{"id":"b91135f7-2c67-4b17-9471-7035846baa71","arxiv_id":"2505.15279","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A robust near-field ISCAP beamforming design that maximizes worst-case secrecy rate while guaranteeing sensing accuracy and energy harvesting under location uncertainty.","lead":"This paper designs a beamforming scheme for a base station that simultaneously sends confidential messages, wirelessly charges devices, and senses a target, while using the same signals to jam eavesdroppers. It shows that near-field antenna arrays can focus energy sharply enough to keep messages secret even when an eavesdropper sits on the same line as the legitimate user but at a different distance.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The worst-case guarantees hinge on φ_k, but Proposition 1 bounds only an approximation of the LoS channel error, and (52) vs (54) are inconsistent, so φ_k may underbound the true error.","rationale":"The reader's conditional verdict is supported by our stress-test. The single most load-bearing premise is that φ_k in (54) upper-bounds the true geometric LoS channel error for every ∆l_k with ∥∆l_k∥≤ε_k. This premise is load-bearing because the S-procedure constraints (59) and (61), the worst-case energy constraints (60)-(61), and hence the claimed robustness all live on the ball ∥∆g_k∥≤φ_k+δ_k. If φ_k underbounds the true error, the design is not worst-case robust. Our reading of the manuscript strengthens, rather than weakens, the reader's concern: Proposition 1's proof bounds only the Taylor/cosine-approximated expression (51), not the original Υ_k(∆l_k); the passage from (48) to (50) drops a denominator term whose sign is not controlled; and the displayed equations are internally inconsistent, with (41) omitting the +1/||ˆl_k−u_n||² term that appears in (40), and (52) using −2Σ1/||·||² while (54) uses −Σ1/||·||². These inconsistencies make it impossible to certify that the φ_k computed in Algorithm 1 is a valid bound. No amount of simulation with 10,000 Monte Carlo draws settles a worst-case statement; a counterexample in the continuous ball would. We also note the reversed inequality in (68) of Appendix B, but that appears to be a typo that can be fixed by flipping the inequality, so it is not the primary blocker. Because the identified gap is concrete and addressable—deriving a rigorous bound or numerically optimizing over the continuous uncertainty ball—the appropriate verdict remains CONDITIONAL, and our stress-test does not change the reader's verdict.","tokens_in":21378,"tokens_out":15983,"duration_ms":123578,"concrete_test":"Take the Section IV setup (N=64, λ=0.017 m, d=0.85 cm, ER/target positions) and for each uncertainty ball ∥∆l∥≤ε_k compute the exact maximum of ∥v(ˆl_k+∆l)⊙b(ˆl_k+∆l)−v(ˆl_k)⊙b(ˆl_k)∥ by dense grid search (≥10^6 points) or global optimization, then compare against φ_k from (54). A violation for any k invalidates the robust design. Also check consistency: at ε_k=0, φ_k must be 0; evaluate the true Υ at ∆l=ε_k u_1 (u_1 the eigenvector of Q_k for λ_max) and compare with (4π/λ)²φ_k².","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's robust guarantees are enforced via S-procedure constraints (59) and (61) on the ball ∥∆g_k∥ ≤ φ_k + δ_k, where φ_k in (54) is supposed to bound the geometric LoS error ∥v(lk)⊙b(lk)−v(ˆlk)⊙b(ˆlk)∥. This bound is not established. Proposition 1 explicitly bounds only 'the approximation in (51)', which is obtained from (i) a first-order Taylor expansion of distances (47), (ii) cos x ≈ 1−x²/2, and (iii) dropping the linear term in the denominator when passing from (48) to (50). None of these steps are inequalities with controlled error, so the approximated expression can be smaller than the true Υ_k(∆l_k). The manuscript also contains internal inconsistencies: (40) expands Υ_k as Σ[1/||l_k−u_n||² + 1/||ˆl_k−u_n||² − 2cos(...)/(...)], while (41) defines Π_k as only Σ1/||l_k−u_n||²; and the Ω bound (52) has −2Σ1/||ˆl_k−u_n||² whereas (54) uses −Σ1/||ˆl_k−u_n||². Consequently, φ_k used in the algorithm is not rigorously derived and may underestimate the true worst-case channel error. If so, the S-procedure constraints do not cover the true uncertainty region, and the claimed worst-case secrecy rate and energy harvesting guarantees in (P1) are not proven. Empirical validation (Figs. 2–3) samples only 10,000 Monte Carlo points, which cannot certify a worst-case bound over a continuous ball.