{"id":"045f86f2-08b9-4110-9578-becfcb5d7832","arxiv_id":"2507.01289","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Rotating the antenna array on a UAV base station can reduce inter-cell interference and improve sum rate by about 10% in LoS multi-cell networks.","lead":"This paper proposes rotating the antenna arrays on drone base stations to reduce interference between neighboring cells. The approach could improve network speed by about 10% in interference-heavy situations without adding hardware.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The sum-rate formula (25) is dimensionally inconsistent with the SINR definition (24); taken literally it predicts that reducing interference lowers the rate, so the reported 10% gain is unsupported unless the simulation uses a different expression.","rationale":"The paper's stated contribution is a 10% sum-rate improvement from UAV rotation. That percentage is a numeric output of a simulation whose governing equations are presented in Section IV-A. Eq. (25) is not a minor assumption; it is the objective function being optimized. A formula that is dimensionally inconsistent and monotonically decreasing in SINR cannot be the basis for the reported gain. My concern is not about environmental scope (the LoS assumption is explicit and defensible for rural/suburban mmWave deployments), but about the internal correctness of the central arithmetic. The reader's verdict of CONDITIONAL is appropriate: the idea is plausible and the qualitative mechanism (rotating a square array moves side lobes away from neighboring users) is supported by Fig. 4, but the paper must correct the sum-rate formula, resolve the rotation-transformation inconsistency, and make the simulation reproducible before the 10% claim can be accepted. I therefore keep the verdict unchanged while disagreeing with the reader's choice of weakest assumption.","tokens_in":17553,"tokens_out":10289,"duration_ms":110364,"concrete_test":"Reimplement the three-cell simulation of Section V using the SINR expression in Eq. (24) and compute the average GU rate in two ways: (a) log2(1 + tilde_eta), the standard Shannon rate; (b) the printed Eq. (25), log2(1 + P/(tilde_eta + L_{c,c,k_c} sigma_n^2)). Compare both against the curves in Fig. 8 for M = 8, W = 32. If the values match (b), the claimed 10.8% improvement is a consequence of a decreasing function of SINR and is invalid. If they match (a), Eq. (25) is a typographical error and the qualitative claim survives, but the paper must be corrected before the result is accepted.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section IV-A defines the SINR in Eq. (24) as tilde_eta = P / (sum_u sum_ku (P L_{c,c,k_c})/(K_u L_{u,c,k_c}) tilde_g + L_{c,c,k_c} sigma_n^2), which is dimensionless. The sum rate in Eq. (25) is then written as R = sum_c sum_kc log2(1 + P / (tilde_eta + L_{c,c,k_c} sigma_n^2)). This mixes watts with a dimensionless SINR and, more seriously, is decreasing in tilde_eta: as interference decreases and tilde_eta rises, the argument P/(tilde_eta + L sigma^2) falls, so the rate falls. Since the proposed rotation is specifically designed to reduce interference, the printed formula would predict that rotation degrades the sum rate, the opposite of the ~10% improvement claimed in Fig. 8 and the abstract. Therefore either (i) the simulations used the correct Shannon expression log2(1+tilde_eta) and Eq. (25) is a typo, in which case the analytical core of Section IV-A is not reliable as written, or (ii) the simulations used Eq. (25), in which case the headline result is an artifact. Support for the central claim therefore cannot be assessed from the paper as presented. A secondary inconsistency: the rotation transformation in Eq. (18) has tilde_beta_i = beta_i - omega, whereas Lemma 2 uses tilde_beta_i = beta_i + pi/2 and assumes alpha_i + beta_i = pi/2, a constraint that does not hold when the GU is not in the positive quadrant; this undermines the claim that the search range can be restricted to [0, pi/2).","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies a downlink multi-cell network in which each cell is served by a UAV acting as an aerial base station with a two-dimensional planar antenna array. Under a line-of-sight position-based channel model, the authors analyze