evidence_quote":"Motivates the statistical-CSI design by jointly optimizing APVs and transmit covariance for average achievable rate, the problem the paper extends to energy efficiency."}],"review_version":1}