{"id":"380d235a-157d-4ece-be93-ecba511e8137","arxiv_id":"2411.08383","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Optimizing fluid antenna positions and receive beamforming at a secondary user increases primary-user detection probability in cognitive radio spectrum sensing, with the sensing objective reducing to maximizing received SNR.","lead":"This paper proposes a fluid antenna system at a cognitive radio secondary user, jointly optimizing antenna positions and receive beamforming to maximize the probability of detecting the primary user's signal. The authors show by simulation that this positioning freedom improves detection over fixed, random, and discrete-selection antenna benchmarks.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Eq. (17) misstates |w^H h|^2: the matrix Φ is defined as ΣΣ^H, but the channel model in Eq. (3) requires Φ = Σ 1_{Lt}1_{Lt}^H Σ^H; for Lt>1 the SCA surrogate is not a lower bound on the true detection SNR.","rationale":"The reader identified the perfect-CSI assumption as the weakest assumption, which is a legitimate practical caveat. However, a more load-bearing concern is that the mathematical reformulation at the heart of the SCA subproblem appears inconsistent with the stated channel model. If Eq. (17) is wrong, the algorithm's objective is not the true SNR, the lower bound property fails, and the central optimality claim is unsupported. This is an internal correctness issue rather than a missing assumption, so it should take priority. I recommend a conditional verdict: the paper can be accepted only if the authors correct the definition of Φ, verify that the SCA surrogate is a valid lower bound under the corrected expression, and confirm that the simulation conclusions still hold. The FAS concept and problem formulation are sensible, and the mistake may be fixable, but the current manuscript does not make the case.","tokens_in":7961,"tokens_out":7665,"duration_ms":73109,"concrete_test":"Independently compute both sides of Eq. (17) for a small case, e.g., Lt = Lr = 2, N = 2, using random Σ and antenna positions t_1, t_2: evaluate |w^H h|^2 directly from Eq. (3) and compare it with α + β(t_n) + 2 Re{f^H(t_n)Ω} using Φ = ΣΣ^H. If they differ for any instance, Eq. (17) is incorrect. Then re-run the simulation of Fig. 3 with the corrected Φ = Σ 1_{Lt}1_{Lt}^H Σ^H; if the 'Proposed' curve no longer consistently exceeds EAS and FPA, the reported gains are an artifact of the algebraic error.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The paper's central claim is that the alternating optimization algorithm with SCA-based antenna positioning maximizes the detection probability. That claim rests on the reformulation of |w^H h|^2 in Eq. (17). Given the channel model h = F(¯t)^H Σ 1_{Lt} (Eq. 3), where 1_{Lt} is the all-ones vector, the true objective is |w^H h|^2 = w^H F^H Σ 1_{Lt}1_{Lt}^H Σ^H F w = Tr(Σ 1_{Lt}1_{Lt}^H Σ^H F w w^H F^H). The paper instead defines Φ = ΣΣ^H and rewrites the objective as Tr(1^H Σ^H F w w^H F^H Σ 1), with α, β(t_n), and Ω built from Φ = ΣΣ^H. Unless Lt = 1, ΣΣ^H ≠ Σ 1_{Lt}1_{Lt}^H Σ^H, so Eq. (17) is not an identity for |w^H h|^2. Consequently the concave lower bound f^l(t_n) in Eq. (21) is not necessarily a lower bound of the true receive SNR, and the monotone convergence and local-optimality claims in Section III are not established. This is an internal inconsistency in the derivation, independent of the CSI availability concern. The simulations may still show gains if the authors implemented the correct expression, but the manuscript as written does not support the claimed optimality of the proposed algorithm.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This letter studies a cognitive radio (CR) network in which a secondary user (SU) equipped with N fluid antennas performs energy detection to sense signals from a primary user (PU). The paper formulates an optimization problem that jointly chooses the detection threshold, the receive beamforming vector, and the fluid antenna positions, subject to a false-alarm constraint, a minimum antenna-spacing constraint, and a positional region constraint. The authors derive a closed-form detection threshold from the false-alarm constraint, a closed-form matched-filter beamformer, and an alternating optimization (AO) algorithm in which the antenna positions are updated using successive convex approximation (SCA). Simulation results are presented to show that the proposed scheme outperforms fixed-position, random-position, and exhaustive-antenna-selection benchmarks. The central claim is that the proposed AO algorithm maximizes the detection probability subject to the false-alarm constraint and provides significant improvements over fixed-position antenna