{"id":"f01e95a6-2b2e-4733-86f6-f4f1f5069ee1","arxiv_id":"2607.13470","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"A posterior-confidence and innovation-gated sensing policy embedded in an SDP beamformer cuts ISAC transmit power by relaxing sensing illumination during high-confidence epochs, at modest tracking cost.","lead":"An ISAC base station uses Kalman-filter confidence and innovation checks to decide when radar probing can be skipped, then solves a convex beamforming problem that keeps only a small illumination floor while sensing is off. The paper reports large transmit-power savings versus always-on and periodic probing schemes while holding communication SINR fixed.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The claimed sensing robustness rests on an underspecified link: Eq. (28c) imposes a beampattern floor ρα_q, but the EKF model uses fixed measurement noise and no detection-probability model, so the track-loss guarantee at ρ=0.3 is not established outside the specific simulation.","rationale":"The reader's weakest assumption identifies the same load-bearing condition: the nonzero illumination floor during skipped epochs must preserve enough observability for the EKF. The paper's optimization formulation enforces a beampattern floor, but the tracking model does not specify how that floor translates into measurement availability or noise. Since the EKF covariance update only consumes measurements, a floor that merely keeps beampattern gain above a threshold has no effect on the filter unless the measurement model depends on that gain. The paper does not provide such a model, so the track-loss results are simulation-specific and not reproducible from the stated equations. This is not an internal mathematical contradiction of the SDP, but it is an unvalidated premise for the central energy-vs-tracking claim. The SINR-preservation part is structural assuming the intended constraint is restored; the sensing-stability part is the weak link. The framework is coherent and the simulation trends are plausible, but the evidence is insufficient for unconditional acceptance. Since the reader already issued a CONDITIONAL verdict and this analysis reinforces that assessment rather than moving it, the verdict should remain UNCHANGED.","tokens_in":13186,"tokens_out":8361,"duration_ms":87371,"concrete_test":"Re-implement the simulator with an explicit coupling between illumination and measurement quality: let the detection probability be p_d = 1 - exp(-β (u_q + ρ(1-u_q)) α_q / σ_R^2) and/or scale the measurement-noise covariance inversely with the achieved beampattern gain. Then rerun the ASP comparison at ρ=0.3 under (a) target maneuvers with 10× process noise q_a and (b) SNR = 0 dB. If the track-loss probability P_loss(ρ=0.3) rises above an acceptable bound or approaches the PLP baseline, the floor is insufficient and the claim must be narrowed.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that ASP saves transmit power without excessive sensing degradation because the safety illumination floor ρα_q preserves observability during skipped epochs (Eq. (28c); ρ=0.3 in Sec. VI.A.3). For this to hold, the floor must actually keep the EKF track from diverging. But the EKF measurement model in Eq. (12) uses a fixed measurement-noise covariance R_z and assumes a measurement z_rad is available whenever NIS gating is passed; no detection-probability or SNR-dependent noise model connects the beampattern gain in Eq. (28c) to the presence or quality of that measurement. Thus the track-loss probability P_loss(ρ) in Eq. (23) is not analytically tied to ρ, and the simulated track-loss curves (Figs. 9, 11) cannot justify robustness outside the tested constant-velocity, fixed-noise, single-ρ setting. There is also an internal tension: Algorithm 1 step 6 says 'set sensing power to zero' for skipped targets, while Eq. (28c) requires a nonzero floor ρ α_q. The mechanism by which the floor prevents divergence is therefore underspecified.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes an energy-efficient beamforming framework for a MIMO dual-functional radar-communication (DFRC) base station. At each epoch, the base station minimizes total transmit power subject to per-user SINR constraints and target-sector beampattern illumination constraints. The novelty is a sensing activation policy driven by EKF-derived posterior confidence and normalized innovation squared (NIS): a target is fully illuminated only when confidence is low or the innovation is inconsistent; otherwise it receives a reduced safety illumination floor ρ α_q. The optimization is posed as a convex SDP, and numerical comparisons are made against full-probing and periodic low-power baselines. The reported results show lower transmit power and equal communication sum-rate for the proposed adaptive skip-aware probing (ASP), with track-loss probability between the baselines.","tokens_in":13503,"tokens_out":8493,"duration_ms":80888,"significance":"If the claims are substantiated, the paper would contribute a practical, tracking-aware mechanism for reducing sensing energy in ISAC systems, a timely topic for 6G green communications. The idea of coupling EKF statistics with transmit-covariance optimization is interesting and the convex formulation is a strength. The paper also provides a