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies a near-field integrated sensing, communication, and powering (ISCAP) system in which an extremely large-scale antenna array (ELAA) base station transmits a confidential information beam w0 together with a dedicated Gaussian signal whose optimizable covariance R1 is recycled as an energy, sensing, and artificial-noise (AN) signal. The energy receivers and the sensing target are treated as potential eavesdroppers whose channels are known only through coarse location estimates with error bound ε_k and bounded NLoS components δ_k. The paper formulates the robust joint beamforming problem (P1), maximizing the worst-case secrecy rate subject to Cramér-Rao bound (CRB) sensing accuracy constraints over a target uncertainty region Θ, worst-case energy-harvesting constraints, a transmit power budget, and a rank-1 constraint on the information covariance. The solution pipeline applies semidefinite relaxation, the Charnes-Cooper transformation, a Taylor-approximation-based bound φ_k on the geometric channel error, and S-procedure conversions of the worst-case SINR and energy constraints into LMIs, followed by a one-dimensional search over the eavesdropper SINR threshold. Numerical results compare the proposed design with separate, MRT, and ZF benchmarks and include a near-field MUSIC localization experiment.","tokens_in":21729,"tokens_out":39254,"duration_ms":304719,"significance":"The problem is timely and well motivated: near-field distance-domain focusing is a promising physical-layer-security mechanism for ISCAP, and the triple-purpose covariance design is a sensible way of integrating the three functionalities. The S-procedure reformulations in (59) and (61) are correct for the ball uncertainty model, the semidefinite relaxation and Charnes-Cooper machinery is applied coherently, the rank-one reconstruction in Proposition 3 is essentially correct, and the numerical study is informative, with three meaningful benchmarks and a separate localization experiment. The load-bearing weakness is the geometric error bound φ_k in (54): Proposition 1 bounds only the approximate expression in (51), the approximation steps in (47)–(50) are uncontrolled, and the equations (42), (52), and (54) are mutually inconsistent. Because the S-procedure constraints (59) and (61) are enforced on the ball of radius φ_k + δ_k, the claimed worst-case secrecy, energy, and sensing guarantees all depend on this unproven bound. If the bound is repaired, or the claims are honestly weakened to heuristic ones, this would be a solid contribution; as it stands the significance is conditional.","major_comments":[{"comment":"The load-bearing geometric error bound φ_k in (54) is not established. Proposition 1 bounds only 'the approximation in (51)', not the true LoS error Υ_k(Δl_k) in (39)–(40). The three steps used to pass from (46) to (51) — the first-order Taylor expansion (47), the cosine approximation cos x ≈ 1 − x²/2 in (50), and the replacement of the denominator ∥l̂_k − u_n∥ + q_{k,n}^T Δl_k by ∥l̂_k − u_n∥ — are not inequalities with controlled error, so the approximated expression can lie below the true Υ_k. The phase argument in (50) is bounded in magnitude only by 4πε_k/λ, about 74 rad for ε_k = 0.1 m and λ = 0.017 m (Section IV), far outside the small-angle regime, so uniform validity of (50) over Ψ_k is not assured. The derivation is also internally inconsistent: (42) defines Ω_k without the term Σ_n 1/∥l̂_k − u_n∥² that (40) contains; Appendix A's (64) and the claimed bound (52) equal twice the approximation (51) (both the quadratic coefficient 4π²/λ² and the constant −2Σ_n 1/∥l̂_k − u_n∥² are double the corresponding terms in (51)); and (54) uses −Σ_n 1/∥l̂_k − u_n∥², matching neither (51) nor (52). The manuscript itself describes the task as 'approximating the upper bound' just before (43), in tension with the conclusion's claim that the solution 'guarantees robustness'. Because the S-procedure constraints (59) and (61) are enforced on the ball ∥Δg_k∥ ≤ φ_k + δ_k, an underestimation of φ_k means the certified constraints do not cover the true uncertainty region Ψ_k, and the claimed worst-case secrecy and energy-harvesting guarantees in (P1) do not follow; the 10,000-sample Monte Carlo in Figs. 2–3 cannot certify a worst case over a continuous ball. The authors should either prove a conservative bound on the true Υ_k