schemes.","tokens_in":8300,"tokens_out":11078,"duration_ms":107615,"significance":"The idea of exploiting fluid antenna position flexibility to improve spectrum sensing is timely and may be of interest to the CR and FAS communities. The paper offers a clean problem formulation, a closed-form threshold, and a structured AO solution. However, the significance is heavily tempered by two issues. First, the derivation in Eq. (17) misstates the objective |w^H h|^2 for the stated channel model with Lt>1, so the SCA lower bound and the resulting antenna-position updates are not optimizing the actual detection SNR. Second, the optimal beamformer and the position updates require perfect knowledge of the PU-SU channel h, an assumption that is never stated and is particularly strong in a spectrum-sensing context. If the objective error is corrected and the CSI assumption is made explicit, the paper could be a useful contribution; as written, the main claims are not supported.","major_comments":[{"comment":"The reformulation of the objective is incorrect for the stated channel model. From Eq. (3), h = F(\\bar{t})^H \\Sigma 1_{Lt}, so |w^H h|^2 = |1_{Lt}^H \\Sigma^H F(\\bar{t}) w|^2 = Tr(\\Sigma 1_{Lt}1_{Lt}^H \\Sigma^H F(\\bar{t}) w w^H F(\\bar{t})^H). The paper defines \\Phi = \\Sigma\\Sigma^H and rewrites the objective as Tr(\\Phi F(\\bar{t}) w w^H F(\\bar{t})^H). These expressions are generally not equal when Lt > 1. Since the simulation setup uses Lt = Lr = 4, the lower bound f^l(t_n) in Eq. (21) is not a lower bound on the true detection SNR, and the monotone convergence and local-optimality claims in Section III are not established for the actual problem. This error is internal to the derivation and does not depend on the CSI availability concern. The authors should either replace \\Phi with \\Sigma 1_{Lt}1_{Lt}^H \\Sigma^H and re-derive \\alpha, \\beta(t_n), and \\Omega, or explicitly restrict the model to Lt = 1.","section":"Section III-B, Eq. (17)"},{"comment":"The optimal beamforming solution w^o = h/||h|| (Eq. (15)) and the SCA position updates in Eq. (17) assume that the SU has exact knowledge of the PU-SU channel vector h, including the path response matrix \\Sigma and the angles embedded in F(\\bar{t}). This perfect-CSI assumption is not stated anywhere in the manuscript. In cognitive radio spectrum sensing, the SU typically does not have a priori knowledge of the PU channel, and this assumption strongly affects the implementability and the interpretation of the reported gains. The authors should state this assumption explicitly, discuss the effect of channel estimation errors, or frame the results as an upper bound for an idealized scenario.","section":"Sections II-A and III-A"},{"comment":"The paper claims that the AO algorithm 'obtains the locally optimal solution when alternately optimizing these sub-problems,' but no convergence proof is provided. While the SCA construction yields a global lower bound that is tight at the current point, which gives monotone improvement of the surrogate objective, convergence to a stationary point or local optimum of the original nonconvex problem (11) is not established. The authors should either add a brief convergence argument (e.g., along the lines of majorization-minimization theory) or soften the claim to state that the algorithm monotonically improves the detection SNR.","section":"Section III, introductory paragraph"}],"minor_comments":[{"comment":"The text says 'H1 is the alternative hypothesis to represent the absence of PU signals'; it should say 'presence' rather than 'absence'.","section":"Section II-B"},{"comment":"The notation in Eq. (17) contains typos and inconsistencies: '1H Lr' should presumably be '1_{Lt}', and the subscripts on \\Sigma and f(t_n) are garbled (e.g., '\\Sigma k' and 'f_k(t_n)'). These should be cleaned up to avoid ambiguity.","section":"Section III-B, Eq. (17)"},{"comment":"The simulation description says 'Monte Carlo simulations are conducted with 100 time average,' but it is unclear whether this means 100 independent channel realizations or 100 noise realizations. Please specify the number of channel realizations and whether the reported curves are averaged over them.","section":"Section IV"},{"comment":"The EAS benchmark selects N antennas from '2N fixed positions.' This is a very small candidate set; please clarify the relationship between these fixed positions and the region S, and consider using a larger discrete grid to make the comparison more meaningful.","section":"Section IV, benchmark descriptions"},{"comment":"Remark 1 states that the FAS scheme demonstrates 'superior performance in both PU and SU detection,' but the manuscript only considers SU detection of the PU signal. This statement should be revised or