clear set of tracking-stability metrics and a comparison with two reasonable baselines. However, the significance is currently limited by several internal inconsistencies and missing modeling links. The main energy-saving result is, to a large degree, built into the relaxed illumination constraint, and the robustness claim for the safety floor is not backed by a detection-probability model or an analytic link between ρ and track-loss probability. These issues must be resolved before the contribution can be fully assessed.","major_comments":[{"comment":"The SINR constraint in the skip-aware SDP is printed incorrectly. It should be tr(H_i W_i) − τ_k(∑_{j≠i} tr(H_i W_j) + tr(H_i F)) ≥ τ_k σ_i^2, as in Eq. (26b). As printed, the constraint is τ_k(∑_{j≠i} tr(H_i W_j) + tr(H_i F)) ≥ τ_k σ_i^2, which omits the desired signal term and reduces to an inequality on interference only. Because Algorithm 1 solves Problem (28), the simulated power and sum-rate results in Section VI are not supported by the stated optimization. Please correct the equation and verify that the simulations use the correct constraint.","section":"Section V, Eq. (28b)"},{"comment":"Step 6 instructs to 'set sensing power to zero' for skipped targets, while Eq. (28c) imposes a nonzero safety floor ρ α_q on the total transmit covariance toward those targets. If F_l is set to zero, the floor must be satisfied by communication beams alone; if F_l is not set to zero, the statement is misleading. This contradiction obscures the mechanism by which the floor preserves observability during skipped epochs. Please clarify the role of F_l and precisely how the safety floor is realized.","section":"Algorithm 1, step 6 vs Eq. (28c)"},{"comment":"The track-loss probability P_loss(ρ) in Eq. (23) is defined as a function of ρ, but no model connects ρ to measurement availability or measurement quality. The EKF in Eq. (12) uses a fixed R_z and assumes a measurement is available whenever the NIS gate passes; there is no detection-probability or SNR-dependent noise term linking the beampattern floor ρ α_q to the probability of obtaining a useful measurement. Thus the claim that ρ=0.3 'guarantees' reliable tracking is only supported by the specific simulation scenario, not by an analytic or semi-analytic argument. Please provide a detection-probability model or a substantially broader sensitivity study (maneuvers, clutter, SNR, ρ) to establish this link.","section":"Section III.B.2 and Section VI.A.3"},{"comment":"The activation rule for epoch l+1 uses Λ_{q,l}, the NIS at epoch l. If epoch l was a skipped epoch, no radar measurement is produced under the model in Eq. (12), so Λ_{q,l} is undefined. The paper does not specify how the NIS is computed following a skipped epoch, nor whether the safety illumination floor yields a measurement. Without this specification, Algorithm 1 is not implementable as written, and the meaning of the NIS trigger is unclear. Clarify the measurement model during skipped epochs or modify the decision rule to depend only on confidence after a skip.","section":"Section V, Eq. (27)"},{"comment":"The reduction in normalized transmit power of ASP relative to FP is a direct consequence of the constraint relaxation: for skipped targets the illumination floor is lowered from α_q to ρ α_q, enlarging the feasible set of the SDP. Therefore the power saving in Fig. 2 is not an empirical discovery but is structurally guaranteed. The substantive claim is the track-loss performance at the selected ρ. To support the claim that the EKF/NIS policy, rather than the relaxation itself, is responsible for the good energy-tracking trade-off, compare ASP against a random or periodic skip policy with the same average sensing rate and the same safety floor.","section":"Section V, Eq. (28c) and Fig. 2"}],"minor_comments":[{"comment":"The text refers to 'SINR constraints (28b)' when discussing the baseline problem; the intended reference is (26b).","section":"Section IV"},{"comment":"Equation numbering is inconsistent: Eq. (3) appears after Eq. (5), and some equation numbers are referenced out of order. Please renumber and update cross-references.","section":"General"},{"comment":"The discussion of the NIS threshold α is ambiguous. A larger α relaxes the acceptance gate (so more measurements are used when taken), but also reduces the frequency of NIS-triggered activations. The net effect on skipping frequency is not monotonic unless the precise interplay between the gate and the activation rule is stated. Please clarify.","section":"Section VI.A.1"},{"comment":"Eq. (22) defines track loss using T_loss consecutive epochs, but the simulation value of T_loss is never specified. Please state the value used in the numerical results.","section":"Section III.B.2"},{"comment":"There are several typos, including 'alwyes' (Section VI.A.2), 'UA V' (references), 'Inspiring by' (Introduction), and 'And Q_w' (Section III.A). A thorough proofread is needed.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The paper has a promising direction, but the current text contains a critical error in the SINR constraint of the main SDP and an internal contradiction in Algorithm 1. The power-saving claim is weakened by being a direct consequence of the relaxation, and the robustness of the safety floor is not analytically connected to track-loss probability. These issues are fixable but require more than minor edits. I recommend major revision rather than rejection, provided the authors can correct the constraint, clarify the skip mechanism, and add the missing detection-probability link or a matched comparison."