directly (e.g., bounding each term of (40) using |cos| ≤ 1 and the triangle inequality) or explicitly present φ_k as a heuristic and validate it accordingly.","section":"§III-B, Eqs. (39)–(54), Appendix A"},{"comment":"The sensing guarantee is stated in (P1) as CRB constraints (29a)–(29b) for all (θ_s, r_s) ∈ Θ, but the algorithm enforces the CRB LMIs only at the M grid points of Θ̄ defined in (27). The text in §II-D says the discrete sampling is used 'to approximate and guarantee' the worst-case performance, yet no discretization-error control (e.g., Lipschitz continuity or monotonicity of the CRB over Θ) is provided, so the guarantee over the continuous region is not established. At minimum, the guarantee should be stated for the grid only, or an argument must show that enforcing (33)–(34) on Θ̄ implies (29a)–(29b) on Θ. The notation also conflates the two regions: (SDR1) and (31)–(32) quantify over 'Θ' while the grid is Θ̄.","section":"§II-D, Eqs. (26)–(29); §III-A, Eqs. (33)–(34)"}],"minor_comments":[{"comment":"The second interval is written as 'θ_s ∈ [r^L_s, r^U_s]'; it should be 'r_s ∈ [r^L_s, r^U_s]'.","section":"Eq. (26)"},{"comment":"The statement says the reconstruction 'satisfies rank(R*_1) = 1', but the construction gives rank(R*_0) ≤ 1, while R*_1 is generally full-rank; the statement should be corrected. In the proof, inequality (68) appears reversed: from (66) we have R⋆_0 − R*_0 ⪰ 0 and hence R*_1 − R⋆_1 ⪰ 0, giving γ_k(R⋆) ≥ γ_k(R*), which is the opposite of the printed sign; this is consistent with the surrounding text, which says the eavesdropper SINR is reduced, and the conclusion of the proposition is correct once the sign is fixed.","section":"Proposition 3, Eq. (62), Eq. (68)"},{"comment":"Describing the proposed bound as an 'exact performance guarantee' based on 10,000 Monte Carlo realizations overstates what the simulations can certify; suggest rewording to 'empirically tight over the sampled realizations'.","section":"§IV, discussion of Fig. 2"},{"comment":"The sensing target's location uncertainty is modeled in Cartesian coordinates as ∥Δl_{K+1}∥ ≤ ε_{K+1} and in polar coordinates as the region Θ with separate angle and range intervals; the mapping between these two descriptions is not stated, and the numerical choice ε_{K+1} = 0.02 m is never related to Θ. Please clarify.","section":"Eqs. (20) and (26)"},{"comment":"The complexity analysis lists '2 LMI constraints with size N', but the CRB LMIs (33)–(34) are 2×2 and the S-procedure LMIs (59) and (61) are (N+1)×(N+1); no N×N LMI appears in (P4), so the count should be corrected.","section":"§III-C"},{"comment":"Minor typos and wording: 'estiblished' (§III-B), 'senarios' (§II-B), the duplicated 'threshold threshold' in the caption of Fig. 8, and 'near-filed' in the caption of Fig. 9.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"To the editor: the manuscript fits the journal's scope and the numerical study is complete in its coverage of benchmarks, but the decisive issue is the geometric channel-error bound φ_k in (54); as documented in Major Comment 1, the proof in Proposition 1 establishes a bound only for an approximation, and the factor-of-two mismatches among (42), (52), and (54) should be explained by the authors. Part of the material appeared in the conference paper [1]; the authors should state clearly what is new beyond [1]. No code or data-availability statement is provided."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague, here's my read.\n\nThis paper is a solid engineering contribution to the ISCAP / near-field physical-layer security line. The new combination is real: ELAA, near-field distance-domain focusing, a triple-purpose AN/energy/sensing covariance matrix, and robust beamforming under location uncertainty. The problem formulation is clear, the benchmarks are sensible, and the numerical results make a plausible case that near-field focusing buys secrecy when an eavesdropper sits in the same angular direction. The SDR/Charnes-Cooper/S-procedure machinery is standard but applied competently. The rank-one reconstruction in Proposition 3 is a neat idea, and the idea of preserving the sum R0+R1 to keep sensing and energy constraints satisfied is sound.