{"id":"52f99a2c-e32c-4ec2-9c40-3198dca7d356","arxiv_id":"2506.07052","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A near-field ISAC beamforming framework that maximizes worst-case target beampattern gain while suppressing inter-target cross-correlation and guaranteeing per-user rates, enabling same-direction, different-range separation.","lead":"This paper designs a beamforming method for a base station that simultaneously communicates with multiple users and detects multiple targets in the near-field, using angle and distance information to separate objects along the same direction. It is a step toward 6G systems that combine communication and radar-like sensing in one transmission.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Equation (19) is not a valid Capon power spectrum as printed: un-squared numerator and R_X^{-1} in denominator make the Fig. 5 sensing evidence non-reproducible.","rationale":"The paper's convex relaxation and rank-one reconstruction are internally sound: preserving the total covariance R_x exactly preserves the beampattern and cross-correlation constraints, and the inequality with the reconstructed beamformers follows from Cauchy-Schwarz for a positive semidefinite matrix. So the beamforming half of the paper is not where the load-bearing risk sits. The risk is in the sensing evaluation: Section IV-C's Capon spectrum is the primary evidence for the headline 'multi-target indication', and Eq. (19) as printed is internally inconsistent. A Capon power spectrum must be a real nonnegative quantity, but the printed formula has no squared modulus and uses an inverse transmit covariance where the cited MIMO-Capon estimator uses the transmit covariance itself. If taken literally, the dB plots in Fig. 5 are undefined and the spectrum can be dominated by directions with small transmitted power. This is a concrete, testable issue. The reader's concern about exact position and channel knowledge is also valid, but it is a scope caveat about applicability; the Capon issue is an internal inconsistency in the evidence presented for the central claim. The recommended verdict stays CONDITIONAL because a straightforward recomputation can settle whether the peaks survive; if they do, only a correction to Eq. (19) is needed, and if they do not, the sensing-resolution claim is unsupported.","tokens_in":958,"tokens_out":2903,"duration_ms":364314,"concrete_test":"Recompute the sensing evaluation using the standard MIMO-Capon estimator from [13]: in Eq. (19), replace the printed denominator term with tilde h_{t,0}^H R_X tilde h_{t,0}^* and add a squared modulus around the numerator. Concretely, use |tilde h_{0,r}^H R_Y^{-1} Y X^H tilde h_{t,0}^*|^2 divided by (tilde h_{0,r}^H R_Y^{-1} tilde h_{0,r}) times (tilde h_{t,0}^H R_X tilde h_{t,0}^*). Regenerate Fig. 5 at the same parameter settings. If Targets 1-3 remain distinct peaks with similar contrast, then Eq. (19) is a typographical error and the sensing claim survives; if peaks merge, shift, or disappear, the multi-target resolution evidence is not reproducible as printed.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that the proposed beamformer resolves same-direction, different-range targets is supported in Section IV-C by Capon spectra. Equation (19) as printed is not a valid Capon power spectrum: the numerator is not squared, so the quantity is generally complex and cannot be plotted in dB; the last denominator factor is written with a stray subscript l and uses R_X^{-1}, whereas the MIMO-Capon estimator from the cited [13] normalizes by b^H R_X b, not by its inverse. Using R_X^{-1} would emphasize directions or ranges where the transmit covariance is small, potentially creating or destroying spectral peaks through transmit-power artifacts rather than through target presence. Since Fig. 5 is the main evidence for the sensing half of the abstract's claim, this is an internal inconsistency in the evidence chain. The SDR and rank-one reconstruction themselves appear sound: the reconstruction preserves R_x exactly, so transmit beampattern, cross-correlation, and rate constraints are unchanged.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper proposes a near-field integrated sensing and communication (ISAC) transmit beamforming design for a multi-user, multi-target scenario. The transmit covariance is optimized to maximize the minimum weighted sensing beampattern gain subject to inter-target cross-correlation suppression, per-user rate guarantees, and a total power constraint. The resulting non-convex problem is relaxed via semidefinite relaxation, and a closed-form rank-one reconstruction is provided following a known theorem. Numerical experiments indicate that the near-field design can separate users and targets lying along the same direction but at different ranges, which far-field beamforming cannot do.","tokens_in":8478,"tokens_out":11840,"duration_ms":130442,"significance":"The SDR relaxation and the rank-one reconstruction appear correct: the reconstruction preserves the transmit covariance and each user's SINR exactly, so all constraints and the objective are unchanged, and the design does not require radar cross-section knowledge. The extension of beampattern and cross-correlation metrics to the joint angle-range space is a natural and practically valuable generalization, and the SINR maps in Figs. 2-3 provide direct evidence of same-direction, different-range separation. The numerical claims are plausible, but