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague, the genuinely new piece here is embedding the sensing skip decision inside the transmit-covariance SDP, driven by EKF confidence and NIS gating. That's a real departure from the cited ISAC beamforming work, and the authors get credit for a clean convex formulation. The soft spots are in the strength of the claims: part of the power saving is structural, and the tracking-robustness evidence is thinner than the conclusions suggest.\n\nThe formulation itself is solid. The min-power SDP with per-user SINR and sector beampattern constraints is standard, and the skip-aware relaxation in (28c) is correctly written as a convex constraint: during skipped epochs the illumination floor drops from alpha_q to rho*alpha_q, which enlarges the feasible set, so the optimal power can only decrease. The comparison against full-probing and periodic low-power baselines is the right experimental design. Give credit for that.\n\nThe soft spots, in rough order of severity. First, the printed SINR constraint (28b) omits the desired user signal term; as written it requires interference to be at least tau_k*sigma_i^2, which cannot be the intended condition. Likely a typo from the baseline (26b), but it has to be fixed before anyone can reproduce the optimization. Second, the rank-one recovery remark is asserted rather than demonstrated for this specific objective and constraint set; I'd want the constructive argument from [32, Theorem 4] made explicit. Third, and most important, the headline result—ASP uses less power with acceptable track-loss—is partly structural, as noted. The substantive question is whether the EKF keeps tracking with the floor at rho=0.3, and that is only shown in simulation with fixed measurement noise and no detection-probability model linking the beampattern gain to the quality or existence of the radar measurement. P_loss(rho) is defined, but no analytic tie from rho to P_loss is provided; the sweep in Fig. 9 is one scenario with one target motion model. That's not a fatal flaw, but it should be stated more modestly. Minor: Algorithm 1 says \"set sensing power to zero\" for skipped targets, which clashes with the nonzero floor rho*alpha_q in (28c). I read it as wording, but it should be cleaned up.\n\nFor a researcher in green ISAC, this is a useful idea and a reasonable starting point. It deserves a serious referee: the concept is new, the framework is coherent, and the gaps are fixable with a corrected constraint, a clearer rank-one argument, and more extensive simulations (varying rho, SNR, maneuvers, with error bars). I'd send it to review and ask for those revisions.","headline":"A genuinely new 'when to sense' decision inside an ISAC beamforming SDP, with a clean convex relaxation—but the power savings are partly baked into the relaxed constraint and track-loss robustness needs more than one simulation.","tokens_in":13987,"tokens_out":3072,"would_cite":true,"duration_ms":31464,"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 sensing-skip policy driven by tracker confidence can cut transmit power in ISAC systems without degrading communication and with acceptable tracking loss.","keywords":["Integrated sensing and communication","Energy-efficient beamforming","MIMO dual-functional radar-communication","Extended Kalman filter tracking","Sensing skipping","Normalized innovation squared","Transmit covariance optimization","Track-loss probability"],"falsifier":"Run the same skip-aware policy on a target executing an abrupt maneuver (e.g., a sudden acceleration) or with ρ set to 0.1; if the EKF loses track quickly or the track-loss probability jumps sharply, the fixed safety-floor assumption fails. Conversely, an analytic posterior Cramér–Rao bound under intermittent illumination showing that the bound remains finite for ρ > 0 would support the claim.","tokens_in":13068,"feed_emoji":"📡","tokens_out":2665,"duration_ms":30593,"temperature":0.7,"pith_summary":"The paper is trying to establish that always-on sensing is wasteful in integrated sensing and communication (ISAC) systems, and that sensing can be selectively skipped when the target state is already well known. It proposes a beamforming framework that minimizes transmit power under per-user SINR constraints and a minimum sensing illumination floor, with sensing activation decided by an extended Kalman filter (EKF) posterior confidence metric and a normalized innovation squared (NIS) gate. The central claim is that this confidence-driven skip policy achieves the lowest normalized transmit power across the whole SNR range while keeping the communication sum-rate identical to baselines and tracking loss between the full-probing and periodic-probing references. A sympathetic reader would care because this points toward a practical path for green 6G ISAC: turning sensing from a continuous drain into an adaptive, information-driven process.","feed_headline":"Pause radar probing when tracks are certain, save power","feed_subtitle":"EKF confidence and innovation