how rotating the UAV's array changes the beamforming gain toward users in neighboring cells while preserving the gain toward the intended user, formulate a sum-rate maximization problem over the rotation angles, and propose a low-complexity alternating optimization algorithm (AUR) benchmarked against exhaustive search. Simulation results are reported for three-cell deployments with random user positions, claiming an average sum-rate improvement of about 10% over a fixed orientation in interference-limited scenarios.","tokens_in":17903,"tokens_out":16949,"duration_ms":286762,"significance":"The idea of exploiting array orientation as an additional controllable degree of freedom for inter-cell interference mitigation is timely and potentially useful, as it requires only position information and no additional RF chains or joint beamforming. The paper provides an explicit analytical framework, an alternating optimization algorithm with complexity analysis, an exhaustive-search benchmark, and a robustness evaluation under user-position errors. However, several load-bearing formulas in the analytical core are incorrect as printed, and the numerical evidence for the headline 10% gain is presented without error bars. The qualitative concept is plausible, but the quantitative claims are not yet supported by the manuscript in its current form.","major_comments":[{"comment":"The sum-rate expression in Eq. (25) is inconsistent with the SINR definition in Eq. (24). Eq. (24) defines \\(\\tilde{\\eta}_{c,k_c}\\) as the SINR, so the Shannon rate must be \\(\\log_2(1+\\tilde{\\eta}_{c,k_c})\\). Instead, Eq. (25) writes \\(\\log_2\\!\\left(1 + P/(\\tilde{\\eta}_{c,k_c} + L_{c,c,k_c}\\sigma_n^2)\\right)\\). This expression mixes a power quantity \\(P\\) with a dimensionless SINR plus a noise-power term, and it is decreasing in \\(\\tilde{\\eta}_{c,k_c}\\): as interference decreases, \\(\\tilde{\\eta}\\) increases, the argument \\(P/(\\tilde{\\eta}+L\\sigma^2)\\) decreases, and the rate decreases. Since the proposed rotation is specifically intended to reduce interference, Eq. (25) as printed predicts that rotation degrades the sum rate, the opposite of the effect claimed in the abstract and in Fig. 8. Please correct Eq. (25) to \\(\\log_2(1+\\tilde{\\eta}_{c,k_c})\\) and confirm that Algorithm 1 and all simulations were computed with the corrected expression. If the simulations used Eq. (25) literally, the reported 10% gain is an artifact and must be recomputed.","section":"Section IV-A, Eq. (25)"},{"comment":"The interference-gain formula in Eq. (16) and Proposition 1 is incorrect for the normalized steering vectors defined in Eq. (1). With \\(\\psi_h(\\theta)=M^{-1/2}[1,e^{j\\pi\\cos\\theta},\\ldots,e^{j(M-1)\\pi\\cos\\theta}]\\), the squared inner product is \\(|\\psi_h(\\theta_{k_c})^H \\psi_h(\\theta_{k_u})|^2 = (1/M^2)\\sin^2(M\\Delta\\Omega/2)/\\sin^2(\\Delta\\Omega/2)\\). The printed Eq. (16) has \\((1/M)\\sin^2(\\cdot)/\\sin^2(\\cdot)\\), and the intermediate expression \\((1/M)\\sum_{k=1}^M |e^{j(k-1)\\Delta\\Omega}|\\) is not equal to a squared modulus; it evaluates to 1 independently of \\(\\Delta\\Omega\\). This overestimates the interference gain by a factor of \\(M\\) in each one-dimensional factor, hence by \\(M^2\\) in the two-dimensional product \\(g_h g_v\\). At \\(\\Delta\\Omega=0\\) the printed formula gives \\(M\\) rather than the correct value 1, which is inconsistent with the unit-norm channel vectors used elsewhere in the paper. Please correct Proposition 1 and clarify whether the simulation code used the corrected normalization; if not, the numerical rates and the 10% improvement need to be recomputed with the correct expression.","section":"Section III-A, Proposition 1 and Eq. (16)"},{"comment":"The proof of Lemma 2 is internally inconsistent with Eq. (18) and relies on an invalid geometric constraint. Eq. (18) states that a counterclockwise rotation by \\(\\omega\\) gives \\(\\tilde{\\alpha}_i=\\alpha_i+\\omega\\) and \\(\\tilde{\\beta}_i=\\beta_i-\\omega\\), but the proof of Lemma 2 sets \\(\\tilde{\\beta}_i=\\beta_i+\\pi/2\\) for \\(\\omega=\\pi/2\\). In addition, the proof assumes \\(\\alpha_i+\\beta_i=\\pi/2\\) for all GUs; this identity holds only for GUs in the first quadrant of the UAV-local coordinate system and is not