clarified.","section":"Remark 1"}],"recommendation":"major_revision","confidential_remarks":"The paper is a straightforward application of fluid-antenna concepts to cognitive radio spectrum sensing, and the main contribution is the joint threshold/beamforming/position optimization. The technical error in Eq. (17) is serious and load-bearing; it must be fixed before the paper can be considered. The unstated perfect-CSI assumption is also a significant limitation that needs to be disclosed. If these issues are addressed, the paper may be suitable as a letter in its scope. I did not find evidence of deliberate circularity or fitted free parameters; the threshold is derived from the false-alarm constraint, and the SCA is a standard lower-bound-based method. The heavy self-citation pattern is not inappropriate for this field but could be trimmed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know about this paper. The idea is genuinely new: using fluid antennas at a secondary user to reposition antennas for better primary-user detection is a sensible combination of two literatures, and the paper provides a complete algorithm. The problem is in the antenna-positioning derivation: the key matrix identity in Eq. (17) is wrong, so the claimed optimality does not follow from the math as written.\n\nWhat's good: the false-alarm threshold in Eq. (12) is the standard energy-detection threshold and follows correctly from the Q-function; the matched-filter beamforming w = h/||h|| is correct under perfect CSI; and the overall AO-SCA structure is reasonable. The simulation results consistently show gains over FPA, RPA, and EAS benchmarks.\n\nThe soft spots are serious. The stress-test note is valid: the channel model in Eq. (3) gives h = F^H Σ 1_{Lt}. Then |w^H h|^2 = Tr(Σ 1_{Lt}1_{Lt}^H Σ^H F w w^H F^H). The paper instead writes the trace with 1_{Lr} and then defines Φ = ΣΣ^H. For Lt > 1 (and the simulation uses Lt = 4), ΣΣ^H ≠ Σ 1_{Lt}1_{Lt}^H Σ^H. Consequently, the SCA surrogate in Eqs. (17)–(21) is not a lower bound on the true SNR, and the local-optimality claim in Section III is not established. This is not a minor typo; it is the load-bearing step of the antenna-position optimization. The simulations might still be correct if the authors implemented the right expression, but the manuscript as written does not support the claim. Also, the abstract says the algorithm maximizes detection probability without the local-optimality qualification.\n\nA second problem is the unstated perfect-CSI assumption. The beamformer and position updates both need h exactly. In cognitive radio, the SU typically does not know the PU channel. The paper never lists this as an assumption and does not discuss estimation error. That alone would merit revision. Minor issues: Eq. (8) has a typo, and the simulation section lacks angle distributions, error bars, and code.\n\nWho is this for? Researchers in fluid antenna systems and cognitive radio. The concept is worth pursuing, and the paper could be valuable after a serious revision. As it stands, I would not cite it. My recommendation: send it to peer review, but the referees should be asked to verify the derivation step by step; the authors need to fix Eq. (17), state the CSI assumption, and provide reproducible simulation details.","headline":"New idea, flawed derivation: Eq. (17) misstates the objective, so the optimality claim is unsupported.","tokens_in":8828,"tokens_out":4456,"would_cite":false,"duration_ms":37421,"reading_group":"maybe","serious_thinker":"no","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper argues that a secondary user with fluid antennas can maximize the probability of detecting a primary user under a false-alarm cap by jointly choosing beamforming and antenna positions, and that the resulting…","keywords":["fluid antenna system","cognitive radio","spectrum sensing","detection probability","false alarm constraint","alternating optimization","successive convex approximation","energy detection"],"falsifier":"Run the proposed algorithm with a channel estimate $\\hat h = h + e$, $e \\sim \\mathcal{CN}(0,\\sigma_e^2 I)$, and plot the achieved $P_d$ against $\\sigma_e^2$ at a fixed false-alarm cap $\\delta$; the paper assumes perfect channel knowledge, so if the gain over fixed-position antennas collapses for realistic estimation errors, the practical reach of the claim is bounded. A hardware check would compare the predicted versus measured $P_d$ for actual fluid-antenna movements in a channel where moving the antenna changes path amplitudes as well as phases.","tokens_in":7778,"feed_emoji":"📡","tokens_out":9232,"duration_ms":88423,"temperature":0.7,"pith_summary":"The paper tries to show that a cognitive radio can detect a primary user more reliably