gating decide when to sense; sum-rate stays flat and track loss stays near full-probing levels.","key_machinery":"The key mechanism is the skip-aware sensing policy and the safety illumination floor embedded in a convex transmit-covariance optimization. The EKF maintains a constant-velocity state with covariance P; the scalar confidence η = 1/tr(P) shrinks as the estimate sharpens, and the NIS Λ = yᵀS⁻¹y gates whether a measurement is consistent. The binary activation variable u_q,l+1 = 1{η_q,l < η_min ∨ Λ_q,l > γ_α,d} triggers full sensing only when confidence drops or innovation is inconsistent. During skipped epochs, only a fraction ρ of the nominal illumination α_q is imposed (Eq. 28c), so the target remains weakly observable. This lets the SDP reallocate power away from unnecessary probing and towa","core_discovery":"The paper's central claim is that radar probing can be temporarily skipped when the EKF indicates high confidence and consistent measurements, without losing the track. The skip decision is made per target per epoch using two complementary statistics: the posterior confidence η_q,l = 1/tr(P_q,l) and the NIS gate Λ_q,l ≤ γ_α,d. When sensing is skipped, a nonzero safety illumination floor ρα_q is still enforced in the beamforming optimization (Problem 28), preserving enough observability for the EKF to remain stable. The resulting transmit-covariance SDP minimizes total radiated power subject to per-user SINR constraints and sector-based beampattern guarantees. Numerical results show that the","pith_inferences":["The fixed ρ = 0.3 safety floor is validated only through simulation; an analytic relation between ρ, target dynamics, and track-loss probability would let a system adapt ρ per target or per scenario rather than picking a constant.","The identical sum-rate across schemes is an artifact of the min-power formulation; under a rate-maximizing objective, the saved sensing power could instead be converted into higher throughput, a regime the paper does not explore.","The confidence/NIS gating could be applied to other tracking filters (e.g., unscented Kalman or particle filters) or to extended-target tracking, where the innovation distribution is non-Gaussian and the NIS gate would need re-calibration.","The paper's claim implies a testable prediction: in a scenario with maneuvering targets or stronger clutter, the fixed ρ floor will become insufficient and track-loss will rise sharply; the optimal ρ would need to increase with target agility."],"forward_implications":["If correct, continuous sensing is unnecessary for reliable tracking; a predictive EKF with a minimal illumination floor can sustain tracks at much lower power.","The communication sum-rate is unaffected by the skip policy because the SINR constraints are fixed; energy savings translate directly into higher energy efficiency, not lower QoS.","The gap in track-loss probability between full probing and adaptive skipping narrows at higher SNR, suggesting that confidence-driven skipping becomes safer as channel conditions improve.","The framework extends naturally to adaptive scheduling of sensing resources across multiple targets, prioritizing only those targets whose estimates carry high uncertainty.","The safety floor ρ is a tunable knob that trades energy against tracking robustness; the paper identifies ρ = 0.3 as a balanced operating point."],"fun_headline_variants":["Sense only when uncertain: EKF gating cuts ISAC power","Skip radar probing when tracks are confident, save transmit power","EKF confidence gates sensing, cutting ISAC power while keeping tracks","Sense only when needed: EKF-based gating slashes ISAC power","Radar skips probing when tracks certain, cuts transmit power"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The load-bearing premise is that the reduced illumination floor ρα_q during skipped epochs keeps the EKF observable enough to avoid track divergence; the paper validates this only through simulation with ρ = 0.3 and provides no analytic guarantee tying ρ to track-loss probability.","fun_headline_variants_meta":{"raw":{"variants":["Sense only when uncertain: EKF gating cuts ISAC power","Skip radar probing when tracks are confident, save transmit power","EKF confidence gates sensing, cutting ISAC power while keeping tracks","Sense only when needed: EKF-based gating slashes ISAC power","Radar skips probing when tracks certain, cuts transmit power"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000678,"raw_usage":{"total_tokens":2927,"prompt_tokens":757,"completion_tokens":2170,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":501,"completion_tokens_details":{"reasoning_tokens":2080}},"tokens_in":501,"tokens_out":2170,"duration_ms":16577,"temperature":1.0,"reasoning_tokens":2080,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T05:04:29.155851+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same skip-aware policy on a target executing an abrupt maneuver (e.g., a sudden acceleration) or with ρ set to 0.1; if the EKF loses track quickly or the track-loss probability jumps sharply, the fixed safety-floor assumption fails. Conversely, an analytic posterior Cramér–Rao bound under intermittent illumination showing that the bound remains finite for ρ > 0 would support the claim.","supporting_citations":[],"review_version":1}