the sensing evidence in Fig. 5 rests on Eq. (19), which is not a valid power spectrum as printed, and the far-field benchmark is weakened in a way that should be disclosed more prominently. The central idea is defensible and worth publishing after these issues are fixed.","major_comments":[{"comment":"As printed, the 'normalized Capon spectrum' is not a power spectrum: the numerator is not squared, so the expression is generally complex-valued and cannot be plotted in dB, and the last denominator factor contains an undefined subscript l and uses R_X^{-1}. Since Fig. 5 is the primary evidence for the sensing half of the central claim, the authors must correct Eq. (19), state whether the plotted quantity is |beta|, |beta|^2, or 20*log10(|beta|), and verify that the peaks in Fig. 5 are not created or destroyed by R_X^{-1} emphasizing directions of small transmit covariance.","section":"Section IV-C, Eq. (19)"},{"comment":"The far-field beamforming (FFBF) benchmark is not solving the same optimization problem as the proposed method: the minimum rate is lowered to 0.95 bps/Hz and the cross-correlation constraint is removed for target pair (1,2). Consequently, the claimed 'significant gains over far-field' comparisons in Figs. 4 and 5c mix the near-field geometric capability with relaxed constraints and a lower rate requirement. The authors should either provide a feasible FFBF baseline under the same constraints when possible, or explicitly frame Fig. 4 as a demonstration of near-field advantage under far-field infeasibility and temper the abstract and conclusion wording accordingly.","section":"Section IV-A"},{"comment":"The relaxed optimization, as written, does not explicitly include F_k ⪰ 0 for all k nor R_x ⪰ 0. These constraints are needed for the problem to be the standard semidefinite relaxation of (15) and for the rank-one reconstruction in (18) to be valid, since the proof relies on [7, Theorem 1] for positive semidefinite matrices. Without these constraints, (17) is not a valid SDR and the convexity statement is incomplete. Please add the missing positive semidefiniteness constraints to the problem statement.","section":"Section III, problem (17)"}],"minor_comments":[{"comment":"The second BS array center is labeled 'pt = (0,0.06,0) m'; this should be the receiver center, presumably 'pr'.","section":"Section IV-A"},{"comment":"The text says 'remove the cross-correlation constraint (15d) for the target pair (1, 2)', but (15d) is the rate constraint; the cross-correlation constraint is (15c).","section":"Section IV-A"},{"comment":"The abstract refers to 'significant gains over far-field and single-target benchmarks', but the simulations compare only the proposed method with NCCS and FFBF; no single-target benchmark appears in Section IV. Please add the intended single-target baseline or adjust the wording.","section":"Abstract and Section IV"},{"comment":"In the text following Eq. (3), 'beta(pk - pt(i, l))' contains a typo; the argument should be 'pk - pt(nt)'.","section":"Section II-C, Eq. (3)"},{"comment":"The denominator factor in Eq. (19) uses 'tilde h_{l,0}' where l is not defined in this context; if this is meant to be the transmit steering vector, it should read 'tilde h_{t,0}'.","section":"Section IV-C, Eq. (19)"}],"recommendation":"major_revision","confidential_remarks":"The paper is technically competent and the central SDR/rank-one argument is sound. The main risk is the sensing evidence: Eq. (19) as printed is not a valid power spectrum, and the FFBF benchmark is weakened in ways that could be seen as unfair. If the authors correct Eq. (19), clarify the plotted quantity, and suitably qualify the far-field comparison, the contribution could be publishable. The reliance on [7, Theorem 1] is acceptable because the paper sketches the proof and the theorem is a general linear algebra result."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nQuick take: this is a solid engineering paper, not a breakthrough. The genuine contribution is the angle-range version of the MIMO beampattern and cross-correlation metrics, which lets a transmit covariance design separate targets that share an angle but differ in range. The SDR relaxation and the rank-one reconstruction are correct and follow [7] cleanly. The communication SINR plots are convincing, and the proposed method clearly outperforms the no-cross-correlation variant.\n\nThe soft spots are real, and one is more serious than the paper suggests. Equation (19) as printed is not a valid Capon spectrum. The numerator is not magnitude-squared, so it is complex and cannot be plotted in dB; the last denominator factor has a stray subscript l and uses R_X^{-1}, which is the wrong normalization. That formula is the basis for Fig. 5, the only sensing evidence for the central claim. It may be a typo, but as written it makes the sensing results non-reproducible. This needs to be fixed before I would trust the multi-target resolution claim.\n\nThe benchmark comparisons are also a bit rigged. The FFBF baseline lowers Rmin to 0.95 bps/Hz and drops the cross-correlation constraint for the same-direction pair, so of course it fails. A fairer comparison would include the closest prior work [3] and maybe a single-target near-field design. The abstract promises \"single-target benchmarks\" that never appear in the simulations. That is a minor overstatement, but it should be cleaned up.