if its receive antennas are movable rather than fixed. It models a secondary user with several fluid antennas that can slide within a small square region; because changing a position changes the phase combination of multipath arrivals, some positions make the primary signal add up more strongly at the receiver. The analysis reduces the sensing problem to maximizing the post-combining signal-to-noise ratio, and provides an alternating optimization algorithm: the detection threshold is set by the false-alarm cap, the beamformer is fixed to point along the primary-user channel, and each antenna position is updated by successive convex approximation. Simulations show this approach finding positions that beat fixed, random, and exhaustive-selection antenna layouts, especially at low primary transmit power and low false-alarm limits.","feed_headline":"Movable antennas boost primary-user detection in cognitive radio","feed_subtitle":"Algorithm tunes fluid antenna positions and beamforming to raise detection while capping false alarms.","key_machinery":"The central object is the fluid antenna system (FAS), an array of $N$ antennas whose positions $\\bar t = [t_1,\\dots,t_N]$ can be adjusted inside a finite spatial region $S$; the engine of the argument is the monotone relationship between the post-combining SNR $\\gamma = P|w^H h|^2/\\sigma_n^2$ and the detection probability $P_d = Q\\big((\\tau - \\sigma_n^2(1+\\gamma))\\sqrt{K}/(\\sigma_n^2(1+\\gamma))\\big)$: raising $\\gamma$ always raises $P_d$ at a fixed false-alarm level. That lets the paper split the original non-convex problem: the threshold $\\tau$ is set by the false-alarm cap, the beamformer $w$ is set to the matched filter, and the antenna-position subproblem becomes maximizing $|w^H h|^2$, with the far-field channel $h = F(\\bar t)^H \\Sigma 1_{L_t}$ changing through phase shifts as antennas move. Each fluid antenna's position update is then solved by successive convex approximation, replacing the objective and the minimum-distance constraints with concave or linear lower bounds obtained from Taylor expansions; the whole procedure iterates between beamforming and position updates until convergence.","core_discovery":"The central claim is that antenna position is itself a resource for spectrum sensing: a secondary user with $N$ fluid antennas can maximize the probability of detecting the primary user under a false-alarm constraint by jointly optimizing the receive beamforming vector and the antenna positions. The paper proves that, because the detection probability $P_d$ is monotonically increasing in the output SNR $\\gamma = P|w^H h|^2/\\sigma_n^2$, the optimal detection threshold is fixed by the false-alarm cap alone ($\\tau^o = \\sigma_n^2 Q^{-1}(\\delta)/\\sqrt{K} + \\sigma_n^2$) and the optimal beamformer is the matched filter $w^o = h/\\|h\\|$. The remaining problem, choosing where to put each fluid antenna, is non-convex and is solved by an alternating-optimization loop whose inner antenna-position subproblem is handled with successive convex approximation based on a concave lower bound of the objective. In the paper's simulations the resulting scheme converges within about thirty iterations and outperforms fixed-position, random-position, and exhaustive-antenna-selection benchmarks.","pith_inferences":["Because the gain in the model comes from phase-only coherence, a natural extension is to relax the far-field assumption: in near-field or blockage-dominated channels, moving an antenna changes path amplitudes and angles, and the SCA update would need a different surrogate objective.","The paper gives no channel-estimation procedure; a testable follow-up would quantify how much of the reported $P_d$ gain survives when the matched-filter beamformer is built from a pilot-based estimate $\\hat h$ rather than $h$.","The same SNR-maximizing view should transfer to other detectors that are monotone in post-combining SNR, such as cyclostationary or eigenvalue-based sensing, though their thresholds and test statistics would need separate derivations.","If the primary signal is wideband or the antenna movement spans more than a small fraction of the wavelength, the single-carrier phase model may miss frequency-selective effects; evaluating detection probability over subbands would test whether position optimization still helps."],"forward_implications":["Detection probability at the secondary user is monotonically increasing in output SNR, so any position or beamforming change that raises $\\gamma$ directly improves $P_d$ without touching the false-alarm cap.","The optimal threshold is $\\tau^o = \\sigma_n^2 Q^{-1}(\\delta)/\\sqrt{K} + \\sigma_n^2$, so threshold setting decouples from antenna placement and depends only on noise