{"id":"f35538d8-f906-416b-8c42-1765219ff20a","arxiv_id":"2505.03090","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A single base station with two static RISs and a user-supplied height can localize a user within 1.5 meters in over 98 percent of simulated trials, using time-multiplexed sensing and communication.","lead":"The paper proposes a wireless system where one base station and two static reflective surfaces locate a user by timing reflected signals and asking the user to send its height. It is a low-cost, low-complexity design for 6G-style integrated sensing and communication, but the supporting evidence is mostly simulation.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Localization accuracy depends on an unverified exactly-three-path assumption; Section V.C's claim that cascade reflections eliminate extra paths is unsupported, so the 98% success rate may rest on perfect path identification.","rationale":"The paper's central claim of localization success above 98% depends on the BS receiving exactly the three modeled paths and reliably associating each ToA estimate with the correct path. The reader's weakest assumption was the cooperative height feedback, which is real but explicitly acknowledged in the manuscript and can be seen as a design requirement rather than a hidden flaw. The more insecure condition is the unproven claim in Section V.C that two or more random reflections in cascade eliminate all alternative paths. This claim is used to justify the three-path model, yet no derivation, ray-tracing study, or quantitative power threshold is provided. The AoA-based path identification in Section III.D is also under-specified: with only three antennas, the paper does not show how the received mixture is resolved into the three signal vectors used in Eqs. (35)-(36), nor whether the max-correlation rule is robust to noise and to the user's position relative to the RIS directions. The Monte Carlo methodology in Section IV.B mentions AoA-based anchor identification, but the reported error model concentrates on ToA errors; no path-classification failure probability is included. Thus the 98% success rate is at risk of being an artifact of perfect path separation. This concern does not require rejecting the paper; a targeted simulation with a realistic received-signal model and an added fourth path would determine whether the concern lands. Since the reader already assigned a conditional verdict and this concern supports that conditionality rather than overturning it, the verdict remains unchanged.","tokens_in":19500,"tokens_out":12180,"duration_ms":136524,"concrete_test":"Re-run the Monte Carlo with a full received-signal model at the 3-antenna BS: y = sum_{i=1..3} a(theta_i) s_i(t) + n, with s_i carrying the ToA delays including error terms, and apply the correlation classifier specified by Eqs. (35)-(36) to classify paths. Compute the success rate of Eq. (42) using only correctly classified paths; if misclassification or unresolvable paths occur in more than 2% of trials, the 98% claim fails. As a second check, add one specular fourth path (e.g., user-to-wall-to-BS) with power 10 dB below the direct path and determine whether the classifier still identifies Path1, Path2, and Path3 correctly.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central 98% localization claim requires that the BS can separate and correctly identify the three modeled paths. Section III.D describes only a simple correlation classifier with M=3 sensing antennas, but does not specify how the three received signals x_i are extracted from the mixture. Section V.C then asserts without derivation that \"two or more random reflections in cascade eliminate any possibility of creating alternative paths beyond those identified as Path1, Path2, and Path3.\" This assertion is not quantified or proven. If a fourth path reaches the BS with non-negligible power, or if the direct path's steering vector is correlated with a RIS steering vector, the max-correlation rules in Eqs. (35)-(36) can mislabel paths. Critically, the Monte Carlo in Section IV.B states that anchors are identified using AoA, but the error model only injects ToA noise; no path-misclassification probability is included. The reported >98% success rate is therefore conditional on an unvalidated path-uniqueness and perfect-AoA assumption. This is load-bearing because a single mislabeled path feeds the wrong distance into Eqs. (29)-(31), and the trilateration solution is then incorrect.