\n\nThe math behind the optimization is sound, and the RCS-agnostic formulation is a nice practical feature. The paper does not cite anything it misrepresents; the self-citation to [7] is appropriate because the rank-one lemma is genuinely reused. My main advice is to fix the Capon formula, add code or at least a detailed simulation setup, and make the baselines less artificial.\n\nFor you: if you work on near-field ISAC, this is worth reading after a revision, but not urgent. I would send it to a serious referee rather than desk reject; the core idea is correct and the flaws are fixable.","headline":"A clean SDR-based extension of MIMO radar covariance design to near-field ISAC with a genuine angle-range metric; the core math holds, but the sensing evidence rests on a misprinted Capon formula and the baselines are handicapped.","tokens_in":9039,"tokens_out":3196,"would_cite":false,"duration_ms":32112,"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":"This paper claims that a near-field ISAC beamformer can resolve and serve users and targets along the same direction but at different distances by optimizing the transmit covariance to maximize worst-case sensing beampattern gain under…","keywords":["near-field ISAC","beamforming","multi-target sensing","cross-correlation suppression","semidefinite relaxation","spherical wavefront","Capon spectrum","range resolution"],"falsifier":"Run the optimization (17) with channels generated by a different physically plausible near-field model—for example a full-wave simulation of the same array—and apply the resulting beamformers to both models; if the Capon spectrum no longer shows distinct peaks at true target positions for same-direction targets, or if user SINRs fall below the required minimum, the central claim is refuted.","tokens_in":8098,"feed_emoji":"📡","tokens_out":7931,"duration_ms":78973,"temperature":0.7,"pith_summary":"The paper aims to establish that a single base station, operating in the near field with a large antenna array, can simultaneously talk to several users and detect several targets even when those users and targets share the same angle from the base station but sit at different distances. It proposes a beamforming design that chooses the transmit covariance to maximize the weakest weighted beampattern gain across all targets, while keeping inter-target cross-correlation below a fixed fraction of that gain and guaranteeing each user a minimum rate. Because the near-field channel depends on both angle and range, the design can place beampattern peaks and nulls at particular distances, not just directions. The payoff, if the claim holds, is that a single transmitted signal serves communication and sensing at the same time without knowing any target's radar cross section, and multi-target localization works where far-field designs merge same-direction targets into one peak.","feed_headline":"Near-field ISAC separates same-direction targets by range","feed_subtitle":"A single covariance optimization maximizes sensing gain, suppresses target cross-talk, and keeps per-user rates.","key_machinery":"The load-bearing object is the transmit covariance matrix $R_x$, shaped by a max-min semidefinite program. The paper extends the standard sensing metrics—beampattern gain $h_{t,l}^T R_x h_{t,l}^*$ and cross-correlation magnitude $|h_{t,l}^T R_x h_{t,l'}^*|$—from far-field angle-only steering vectors to near-field steering vectors whose entries are $[h_{t,a}]_{n_t}=\\sqrt{F(p_a-p_t(n_t),w_t)}\\,\\beta(p_a-p_t(n_t))$, so each entry carries a distance-dependent amplitude and phase. The optimization in (15) maximizes the worst-case weighted gain $\\mu$, caps cross-correlations at $\\varepsilon\\mu$, enforces rate constraints via the SINR-to-covariance transformation in (16), and is relaxed to the convex program (17) by replacing $f_kf_k^H$ with a positive semidefinite matrix $F_k$. A closed-form reconstruction, $\\tilde f_k=(h_{t,k}^T \\hat F_k h_{t,k}^*)^{-1/2}\\hat F_k h_{t,k}^*$, recovers rank-one beamformers that preserve the objective and all constraints.","core_discovery":"In the paper's own terms, the discovery is that classical far-field MIMO beampattern and cross-correlation criteria can be generalized to the near field by letting the steering vectors depend on range as well as angle, and that a transmit covariance optimized under these generalized criteria simultaneously resolves and serves users and targets aligned along the same direction but at different distances. The optimization maximizes the minimum weighted beampattern gain $w_l\\,h_{t,l}^T R_x h_{t,l}^*$ subject to cross-correlation bounds $w_{l,l'}|h_{t,l}^T R_x h_{t,l'}^*| \\le \\varepsilon \\mu$, per-user rate constraints, and a total power budget. The resulting non-convex problem is relaxed by rank-one lifting into a semidefinite program, and a closed-form reconstruction recovers rank-one beamformers that achieve the same objective. Simulations show the near-field design produces Capon spectra with distinct peaks at all three true target locations, whereas a version without cross-correlation suppression yields a cluttered spectrum and a far-field version merges same-direction targets.","pith_inferences":["If the claim holds, the same angle-range discrimination could be used for single-base-station 3D positioning of users and reflectors, because range becomes an observable degree of freedom rather than a nuisance parameter.","A testable extension is resolution scaling: the minimum range separation needed to resolve two same-direction targets should