\n\nThe soft spots are in the robustness claims, and they are not cosmetic. The load-bearing quantity is φ_k, supposedly an upper bound on the geometric LoS channel error. Proposition 1 does not actually bound the true error; it bounds the Taylor/cosine approximation in (51). The steps from (47) to (50) are approximations, not inequalities with controlled error, so the approximated expression can be smaller than the true Υ_k(Δl_k). If φ_k under-estimates the error, the S-procedure ball in (59) and (61) does not cover the true uncertainty region, and the claimed worst-case secrecy and energy guarantees are not proven. On top of that, (52) and (54) are inconsistent: (52) has a -2Σ1/||...||² term per element, while (54) has only -Σ1/||...||². That is either a typo or a sign error, but as written the formula does not follow from Proposition 1. The Monte Carlo checks in Figs. 2-3 are reassuring but 10,000 samples cannot certify a continuous worst case.\n\nTwo smaller issues: the CRB constraints are enforced on a finite grid, and there's no bound on the discretization error. And in Appendix B, inequality (68) appears reversed; the claim that eavesdropper SINR is reduced may still be true, but the proof needs to be fixed.\n\nNet: the design idea is good and likely correct in spirit, but the advertised worst-case guarantees are not rigorously established. This is a revise-and-resubmit, not a reject: the authors need to either make the bound rigorous or explicitly reframe the paper as a heuristic robust design with simulation validation. I'd send it to a serious referee.","headline":"Novel near-field ISCAP beamforming design whose worst-case robustness rests on an unproven approximate error bound—fixable, but not as is.","tokens_in":22257,"tokens_out":4333,"would_cite":false,"duration_ms":36453,"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":"The paper proposes a robust joint beamforming scheme to maximize the worst-case secrecy rate in near-field ISCAP systems under bounded eavesdropper location uncertainty, with guaranteed sensing and energy harvesting.","keywords":["near-field communications","extremely large-scale antenna arrays (ELAA)","integrated sensing communication and powering (ISCAP)","physical layer security","artificial noise","robust beamforming","Cramér-Rao bound","location uncertainty"],"falsifier":"Fix an array, an estimated eavesdropper location $\\hat{l}_k$, and an uncertainty radius $\\varepsilon_k$; numerically maximize $\\|v(l)\\odot b(l)-v(\\hat{l}_k)\\odot b(\\hat{l}_k)\\|$ over all $l$ with $\\|l-\\hat{l}_k\\|\\le\\varepsilon_k$ and compare the maximum with $\\varphi_k$ from (54): if the true maximum exceeds $\\varphi_k$, the robust guarantees fail. Similarly, for the sensing side, evaluate the true CRB at a target location lying between the grid samples of $\\bar{\\Theta}$: if it exceeds $\\Gamma_\\theta$ or $\\Gamma_r$, the worst-case sensing claim is violated.","tokens_in":21152,"feed_emoji":"📡","tokens_out":6068,"duration_ms":45250,"temperature":0.7,"pith_summary":"This paper proposes a robust joint beamforming design for an integrated sensing, communication, and powering (ISCAP) system where a base station with an extremely large-scale antenna array serves one legitimate user, several energy-harvesting receivers, and a sensing target at the same time. Because the energy receivers and the sensing target may eavesdrop, the design splits the transmitted signal into an information beam and a dedicated beam that plays three roles at once: energy delivery, sensing illumination, and artificial noise. The paper's central claim is that maximizing the worst-case secrecy rate of the legitimate user is tractable even when only coarse locations of the potential eavesdroppers are known, and that this can be done while guaranteeing worst-case localization accuracy via Cramér-Rao bound constraints and worst-case harvested energy. If the claim holds, near-field distance-domain focusing becomes a practical physical-layer security mechanism that can protect communication even when an eavesdropper sits along the same angular direction as the legitimate user but at a different distance.","feed_headline":"Near-field beams shield ISCAP from eavesdroppers","feed_subtitle":"One beam doubles as energy, sensing, and jamming; robust design maximizes worst-case secrecy under location error.","key_machinery":"The argument rests on near-field steering vectors $v(\\theta, r)$ that depend on both angle and distance through spherical wavefronts, which lets beamforming focus energy in the distance domain rather than only in angle. On top of that, the paper builds a