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes an ISAC system in which a single base station (BS) and two static RISs perform three-dimensional user localization by time-multiplexing communication and sensing frames. A sensing signal is reflected by the user along three modeled paths: a direct path back to the BS, a path via RIS 1, and a path via RIS 2. The BS estimates ToA and uses a simplified AoA correlation procedure to identify the three paths, then solves a linearized trilateration system with Cramer's rule; the two-candidate ambiguity in the height coordinate is resolved using user-reported height in a Permission and Response Frame. The RISs are configured once using passive beamforming, and the paper reports O(1) sensing complexity, 0.4 mJ energy per sensing cycle, and a Monte Carlo localization success rate above 98% within a 1.5 m error threshold. A probabilistic ISAC error model combining localization success and BER is also presented and compared with joint simulation.","tokens_in":19753,"tokens_out":4769,"duration_ms":53737,"significance":"If the central claims hold, the system offers an interesting single-BS, two-RIS ISAC architecture with very low sensing energy and constant localization complexity, which would be a useful data point for low-cost ISAC deployment. The explicit time-frame design, the analytic linearization of the trilateration problem, the complexity comparison against MUSIC- and STCM-based systems, and the 5000-trial Monte Carlo evaluation are concrete strengths. The significance is weakened, however, by three issues: the 98% success claim is conditional on an unverified exactly-three-path assumption with no modeled path-misclassification; the 'probabilistic' validation model shares its localization component with the simulation, so the agreement in Fig. 7.c does not independently validate the localization error model; and the sensing operation is cooperative rather than passive because it relies on the user reporting height. These limitations are addressable, but they currently bound the strength of the paper's conclusions.","major_comments":[{"comment":"The claim that 'two or more random reflections in cascade eliminate any possibility of creating alternative paths beyond those identified as Path1, Path2, and Path3' is unsupported. The paper provides no quantitative path-loss or delay-spread analysis for higher-order reflections, and the Monte Carlo error model in Section IV.B injects only ToA noise while assuming that the AoA-based path identification in Eqs. (35)-(36) is always correct. If a fourth path arrives with non-negligible power, or if the direct path is mislabeled as a RIS path, the max-correlation rules can feed the wrong distance into Eqs. (29)-(31) and produce an incorrect trilateration solution. The >98% success rate is therefore conditional on perfect path identification. This assumption should be tested explicitly, e.g., by simulating cascaded reflections with a multipath model and by adding a path-misclassification probability to the error model.","section":"§V.C, last paragraph; §III.D"},{"comment":"Equation (34) contains an apparent typo: the round-trip time for Path 3 is written as t3 = (d1 + d2 + dR2-BS)/c + ε3, but it should use d3, the distance between the user and RIS 2, in place of d2. As written, the formula describes the same user-RIS distance as Path 2 and would produce a systematically wrong distance estimate for the third anchor. The authors should correct Eq. (34) and confirm that the reported Monte Carlo results, especially the 98% success rate in Section V.C, are reproduced with the corrected formula.","section":"§III.D, Eq. (34)"},{"comment":"The validation of the probabilistic ISAC error model is partially circular. The 'probabilistic' curve in Fig. 7.c is obtained from the complement of Eq. (38), but the localization term P({unsuccessful location}) is taken from the simulated success rate SRSENS_sim in Eq. (42), not from an independent analytical model. Consequently, the agreement between the probabilistic and simulated curves in Fig. 7.c validates only the substitution of the theoretical BER for the simulated BER, not the localization error model itself. To support the stated validation claim, the authors should derive an analytic or semi-analytic characterization of the localization error distribution that does not reuse the same Monte Carlo samples, or alternatively re-label the comparison as a consistency check.","section":"§IV.B, §IV.C, Fig. 7.c"},{"comment":"The localization mechanism requires the user to transmit its height inside the Permission and Response Frame whenever it detects the Sensing Frame; the paper itself states that this information 'is critical for this ISAC model to function effectively.' This makes the sensing operation cooperative and limits the system to users that respond correctly. The failure mode for a non-cooperative user, an out-of-coverage user, or a user reporting an incorrect height is not quantified, and the paper's framing of the system as replacing three additional sensing BSs should be qualified accordingly. This limitation should be stated explicitly and, if possible, analyzed with a sensitivity study on height-report error.","section":"§III.E, §III.A, Fig. 2"},{"comment":"The RIS configuration procedure is post-hoc. The paper reports that the first attempt, programming the RIS to cover the first quadrant (1Q), was inadequate, and that the RIS elements were reprogrammed based on the BS-to-RIS path to cover the second quadrant (2Q) before the system worked. This means the static RIS configuration is not derived from a systematic optimization or a stated design rule; the reported coverage heat map in Fig. 6 and the subsequent 98% localization success are tied to a configuration found