{"id":"f46f9f49-b481-4461-9872-377559bcbfc0","arxiv_id":"2412.06456","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":9,"one_line_summary":"A UAV swarm acting as a virtual antenna array with optimized positions, weights, and transmission order can cut interference to non-receiving base stations, though the reported gain claims rely on dubious dB arithmetic.","lead":"This paper treats a swarm of UAVs as a single virtual antenna array to reduce interference to ground base stations during data upload. It formulates a multi-objective optimization problem and proposes a chaotic genetic algorithm to solve it.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 4.95× SINR gain is the ratio of two dB values (857.2498/173.1319≈4.95); dB ratio is meaningless, so the headline interference-mitigation claim lacks quantitative support.","rationale":"The reader's verdict is CONDITIONAL, with the weakest assumption identified as imperfect synchronization/CSI degrading the ideal beam pattern. That is a legitimate structural concern about physical feasibility. However, the most load-bearing weakness in the paper's argument is more basic and internal: the quantitative evidence for the central claim is invalid because Table IV's 'increase by 4.95 times' is computed as the ratio of two dB values (857.2498/173.1319). This is not a question of outside consensus or missing experimental validation; it is an internal inconsistency in the presented numbers. The reader's own strongest_claim notes that 'dB-based arithmetic is questionable,' but the reader did not make this the primary load-bearing concern, instead emphasizing synchronization. In good faith, the synchronization concern matters only if the reported numbers can be taken at face value; since they cannot, the headline claim is unsupported regardless of synchronization quality. I therefore agree partially: the reader saw the issue but placed the weight on a different assumption. A concrete re-computation of Table IV using linear SINR summation is a single, decisive check that would settle whether the 4.95× factor survives. If it does not, the conditional verdict should remain or harden, but I do not see grounds for outright rejection because the qualitative direction—that spatial nulling can reduce interference—is consistent with established CB/beamforming principles. The correct resolution is to require the authors to report linear-domain SINR sums and a precise baseline before the quantitative claim is accepted, which is exactly a CONDITIONAL posture. Hence, verdict_should_be is UNCHANGED relative to the reader's CONDITIONAL.","tokens_in":23919,"tokens_out":3345,"duration_ms":36520,"concrete_test":"Recompute Table IV from the per-BS SINR values: convert each SINR from dB to linear, sum linearly across interfered BSs (per Eq. (8)/(12)), and express as 10log10 of the sum for both Before and After. If the linear-domain before/after ratio is not 4.95 (and per-link values are not itemized), the headline gain is an artifact of dB arithmetic. Also require an explicit definition of the Before baseline.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim in the abstract and conclusion—that CB-based uplink interference mitigation significantly reduces interference to non-receiving BSs—rests on Section VI-B-3 and Table IV. Table IV reports total SINR before/after CB as 173.1319 dB and 857.2498 dB (8 UAVs), and the text says total SINR increases by 4.95 times. Numerically, 857.2498/173.1319 = 4.95, so the claimed factor is the ratio of two dB readings. SINR in dB is 10log10(linear SINR), and the ratio of two dB values is not a power ratio; converting to linear, the difference of 684.1 dB is a factor of roughly 10^68.4, not 4.95. If the table entries are sums of per-BS SINR values expressed in dB, then adding dB values is also invalid—it corresponds to multiplying linear SINRs, not summing interference contributions. Furthermore, the 'Before' baseline is undefined: the paper does not specify how the non-CB received power or array gain is computed (e.g., equal weights with non-coherent combining, or a single UAV), and the text says SINR values are 'cumulative across all interfered BSs' but gives no per-link breakdown. Thus, even granting the reader's synchronization concern, the reported 4.95× and 4.23× improvements are artifacts of invalid dB arithmetic and cannot support the paper's headline claim as stated.