location-error-to-channel-error bound: the channel mismatch $\\Delta g_k$ is decomposed into a geometric line-of-sight part depending on $\\Delta l_k$ and an NLoS part, and Proposition 1 supplies an approximate upper bound $\\varphi_k$ from a first-order Taylor expansion of the distances and a second-order cosine approximation. This bound turns the worst-case eavesdropping SINR and harvested-energy constraints into constraints over a ball $\\|\\Delta g_k\\| \\le \\varphi_k + \\delta_k$, which the S-procedure converts into linear matrix inequalities. Semidefinite relaxation with the Charnes-Cooper transformation handles the fractional secrecy-rate objective, and Proposition 3 reconstructs a rank-one beam with equal secrecy performance.","core_discovery":"On its own terms, the paper establishes that in a near-field ISCAP system, the transmit covariance separating the information beam $R_0 = w_0 w_0^H$ from a general auxiliary covariance $R_1$ can be optimized so that the worst-case secrecy rate in (24) is maximized subject to worst-case angle and range Cramér-Rao bound constraints on sensing, worst-case harvested power constraints at each energy receiver, and a total power budget, under bounded location errors $\\|\\Delta l_k\\| \\le \\varepsilon_k$ and bounded non-line-of-sight components. The key structural result, Proposition 3, says that the semidefinite relaxation solution can be converted to a rank-one information beam with no loss of secrecy rate. Numerically, the proposed design outperforms separate, maximum-ratio-transmission, and zero-forcing benchmarks in achievable secrecy rate, and the beam patterns show a sharp energy focus at the legitimate user with a deliberate null at the user location for the artificial-noise beam.","pith_inferences":["A natural extension, not in the paper, is to replace the finite grid $\\bar{\\Theta}$ in the CRB constraints with the full continuous uncertainty region; the paper's S-procedure machinery could likely be adapted to yield genuine worst-case sensing guarantees across all $(r_s, \\theta_s)$, rather than only at sampled points.","The Taylor-based error bound $\\varphi_k$ is derived from approximations, so the true robustness guarantee could be checked offline by an exhaustive search over all $\\|\\Delta l_k\\| \\le \\varepsilon_k$; if such a check ever finds a larger true error, the S-procedure constraints would need a tighter bound.","The same geometric-versus-NLoS channel-error decomposition could be reused in other near-field robust designs, such as pure integrated sensing and communication beamforming or wireless power transfer, where location uncertainty also converts into line-of-sight channel mismatch."],"forward_implications":["A single auxiliary beam can serve as energy signal, sensing waveform, and artificial noise simultaneously, so secrecy protection does not consume extra power beyond what sensing and powering already use.","Because the reconstructed beam is rank one with no secrecy-rate loss, the robust design is directly implementable as one information beamforming vector plus a covariance beam.","Worst-case guarantees on sensing accuracy and harvested energy hold for every channel realization inside the modeled uncertainty region, not just for the nominal channel.","Eavesdroppers along the same angle as the legitimate user are still suppressed when they are at different distances, which far-field beamforming cannot do."],"supporting_citations":[{"why":"Supplies the worst-case robust secure transmission formulation in MISO channels that underlies the secrecy-rate objective in (24).","marker":"[30]"},{"why":"Provides the robust secure beamforming framework for wireless information and power transfer, including the S-procedure treatment of eavesdropper CSI errors that this paper adapts.","marker":"[18]"},{"why":"Delivers the robust secure integrated sensing and communication beamforming approach and the S-procedure machinery for bounded eavesdropper CSI errors.","marker":"[9]"},{"why":"Gives the near-field integrated sensing and communication model and the Cramér-Rao bound expressions in (18) and (19) used for the sensing constraints.","marker":"[25]"},{"why":"Motivates the Taylor-series approximation for bounding location-induced channel errors in the derivation of Proposition 1.","marker":"[31]"},{"why":"Establishes the near-field propagation model and Rayleigh distance that justify the spherical-wavefront steering