satisfied by the random circular user distributions used in Section V. The identity \\(\\cos(\\pi-\\beta)=\\cos\\beta\\) used in Eq. (20) is also false; the correct identity is \\(\\cos(\\pi-\\beta)=-\\cos\\beta\\). The \\(\\pi/2\\)-periodicity claim may be salvageable by the symmetry of the square array (the horizontal and vertical factors swap), but the argument as written does not establish it. Since the search-range restriction \\([0,\\pi/2)\\) in Eq. (26) and the complexity reduction in Remark 2 rely on this lemma, the proof must be corrected or the algorithm must be applied on \\([0,\\pi)\\) unless periodicity is rigorously established.","section":"Section III-B, Lemma 2 and Eq. (18)"},{"comment":"The headline improvement of approximately 10% is presented in Figs. 8a–8c without any confidence intervals, error bars, or statistical significance tests, even though Table I reports only 50 Monte Carlo trials with randomly dropped users in each trial. Given that the average-rate curves are close (e.g., 4.55 vs. 5.05 bps/Hz for \\(M=8\\)), the reported gain could be within the trial-to-trial variation. Please report the standard error or confidence intervals of the average rate, and preferably also box plots or a paired comparison, so that the existence and magnitude of a nonzero gain can be assessed.","section":"Section V, Fig. 8"}],"minor_comments":[{"comment":"The spacing in 'UA V' is inconsistent (sometimes 'UAV' appears, e.g., in the references); please standardize the notation.","section":"Throughout"},{"comment":"The X2 interface is described as part of 3GPP New Radio; X2 is the LTE inter-base-station interface, while NR uses the Xn interface. Please correct this technical detail.","section":"Section II, X2 interface"},{"comment":"The discrete search set is written as \\(\\omega_u \\in \\{0, \\pi/(2W), \\ldots, \\pi/2\\}\\). With the claimed \\(\\pi/2\\) periodicity, \\(\\pi/2\\) duplicates 0, so the set should be \\(\\{0, \\pi/(2W), \\ldots, \\pi/2 - \\pi/(2W)\\}\\) or the endpoints should be explicitly identified as equivalent.","section":"Algorithm 1"},{"comment":"The caption says 'Black dashed lines indicate the boundaries ... while black dashed lines represent the user distribution regions'; the second phrase should refer to different line types or markers, as the current wording describes both elements identically.","section":"Fig. 7 caption"},{"comment":"There are several typographical errors, including 'beaforming' in Section III-A and 'thea' in Section III-B; a careful proofread is needed.","section":"General"},{"comment":"The abstract and conclusion state 'about 10%' improvement, while Fig. 8 shows values from 9.1% to 11.1%; please either state the range or use a consistent phrasing.","section":"Abstract and Section V"}],"recommendation":"major_revision","confidential_remarks":"The manuscript reports a useful and simple idea, but the analytical errors in Eq. (25), Eq. (16), and Lemma 2 are load-bearing for the main claim. The authors should be asked to provide corrected derivations and rerun the simulations with the correct formulas, and to include a reproducibility statement (e.g., releasing simulation code) given that the numerical results are the main evidence for the claimed gain."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's the short version: the core idea—rotating a UAV's planar array and recomputing MRT to steer sidelobes away from neighboring cells—is sound and worth pursuing. But as written, the paper's own sum-rate formula (25) predicts the opposite of the claimed 10% gain, so the headline result is not supported by the equations. I'd send it back for major revision.\n\nWhat's new: this extends the 6DMA rotation idea to UAV-mounted arrays with a simple alternating search, benchmarked against exhaustive search in a three-cell LoS deployment. The physical picture is credible: with MRT recomputed after rotation, the main lobe stays on the intended user while sidelobes rotate, giving a few dB of SINR and roughly 10% rate gain in interference-limited regimes. That's a legitimate contribution, though incremental—standard array-factor math, no new hardware.