by trial and error. The authors should either provide a deterministic rule for choosing the RIS phase configuration for a given environment or include a sensitivity analysis over RIS configurations to show that the result is robust rather than configuration-specific.","section":"§V.A, Table II"}],"minor_comments":[{"comment":"The manuscript contains numerous typographical and grammatical errors, including the section heading 'Beanforming passivo' and the phrase 'with a W GN channel'; a thorough language edit is needed before publication.","section":"Throughout"},{"comment":"The notation for measured versus actual distances is inconsistent: Eq. (22) minimizes over di, described as 'actual distances,' while Eq. (24) and the surrounding text refer to measured distances d. Using distinct symbols for measured and true distances would clarify the optimization problem.","section":"§III.C, Eqs. (22)-(25)"},{"comment":"The Monte Carlo curves are reported without error bars or confidence intervals. Given that the central claim is a success rate above 98% from 5000 trials, the authors should report the statistical uncertainty of that estimate.","section":"§V.C, Fig. 7"},{"comment":"The BER expressions in Eq. (40) are written for the normalized channel h, but the SNR variable in the figures is labeled |h|^2 Eb/N0; the connection between these quantities should be stated explicitly so that the reader can relate the BER curves to the localization SNR thresholds.","section":"§IV.A, Eq. (40)"},{"comment":"The 'Passive Beamforming Target (RIS)' parameter is listed as (-55, 5, 0), which lies outside the monitored region shown in Fig. 4 and Fig. 6; the meaning of this parameter and its role in the configuration should be explained.","section":"§V.A, Table II"}],"recommendation":"major_revision","confidential_remarks":"The paper is readable as an early-stage system-design contribution, but the comparison with prior ISAC localization work is narrow and relies heavily on the authors' own prior publications; a more thorough positioning against recent RIS-ISAC localization literature would help establish novelty. The manuscript also appears to be an early arXiv version, with several typos and an obvious error in Eq. (34); the authors should be encouraged to revise carefully before resubmission."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a clearly written engineering proposal for an ISAC localization system with one BS, two static RISs, and user-supplied height to pick between two trilateration solutions. The architecture is concrete and the complexity analysis is sane. But the central performance number is softer than it looks: the 98% success rate comes from a simulation that injects ToA noise only and assumes the BS always identifies the three paths correctly. The paper's own claim that cascade reflections cannot create extra paths (Section V.C) is asserted, not shown, and the Monte Carlo does not include any path-misclassification probability. That makes the headline accuracy conditional on an unverified assumption. This is the biggest soft spot.\n\nWhat's new: the specific combination of one BS, two static RISs, and height feedback to remove the fourth sensing anchor. The frame structure and energy calculation are concrete. The complexity comparison (O(1) after fixing M=N=3) is fair.\n\nOther issues: Eq. (34) has a likely typo — t3 uses d2 where it should use d3. The RIS configuration was reprogrammed after the initial 1Q design failed; that is post-hoc tuning, and there is no sensitivity analysis to show the result holds across environments. The probabilistic model in Eq. (38) is validated against the joint simulation, but both sides share the same noise model and threshold, so the match is not independent confirmation. And the system depends on the UE sending its height in the Permission and Response Frame; the paper admits this is critical, so calling it 'passive' sensing would be misleading.\n\nThe paper does its own homework: it cites the trilateration source [20], acknowledges the ambiguity, and the algebra reduces to standard trilateration. That is not a flaw, just a limit on novelty.\n\nWho gets value: people working on low-complexity ISAC positioning or RIS-assisted localization will find a useful reference design. It deserves peer review because the architecture is testable and the flaws are fixable, but the reported numbers need to be re-derived with a realistic path model and direct baseline comparisons.\n\nMy recommendation: send it to review, but expect revision. If the authors add a path-identification failure model and correct Eq. (34), the paper could be a solid incremental contribution.","headline":"A concrete but modest single-BS ISAC architecture; the 98% localization claim rests on an untested perfect-path-identification assumption, so treat the headline number as optimistic.","tokens_in":20276,"tokens_out":2833,"would_cite":false,"duration_ms":29304,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":null,"created_at":"2026-08-15T23:59:41.756279+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":null,"supporting_citations":[],"review_version":1}