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a collaborative-beamforming (CB) approach for uplink interference mitigation in a multi-UAV data collection network, where UAVs form a virtual antenna array to transmit collected data to multiple base stations. It formulates a three-objective optimization problem (transmission time, total SINR of interfered BSs, and propulsion energy) with mixed continuous/discrete decision variables, proves NP-hardness of the problem, and introduces CNSGA-II, an evolutionary algorithm with chaotic initialization, crossover, mutation, and an elimination mechanism. Simulations compare CNSGA-II against MOPSO, MOMVO, MODA, and NSGA-II, and the paper reports that the total SINR of interfered BSs improves by 4.95 times and 4.23 times after applying CB in the 8-UAV and 16-UAV cases, respectively.","tokens_in":24298,"tokens_out":5729,"duration_ms":58069,"significance":"If properly supported, the idea of using a UAV-enabled virtual antenna array to shape beams toward target BSs while suppressing interference to non-receiving BSs is a useful complement to power-control and trajectory-design approaches. The paper provides a detailed system model, a formal NP-hardness argument, and a custom multi-objective evolutionary algorithm with clearly described operators. The claimed algorithmic superiority of CNSGA-II over four baselines is plausible and is supported by Pareto-front and objective-value plots, although the absence of statistical repeats weakens the strength of that claim. The central quantitative claim about interference mitigation, however, rests on the dB arithmetic in Table IV, which is currently invalid as stated.","major_comments":[{"comment":"The statement that 'total SINR increases by 4.95 times after applying the CB method' is obtained by dividing two quantities expressed in dB (857.2498 / 173.1319 ≈ 4.95). A ratio of dB values is not a power ratio. If the entries are 10log10 of cumulative linear SINR, the linear improvement factor is 10^((857.2498−173.1319)/10) ≈ 10^68.4, which is implausibly high; if the entries are sums of per-link dB values, adding dB values corresponds to multiplying linear SINRs, not summing interference contributions. The authors must report linear SINR values or correctly converted ratios, and must specify whether the cumulative SINR is computed in the linear or dB domain.","section":"§VI-B-3, Table IV"},{"comment":"The 'Before' configuration is not specified. The paper states that comparisons are made between 'A2G transmission without adopting CB' and 'A2G transmission with CB', but it does not define how received power and array gain are computed in the non-CB case (e.g., equal weights with non-coherent combining, a single UAV, or an omnidirectional antenna at the VAA centroid). Without this baseline definition and without a per-link or per-BS breakdown, the reported improvement cannot be reproduced or attributed to CB. The authors should define the baseline model and provide per-link SINR results.","section":"§VI-B-3"},{"comment":"The array factor in Eq. (1) and the gain in Eq. (6) assume ideal coherent combining, and the text states that the synchronization protocols in [37]–[40] and CSI quantization in [42] can achieve this. However, no residual phase-error, frequency-error, or position-error model is provided. Since the claimed interference mitigation relies on deep pattern nulls toward non-receiving BSs, the paper should include a sensitivity analysis (e.g., total SINR versus phase-error variance or UAV position error) or explicitly frame the claims as ideal-synchronization upper bounds.","section":"§III-A-1 and §V-C"},{"comment":"The after-CB SINR is the value of objective f2 that CNSGA-II directly optimizes, so the reported improvement is partly a statement about the optimizer's convergence rather than an independent validation of interference mitigation. A comparison against a theoretical beamforming benchmark (e.g., maximum array gain toward the target BS subject to null constraints) or against a fixed heuristic would allow the reader to judge how much of the gain comes from the CB concept itself versus the specific optimization procedure.","section":"§VI-B-3"}],"minor_comments":[{"comment":"The pseudocode caption spells the algorithm name as 'CNASGA-II'; this should be corrected to 'CNSGA-II'.","section":"Algorithm 1"},{"comment":"The placement of ω(θ,ϕ)^2 and η in the gain formula is unclear; please check the parenthesization and define all terms explicitly.","section":"Eq. (6)"},{"comment":"Several parameter values appear to have inconsistent units relative to their definitions (e.g., ρ is listed as 0.503 m^2, and d0 as 1.225 kg/m^3). These should be verified against the cited references.","section":"Table III"},{"comment":"The phrase 'cumulative across all interfered BSs' should state whether the accumulation is performed in the linear