\n\nThe soft spots are not minor. Equation (25) defines the sum rate as log2(1 + P/(tilde_eta + L sigma^2)), where tilde_eta is the dimensionless SINR from (24). That's dimensionally wrong (P is watts, tilde_eta is unitless) and, worse, it decreases as tilde_eta rises—so reducing interference would lower the rate. Unless the simulations used the correct log2(1+tilde_eta), the formula contradicts the abstract's 10% claim. The authors need to correct (25) and state which expression generated Figs. 8-10. Second, the rotation transformation in (18) uses tilde_beta_i = beta_i - omega, while Lemma 2's proof uses tilde_beta_i = beta_i + pi/2 and assumes alpha_i + beta_i = pi/2, a constraint that doesn't hold for arbitrary GU positions. So the claimed restriction to [0, pi/2) is not established. There are also no error bars on the 10% figure, and the pure-LoS assumption, while acknowledged, is load-bearing.\n\nNone of this kills the idea; a corrected version could well confirm the trend. The analytical core just isn't reliable in this version.\n\nWho's this for: researchers in UAV placement, 6DMA, or interference management in aerial networks. It deserves a serious referee—the idea is timely—but I'd recommend major revision: fix (25), reconcile (18) with Lemma 2, justify the symmetry restriction, and share code or at least standard deviations for the headline number.","headline":"Plausible idea with a broken printed sum-rate formula; the 10% gain cannot be assessed as written, but the mechanism is worth a major revision.","tokens_in":18469,"tokens_out":6572,"would_cite":false,"duration_ms":72599,"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":"This paper claims that rotating the antenna array of a UAV base station, while keeping the same maximum-ratio beamforming, reduces interference to neighboring cells and improves the multi-cell sum rate by about 10 percent in line-of-sight…","keywords":["UAV rotation","inter-cell interference","aerial base station","beamforming gain","maximum ratio transmission","mmWave MIMO","sum-rate optimization","line-of-sight channel"],"falsifier":"A field experiment or ray-tracing simulation in a moderately scattering environment (for example, an urban setting with Rician K-factor near 0 dB) that compares a fixed-orientation ABS with a rotated ABS under identical MRT beamforming: if the measured interference at neighboring-cell users does not drop by roughly the predicted amount, or the intended user's gain degrades noticeably, the central claim fails. A simpler laboratory check would measure the two-user array factor of a square patch array as a function of mechanical rotation and compare the null depths and side-lobe levels against Proposition 1.","tokens_in":17325,"feed_emoji":"🛸","tokens_out":4699,"duration_ms":48777,"temperature":0.7,"pith_summary":"The paper argues that physically rotating a UAV-mounted antenna array provides a nearly free degree of freedom for interference management in multi-cell aerial networks. Under line-of-sight channels, the beamforming weights of a maximum-ratio transmitter are determined purely by the relative angles between the UAV and its users, so rotating the array changes the gain pattern toward users in neighboring cells while leaving the intended user's beamforming gain untouched. The authors derive a closed-form interference gain that separates into horizontal and vertical factors and use it to formulate a sum-rate maximization over rotation angles, solved by a low-complexity alternating scheme. Simulations show that in interference-limited regimes, rotating the three UAV base stations improves the average multi-cell user rate by roughly 10 percent over fixed orientations, with performance close to exhaustive search. The significance is that interference mitigation is achieved without extra RF chains, CSI feedback, or joint beamforming optimization.","feed_headline":"Rotating UAV antennas lifts multi-cell sum rate by ~10 percent","feed_subtitle":"Spinning the array keeps the served user's gain while reshaping side lobes away from neighboring cells.","key_machinery":"The object that carries the argument is the interference gain $g_{\\{k_c,k_u\\}}(\\omega)$, derived in Proposition 1 as a product of two Fejér-kernel-type factors: $g_h(\\Phi) g_v(\\Phi)$, where each factor is $\\sin^2(M\\Delta\\Omega/2)/(M\\sin^2(\\Delta\\Omega/2))$ with $\\Delta\\Omega$ the difference of directional cosines between the intended and interfered