power, sample count, and the allowed false-alarm probability.","For any fixed antenna layout, the best receive beamformer is the matched filter $w=h/\\|h\\|$; no other unit-norm combining vector can give a higher detection probability.","In the simulated setup the fluid-antenna optimized layout converges within about thirty iterations and outperforms fixed-position, random-position, and exhaustive-antenna-selection benchmarks, with the largest margins at low primary transmit power and low $\\delta$.","Because the false-alarm constraint is always active at its upper limit, all remaining optimization effort is spent on maximizing the received SNR, making the FAS benefit an SNR gain rather than a threshold or diversity gain."],"supporting_citations":[{"why":"Supplies the Gaussian approximation of the energy-detection test statistic under $H_0$ and $H_1$ that underlies the false-alarm and detection probability formulas.","marker":"[5]"},{"why":"Introduces the fluid antenna system concept and the spatial degree of freedom that the paper repurposes for spectrum sensing.","marker":"[9]"},{"why":"Gives the planar far-field fluid-antenna channel model in which antenna movement changes only path phases, which is the basis for the position-optimization formulation.","marker":"[11]"},{"why":"Provides the convex solver used to run the SCA-based antenna-position subproblem in each alternating-optimization iteration.","marker":"[31]"}],"fun_headline_variants":["Fluid antennas outdo fixed positions for cognitive radio sensing","Tune antenna spots and beamformer to catch primary users","Joint antenna-position and beamforming optimization lifts detection","FAS sensing: where you put the antenna matters for CR","Move antennas to maximize primary detection with low false alarms"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the secondary user knows the exact channel vector $h$ from the primary user and that sliding a fluid antenna changes only the phases of the multipath components, not their strengths or arrival angles; if either condition fails, the matched-filter receiver and the position-update rule no longer deliver the claimed detection gains.","fun_headline_variants_meta":{"raw":{"variants":["Fluid antennas outdo fixed positions for cognitive radio sensing","Tune antenna spots and beamformer to catch primary users","Joint antenna-position and beamforming optimization lifts detection","FAS sensing: where you put the antenna matters for CR","Move antennas to maximize primary detection with low false alarms"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000228,"raw_usage":{"total_tokens":1479,"prompt_tokens":951,"completion_tokens":528,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":567,"completion_tokens_details":{"reasoning_tokens":449}},"tokens_in":567,"tokens_out":528,"duration_ms":6099,"temperature":1.0,"reasoning_tokens":449,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T21:39:18.862880+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the proposed algorithm with a channel estimate $\\hat h = h + e$, $e \\sim \\mathcal{CN}(0,\\sigma_e^2 I)$, and plot the achieved $P_d$ against $\\sigma_e^2$ at a fixed false-alarm cap $\\delta$; the paper assumes perfect channel knowledge, so if the gain over fixed-position antennas collapses for realistic estimation errors, the practical reach of the claim is bounded. A hardware check would compare the predicted versus measured $P_d$ for actual fluid-antenna movements in a channel where moving the antenna changes path amplitudes as well as phases.","supporting_citations":[{"cited_title":"Enhancing spectrum sensing via re- configurable intelligent surfaces: Passive or active sensing and how many reflecting elements are needed?","cited_arxiv_id":null,"evidence_quote":"Supplies the Gaussian approximation of the energy-detection test statistic under $H_0$ and $H_1$ that underlies the false-alarm and detection probability formulas."},{"cited_title":"Fluid antenna system,","cited_arxiv_id":null,"evidence_quote":"Introduces the fluid antenna system concept and the spatial degree of freedom that the paper repurposes for spectrum sensing."},{"cited_title":"A tutorial on fluid antenna system for 6G networks: Encompassing communication theory, optimization methods and hard- ware designs,","cited_arxiv_id":null,"evidence_quote":"Gives the planar far-field fluid-antenna channel model in which antenna movement changes only path phases, which is the basis for the position-optimization formulation."},{"cited_title":"CVX: MATLAB software for disciplined convex programming,","cited_arxiv_id":null,"evidence_quote":"Provides the convex solver used to run the SCA-based antenna-position subproblem in each alternating-optimization iteration."}],"review_version":1}