domain or in dB, and ideally a per-BS table should be provided.","section":"§VI-B-1 and §VI-B-3"},{"comment":"The statements that Pareto solutions are 'closer to the PF direction' are not quantitative; consider reporting established multi-objective metrics such as inverted generational distance or hypervolume.","section":"§VI-B-1 and §VI-B-2"},{"comment":"Ranking the population by f1 and then f2 and removing the worst τ1 and τ2 solutions may discard non-dominated solutions; the choice of τ1 and τ2 should be justified, and its effect on solution diversity should be discussed.","section":"Algorithm 4"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the scope of the journal and the underlying problem and algorithm are of interest. The headline interference-mitigation claim in the abstract and conclusion rests on Table IV, and the current dB-based arithmetic is invalid; this is a load-bearing issue that should be fixed by re-analyzing the raw simulation data in the linear domain and clearly defining the baseline. If the authors can provide corrected numbers and a reproducible baseline, the paper could become suitable for publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the paper's core quantitative claim is not supported by its own Table IV, because the reported \"4.95 times\" is just the ratio of two numbers in dB (857.25/173.13), which is meaningless. That said, the paper is not a throwaway: it extends the authors' WASA 2022 single-BS collaborative beamforming work to multiple BSs plus a transmission-order variable, and it does that cleanly. The MOOP formulation is sensible, the NP-hardness argument via reduction to TSP is standard but fine, and CNSGA-II is a reasonable chaotic variant of NSGA-II with hybrid crossover/mutation for the mixed discrete-continuous space. The comparison against MOPSO, MOMVO, MODA, and NSGA-II is the usual evolutionary-computation horse race, but the algorithm section is competently written.\n\nWhere it falls down is the verification of the headline interference-mitigation claim. Table IV lists total SINR \"before\" and \"after\" CB as 173.1319 dB and 857.2498 dB. The text then says this is a 4.95× increase. Dividing two dB values gives a unitless 4.95 but it has no physical meaning. If these are sums of per-BS SINRs in dB, summation across dB is itself invalid, because adding dB values corresponds to multiplying linear ratios, not summing them. Also, the \"before\" baseline is not defined anywhere: no formula for how the non-CB received power or array gain is computed. And since f2 is the total SINR objective the algorithm optimizes, the \"after\" value is essentially the optimizer's own objective, so this is not an independent validation; it is circular if presented as evidence of interference suppression. No error bars, no repeated runs, no sensitivity to k1, k2, sigma^2, and so on.\n\nThe synchronization assumption is cited to prior protocols, which is honest, but it remains an idealization; residual phase errors would raise sidelobes. That concern is real but secondary compared to the dB arithmetic.\n\nBottom line: the problem formulation and algorithm design are worth reading by anyone working on UAV collaborative beamforming or multi-objective deployment. The paper needs major revision on the evaluation side before the 4.95× claim can be taken seriously. I would send it to peer review; a competent reviewer will catch the same arithmetic issue, and the authors should be asked to redo the evaluation in linear SINR, define the baseline, and add statistical runs.","headline":"The 4.95× interference-mitigation gain is an arithmetic artifact (ratio of two dB values), but the multi-BS problem formulation and the CNSGA-II algorithm are solid enough to send to peer review.","tokens_in":24849,"tokens_out":1741,"would_cite":false,"duration_ms":16948,"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":"A UAV swarm that beamforms as a virtual antenna array can cut uplink interference to non-target base stations by roughly a factor of five.","keywords":["UAV network","interference mitigation","collaborative beamforming","virtual antenna array","multi-objective optimization","evolutionary computation","CNSGA-II","uplink transmission"],"falsifier":"A deployment or high-fidelity simulation that sweeps phase-synchronization error and position jitter across the swarm, and measures the SINR at a non-target base station, would settle it: if the cumulative SINR improvement over the no-collaborative-beamforming baseline falls well below the reported 4.95x under realistic error, the central claim