user. Under MRT beamforming the gain is exactly the squared channel correlation $|h_{u,c,k_c} h_{u,u,k_u}^H|^2$, so rotation enters because a counterclockwise rotation $\\omega$ maps $(\\alpha_i,\\beta_i)$ to $(\\alpha_i+\\omega,\\beta_i-\\omega)$ for every user while keeping the pitch $\\gamma_i$ fixed. This mapping makes the interference gain a one-dimensional function of $\\omega$ for each UAV, lets the authors restrict the search to $[0,\\pi/2)$ via Lemma 2, and yields the alternating AUR algorithm whose per-UAV update is a one-dimensional search over $W$ discrete angles.","core_discovery":"The central claim is that rotating a UAV base station's downward-facing square antenna array, while recomputing the same position-based MRT beamformer, can reshape the array's interference pattern on the ground and thereby reduce inter-cell interference without sacrificing the beamforming gain toward the intended ground user. The paper establishes this through an interference-gain analysis: under a pure line-of-sight channel, the correlation between the serving and interfering channel steering vectors factors into a product of two Dirichlet-like gains, one horizontal and one vertical, each a function of the directional cosines of the UAV-user angles. Because the azimuth angles of all users transform by the same rotation angle while the pitch angle stays fixed, the interference gain becomes a controllable function of the rotation angle, and a $\\pi/2$ rotational symmetry of the square array limits the search interval. The proposed alternating UAV rotation algorithm greedily updates one rotation angle at a time and is shown in simulation to capture most of the exhaustive-search gain, improving the average per-user rate by about 10 percent in the interference-limited high-SNR regime.","pith_inferences":["The same angular-rotation mechanism could transfer to terrestrial base stations with mechanically steerable arrays or to reconfigurable intelligent surfaces, where the phase profile rotates instead of the physical array.","The factorization of the interference gain suggests an analytic shortcut: choose the rotation angle that nulls the dominant interferer's directional-cosine difference, potentially avoiding the discrete search altogether.","Because rotation preserves the intended user's gain while reshuffling side-lobe exposure, deployments with clustered users should see larger gains, a testable prediction beyond the paper's random-user simulations.","Under Rician fading with significant multipath, the predicted gain would likely shrink as the K-factor drops; the same correlation framework could quantify that degradation."],"forward_implications":["In interference-limited line-of-sight deployments, rotating UAV base stations can yield roughly 10 percent average per-user rate gains without additional hardware or CSI feedback.","The $\\pi/2$ symmetry of square arrays cuts the search space, making rotation optimization practical with linear complexity $O(LNW)$ in the number of UAVs.","The rotation gain grows with antenna array size because narrower beams allow finer interference steering, and shrinks as user density rises because one rotation angle must serve many interfered users.","The scheme remains robust to positioning errors up to about 20 meters RMS, retaining nearly full gain, which supports low-rate X2-based coordination between UAVs.","UAV rotation can be integrated with trajectory planning and user scheduling as an additional spatial control dimension for future aerial networks."],"supporting_citations":[{"why":"Establishes that beamforming vectors in mmWave UAV communications can be computed from UAV and GU positions, the basis for the position-based MRT design.","marker":"[8]"},{"why":"Derives beam weights from user-UAV geometry, a precedent for the angle-based beamforming used here.","marker":"[31]"},{"why":"Models UAVs with directional antennas to characterize ground coverage, motivating orientation as a design parameter that the paper extends to interference shaping.","marker":"[32]"},{"why":"Introduces 6D movable antenna technology with flexible antenna position and rotation, the conceptual source for treating UAV rotation as a new degree of freedom.","marker":"[33]"},{"why":"Provides discrete position and rotation optimization