shrink as the array aperture grows relative to the Rayleigh distance, so the approach predicts a concrete aperture-versus-resolution tradeoff that could be measured.","The dependence on exact channel knowledge suggests a natural stress test: feeding the optimizer positions with a small offset, or replacing the idealized radiation profile with measured element patterns, should reveal how much of the claimed separation is robust to model mismatch."],"forward_implications":["Users or targets at the same angle but different ranges can be separated and served simultaneously, something far-field planar-wavefront beamforming cannot do.","Multi-target sensing no longer needs radar cross-section knowledge: only the array steering functions are required for the beampattern and cross-correlation constraints.","The semidefinite relaxation with rank-one reconstruction yields an implementable beamformer that attains the optimum of the relaxed problem without violating per-user rate or power constraints.","In the simulated 30 GHz configuration, near-field designs give user SINR peaks above 40 dB while the far-field benchmark stays below 0 dB, and the Capon spectrum shows three distinct target peaks only when cross-correlation suppression is included."],"supporting_citations":[{"why":"supplies the near-field ISAC system model, the Rayleigh distance defining the near-field region, and the rate-constraint reformulation used in (16).","marker":"[2]"},{"why":"provides the multi-target near-field ISAC SINR baseline that the paper argues needs RCS knowledge and lacks cross-correlation suppression.","marker":"[3]"},{"why":"furnishes the joint transmit beamforming framework and the rank-one reconstruction theorem used to recover closed-form beamformers from the relaxed solution.","marker":"[7]"},{"why":"defines the MIMO radar beampattern and cross-correlation probing criteria that the paper generalizes to the near field.","marker":"[9]"},{"why":"supplies the spherical-wave near-field channel model for downlink communication used in equation (3).","marker":"[11]"},{"why":"supplies the radiation-profile function $F(p,w)$ used in the near-field channel model.","marker":"[12]"},{"why":"provides the Capon spectrum estimator used to evaluate multi-target indication in the simulations.","marker":"[13]"}],"fun_headline_variants":["Near-field ISAC separates targets by range, not angle","Range-aware beampatterns resolve same-direction targets","Near-field beamforming distinguishes targets at different distances","Same direction, different range: near-field ISAC resolves both","Near-field ISAC spots each target despite alignment"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The design assumes the base station knows the exact position of every user and target, and that the spherical-wave channel model—radiation profile, path attenuation, and phase from equations (3)–(5)—describes the physical propagation precisely; if positions are uncertain or the model is inaccurate, the promised same-direction, different-range separation and rate guarantees may fail.","fun_headline_variants_meta":{"raw":{"variants":["Near-field ISAC separates targets by range, not angle","Range-aware beampatterns resolve same-direction targets","Near-field beamforming distinguishes targets at different distances","Same direction, different range: near-field ISAC resolves both","Near-field ISAC spots each target despite alignment"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000179,"raw_usage":{"total_tokens":1320,"prompt_tokens":988,"completion_tokens":332,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":604,"completion_tokens_details":{"reasoning_tokens":255}},"tokens_in":604,"tokens_out":332,"duration_ms":3313,"temperature":1.0,"reasoning_tokens":255,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T05:43:07.021058+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the optimization (17) with channels generated by a different physically plausible near-field model—for example a full-wave simulation of the same array—and apply the resulting beamformers to both models; if the Capon spectrum no longer shows distinct peaks at true target positions for same-direction targets, or if user SINRs fall below the required minimum, the central claim is refuted.","supporting_citations":[{"cited_title":"Near-field integrated sensing and communications,","cited_arxiv_id":null,"evidence_quote":"supplies the near-field ISAC system model, the Rayleigh distance defining the near-field region, and the rate-constraint reformulation used in (16)."},{"cited_title":"Near- field ISAC: Beamforming for multi-target detection,","cited_arxiv_id":null,"evidence_quote":"provides the multi-target near-field ISAC SINR baseline that the paper argues needs RCS knowledge and lacks cross-correlation suppression."},{"cited_title":"Beam focusing for near-field multiuser MIMO communications,","cited_arxiv_id":null,"evidence_quote":"supplies the spherical-wave near-field channel model for downlink communication used in equation (3)."},{"cited_title":"Path loss in reconfigurable intelligent surface-enabled channels,","cited_arxiv_id":null,"evidence_quote":"supplies the radiation-profile function $F(p,w)$ used in the near-field channel model."},{"cited_title":"Target detection and parameter estimation for MIMO radar systems,","cited_arxiv_id":null,"evidence_quote":"provides the Capon spectrum estimator used to evaluate multi-target indication in the simulations."}],"review_version":1}