collapses.","tokens_in":23730,"feed_emoji":"📡","tokens_out":6326,"duration_ms":58179,"temperature":0.7,"pith_summary":"This paper argues that a swarm of UAVs collecting data can keep the interference it causes to non-target base stations low by acting as a virtual antenna array: the drones jointly set their hover positions and excitation currents so the beam points at whichever base station is receiving data, while the transmission order to multiple base stations is optimized. The authors formulate this as a three-objective optimization problem (transmission time, total SINR of interfered base stations, and propulsion energy), prove that the problem is NP-hard, and solve it with a chaotic evolutionary algorithm called CNSGA-II. Simulation results show the cumulative SINR of interfered base stations improving by 4.95 times with 8 UAVs and 4.23 times with 16 UAVs after collaborative beamforming is applied. If this holds, data-collection UAVs could serve several base stations for backup and distributed processing without heavily degrading other ground communications.","feed_headline":"UAV swarm beamforming cuts uplink interference 5 times","feed_subtitle":"Drones acting as one virtual antenna can serve multiple base stations while shielding others, lifting total SINR 4.95 times.","key_machinery":"The load-bearing object is the array factor $F(\\theta,\\phi)=\\sum_{i=1}^{N_U} I_i e^{j (2\\pi/\\lambda)(x_i\\sin\\theta\\cos\\phi + y_i\\sin\\theta\\sin\\phi + z_i\\cos\\theta)}$, which describes how the UAV swarm's signals combine in each direction; combined with the array gain formula it gives the power toward the target base station versus other base stations. Around this, the paper builds a multi-objective optimization problem (MOOP) whose objectives are total transmission time $f_1$, total SINR of interfered base stations $f_2$, and propulsion energy $f_3$, with a proof that the problem is NP-hard by reduction to the traveling salesman problem. The solver is CNSGA-II, a non-dominated sorting genetic algorithm modified with chaotic initialization, crossover, and mutation operators plus an elimination mechanism, which searches the mixed continuous-discrete solution space for Pareto-optimal trade-offs.","core_discovery":"The central discovery is that uplink interference from a UAV-enabled virtual antenna array to non-receiving base stations can be substantially reduced by jointly optimizing three things: the excitation current weights of each UAV, the three-dimensional hover positions, and the order in which the swarm transmits to multiple base stations. Quantitatively, the paper reports that in simulations the total SINR across interfered base stations rises by a factor of 4.95 in an 8-UAV network and 4.23 in a 16-UAV network compared with the same transmission without collaborative beamforming. The mechanism is that the virtual array's mainlobe is steered toward the intended base station while sidelobes toward other base stations are suppressed, and the multi-objective formulation lets the system trade interference mitigation against transmission time and propulsion energy.","pith_inferences":["If the reported gains are sensitive to synchronization error, then the 4.95x figure is an upper bound for real deployments; a plausible testable extension is to add phase-noise and position-jitter terms to the array factor and re-measure the cumulative SINR gain.","The same joint position-weight optimization could be combined with movable antenna arrays or with power control to handle the residual sidelobe interference that imperfect CSI leaves behind.","The NP-hardness reduction suggests that any exact or near-exact solution must exploit the traveling-salesman structure of the transmission order, so heuristics that solve TSP subproblems efficiently may dominate generic evolutionary search in larger networks."],"forward_implications":["Interference from data-collection UAVs to terrestrial base stations can be reduced without cutting transmit power or extending flight paths, since beamforming replaces power control and trajectory avoidance.","A UAV swarm can transmit the same or different data to several base stations in sequence for backup and distributed processing while keeping the SINR of non-target base stations high enough for their own ground users.","The trade-off between interference mitigation and energy is explicit: the Pareto front lets a network operator choose how much propulsion energy to spend moving drones for better beam patterns.","The proposed CNSGA-II solver, if the comparison simulations are representative, finds these trade-offs more effectively than MOPSO, MOMVO, MODA, and standard NSGA-II, at the cost of higher