vectors and distance-domain focusing.","marker":"[21]"},{"why":"Provides the near-field physical layer security context showing why distance-domain resolution can protect against eavesdroppers along the same angle.","marker":"[28]"}],"fun_headline_variants":["Triple-duty beams jam eavesdroppers in near-field ISCAP","Robust beamforming maximizes secrecy in near-field ISCAP","One beam powers, senses, and jams in near-field ISCAP","Worst-case secrecy rate maximized in near-field ISCAP","ELAA beams secure near-field ISCAP against location errors"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the approximate upper bound $\\varphi_k$ from Proposition 1 really is an upper bound on the geometric line-of-sight channel error for every location error within $\\|\\Delta l_k\\| \\le \\varepsilon_k$; if it ever underestimates the true error, the claimed worst-case secrecy, energy, and sensing guarantees are not assured.","fun_headline_variants_meta":{"raw":{"variants":["Triple-duty beams jam eavesdroppers in near-field ISCAP","Robust beamforming maximizes secrecy in near-field ISCAP","One beam powers, senses, and jams in near-field ISCAP","Worst-case secrecy rate maximized in near-field ISCAP","ELAA beams secure near-field ISCAP against location errors"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000523,"raw_usage":{"total_tokens":2557,"prompt_tokens":1005,"completion_tokens":1552,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":621,"completion_tokens_details":{"reasoning_tokens":1465}},"tokens_in":621,"tokens_out":1552,"duration_ms":11721,"temperature":1.0,"reasoning_tokens":1465,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T15:20:59.852099+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Fix an array, an estimated eavesdropper location $\\hat{l}_k$, and an uncertainty radius $\\varepsilon_k$; numerically maximize $\\|v(l)\\odot b(l)-v(\\hat{l}_k)\\odot b(\\hat{l}_k)\\|$ over all $l$ with $\\|l-\\hat{l}_k\\|\\le\\varepsilon_k$ and compare the maximum with $\\varphi_k$ from (54): if the true maximum exceeds $\\varphi_k$, the robust guarantees fail. Similarly, for the sensing side, evaluate the true CRB at a target location lying between the grid samples of $\\bar{\\Theta}$: if it exceeds $\\Gamma_\\theta$ or $\\Gamma_r$, the worst-case sensing claim is violated.","supporting_citations":[{"cited_title":"Robust secure transmission in MISO channels based on worst-case optimization,","cited_arxiv_id":null,"evidence_quote":"Supplies the worst-case robust secure transmission formulation in MISO channels that underlies the secrecy-rate objective in (24)."},{"cited_title":"Robust beamforming for secure communication in systems with wireless information and power transfer,","cited_arxiv_id":null,"evidence_quote":"Provides the robust secure beamforming framework for wireless information and power transfer, including the S-procedure treatment of eavesdropper CSI errors that this paper adapts."},{"cited_title":"Robust transmit beamforming for secure integrated sensing and communication,","cited_arxiv_id":null,"evidence_quote":"Delivers the robust secure integrated sensing and communication beamforming approach and the S-procedure machinery for bounded eavesdropper CSI errors."},{"cited_title":"Near-field integrated sensing and communication: Performance analysis and beamforming design,","cited_arxiv_id":null,"evidence_quote":"Gives the near-field integrated sensing and communication model and the Cramér-Rao bound expressions in (18) and (19) used for the sensing constraints."},{"cited_title":"Location information assisted beamforming design for reconfigurable intelligent surface aided com- munication systems,","cited_arxiv_id":null,"evidence_quote":"Motivates the Taylor-series approximation for bounding location-induced channel errors in the derivation of Proposition 1."},{"cited_title":"Near-field MIMO communications for 6G: Fundamentals, challenges, potentials, and future directions,","cited_arxiv_id":null,"evidence_quote":"Establishes the near-field propagation model and Rayleigh distance that justify the spherical-wavefront steering vectors and distance-domain focusing."},{"cited_title":"Physical layer security in near-field communications,","cited_arxiv_id":null,"evidence_quote":"Provides the near-field physical layer security context showing why distance-domain resolution can protect against eavesdroppers along the same angle."}],"review_version":1}