for movable antennas, the methodology that the alternating UAV rotation algorithm adapts.","marker":"[34]"},{"why":"Supplies the X2 interface protocol that lets neighboring ABSs exchange position and control information, enabling the distributed coordination the algorithm relies on.","marker":"[39]"}],"fun_headline_variants":["Rotating UAV antennas lift multi-cell sum rate by ~10%","UAV rotation trims inter-cell interference, lifts sum rate 10%","Spinning UAV arrays reshape interference, improving multi-cell rates","Rotating UAV base stations cut interference, add 10% sum rate","UAV antenna rotation reduces interference, raising sum rate ~10%"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The analysis assumes every UAV-to-ground link is a pure line-of-sight channel with negligible scattering, so the channel is fully described by the array steering vector; if multipath or blockage is significant, position-based MRT is no longer optimal and rotating the array may not produce the predicted interference reduction.","fun_headline_variants_meta":{"raw":{"variants":["Rotating UAV antennas lift multi-cell sum rate by ~10%","UAV rotation trims inter-cell interference, lifts sum rate 10%","Spinning UAV arrays reshape interference, improving multi-cell rates","Rotating UAV base stations cut interference, add 10% sum rate","UAV antenna rotation reduces interference, raising sum rate ~10%"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000933,"raw_usage":{"total_tokens":4030,"prompt_tokens":1017,"completion_tokens":3013,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":633,"completion_tokens_details":{"reasoning_tokens":2921}},"tokens_in":633,"tokens_out":3013,"duration_ms":21776,"temperature":1.0,"reasoning_tokens":2921,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T20:57:06.586025+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A field experiment or ray-tracing simulation in a moderately scattering environment (for example, an urban setting with Rician K-factor near 0 dB) that compares a fixed-orientation ABS with a rotated ABS under identical MRT beamforming: if the measured interference at neighboring-cell users does not drop by roughly the predicted amount, or the intended user's gain degrades noticeably, the central claim fails. A simpler laboratory check would measure the two-user array factor of a square patch array as a function of mechanical rotation and compare the null depths and side-lobe levels against Proposition 1.","supporting_citations":[{"cited_title":"A Survey on Millimeter-Wave Beamforming Enabled UA V Communica- tions and Networking,","cited_arxiv_id":null,"evidence_quote":"Establishes that beamforming vectors in mmWave UAV communications can be computed from UAV and GU positions, the basis for the position-based MRT design."},{"cited_title":"Lightweight 3-D Beamform- ing Design in 5G UA V Broadcasting Communications,","cited_arxiv_id":null,"evidence_quote":"Derives beam weights from user-UAV geometry, a precedent for the angle-based beamforming used here."},{"cited_title":"Position and Orientation Planning of the UA V With Rectangle Coverage Area,","cited_arxiv_id":null,"evidence_quote":"Models UAVs with directional antennas to characterize ground coverage, motivating orientation as a design parameter that the paper extends to interference shaping."},{"cited_title":"6DMA Enhanced Wireless Network with Flexible Antenna Position and Rotation: Opportunities and Challenges,","cited_arxiv_id":null,"evidence_quote":"Introduces 6D movable antenna technology with flexible antenna position and rotation, the conceptual source for treating UAV rotation as a new degree of freedom."},{"cited_title":"6D Movable Antenna Enhanced Wireless Network via Discrete Position and Rotation Opti- mization,","cited_arxiv_id":null,"evidence_quote":"Provides discrete position and rotation optimization for movable antennas, the methodology that the alternating UAV rotation algorithm adapts."},{"cited_title":"Evolved Universal Terrestrial Radio Access Network (E- UTRAN); X2 Application Protocol (X2AP),","cited_arxiv_id":null,"evidence_quote":"Supplies the X2 interface protocol that lets neighboring ABSs exchange position and control information, enabling the distributed coordination the algorithm relies on."}],"review_version":1}