computation."],"supporting_citations":[{"why":"Supplies the array factor model that defines how the UAV swarm's signals combine in each direction.","marker":"[36]"},{"why":"Provides the probabilistic LoS channel model and the 3D propulsion energy formulation used in the SINR and energy objectives.","marker":"[2]"},{"why":"Together with [2], supplies the probabilistic LoS air-to-ground channel model used for link gains.","marker":"[43]"},{"why":"Supplies the rotary-wing UAV propulsion power and energy equations behind the energy objective.","marker":"[47]"},{"why":"Provides the codebook-based CSI quantization that the paper assumes gives near-perfect channel feedback for beamforming weights.","marker":"[42]"},{"why":"Experimental demonstration of distributed beamforming by a UAV swarm, used to justify that the swarm can build the virtual antenna array.","marker":"[38]"},{"why":"Timing, phase and frequency synchronization protocol for distributed wireless networks, assumed to hold for the UAV array.","marker":"[40]"},{"why":"Standard NSGA-II, the baseline algorithm that CNSGA-II modifies and compares against.","marker":"[66]"},{"why":"Earlier collaborative-beamforming UAV swarm data collection method that this paper extends from a single target base station to multiple base stations.","marker":"[15]"}],"fun_headline_variants":["UAV swarm virtual antenna cuts uplink interference 5x","Drones as virtual antenna array shield base stations from uplink interference","Joint UAV positioning and beamforming reduces base station interference","UAV swarm beamforming boosts SINR 4.95x for data collection","Optimized UAV virtual array cuts interference to non-receiving base stations"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole beam-pattern argument assumes the UAVs can share data and synchronize their clocks, phases, and frequencies accurately enough that the idealized virtual-antenna array factor actually forms, so if synchronization or channel feedback is imperfect, the deep nulls toward non-target base stations shallow and the reported SINR gain drops.","fun_headline_variants_meta":{"raw":{"variants":["UAV swarm virtual antenna cuts uplink interference 5x","Drones as virtual antenna array shield base stations from uplink interference","Joint UAV positioning and beamforming reduces base station interference","UAV swarm beamforming boosts SINR 4.95x for data collection","Optimized UAV virtual array cuts interference to non-receiving base stations"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001094,"raw_usage":{"total_tokens":4585,"prompt_tokens":980,"completion_tokens":3605,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":596,"completion_tokens_details":{"reasoning_tokens":3514}},"tokens_in":596,"tokens_out":3605,"duration_ms":27363,"temperature":1.0,"reasoning_tokens":3514,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T19:37:30.982899+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A deployment or high-fidelity simulation that sweeps phase-synchronization error and position jitter across the swarm, and measures the SINR at a non-target base station, would settle it: if the cumulative SINR improvement over the no-collaborative-beamforming baseline falls well below the reported 4.95x under realistic error, the central claim collapses.","supporting_citations":[{"cited_title":"Energy efficient collaborative beamforming for reducing side- lobe in wireless sensor networks,","cited_arxiv_id":null,"evidence_quote":"Supplies the array factor model that defines how the UAV swarm's signals combine in each direction."},{"cited_title":"Energy minimization for wireless communication with rotary-wing UA V,","cited_arxiv_id":null,"evidence_quote":"Supplies the rotary-wing UAV propulsion power and energy equations behind the energy objective."},{"cited_title":"Outage probability and ergodic capacity of distributed transmit beam- forming with imperfect CSI,","cited_arxiv_id":null,"evidence_quote":"Provides the codebook-based CSI quantization that the paper assumes gives near-perfect channel feedback for beamforming weights."},{"cited_title":"AirBeam: Experimental demonstration of distributed beamforming by a swarm of UA Vs,","cited_arxiv_id":null,"evidence_quote":"Experimental demonstration of distributed beamforming by a UAV swarm, used to justify that the swarm can build the virtual antenna array."},{"cited_title":"RFClock: Timing, phase and frequency synchronization for distributed wireless networks,","cited_arxiv_id":null,"evidence_quote":"Timing, phase and frequency synchronization protocol for distributed wireless networks, assumed to hold for the UAV array."}],"review_version":1}