{"id":"6342c7a0-16c1-4655-b43e-ea7f2943969a","arxiv_id":"2511.23201","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A bistatic ISAC channel framework decomposes the channel into target and background parts, adds deterministic geometric clusters to TR38.901, and claims BER/capacity parity while enabling sensing evaluation.","lead":"This paper builds a radio channel model for 6G systems that can communicate and sense at the same time, splitting the channel into reflections from a target and a background of everything else. It extends the standard 3GPP channel model with fixed-geometry scatterers mixed with random clusters, and reports equal communication performance plus radar-like ranging and detection.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Stochastic-to-deterministic cluster conversion in Step 7S.a may silently alter delay/angle joint statistics; the paper's aggregate BER/capacity checks cannot detect this, so 'full compatibility with TR38.901' is not yet established for communication metrics sensitive to small-scale structure.","rationale":"After reading the full text, the central construction is a plausible extension of TR38.901: the background channel remains TR38.901, and the target channel is added via deterministic/stochastic clusters. The most load-bearing unsupported step is exactly the one the reader singled out: Step 7S.a converts stochastic clusters to geometric coordinates by forcing delay-angle consistency. TR38.901's cluster generation deliberately decorrelates delays and angles (beyond LSP correlations); forcing them onto an ellipse changes their joint distribution. The paper's validation metrics (BER, ergodic capacity) are aggregate and cannot detect this. I considered other possible objections—the measured-vs-simulated BER inconsistency, RCS values fitted in the authors' prior work, and the sensing validation being largely self-referential—but those are secondary or addressable within the paper's own claims. The statistical-preservation question is the one that, if wrong, undermines the headline 'drop-in TR38.901 replacement' while still allowing all the reported curves to look the same. My proposed test—direct before/after comparison of the converted clusters' small-scale statistics—would settle it. Because the reader's verdict is already CONDITIONAL and this concern is the same one, no verdict change is needed. The paper is not fatally flawed; it needs a statistical validation section.","tokens_in":23392,"tokens_out":7567,"duration_ms":77847,"concrete_test":"Run a Monte-Carlo statistical validation of Step 7S.a: generate N=1000 independent TR38.901 channel realizations for UMi, UMa, and InF using the Table II parameters; for each, apply the law-of-cosines mapping to the D=5 deterministic clusters; compare the empirical distributions of per-cluster delay, AoA/ZoA, AoD/ZoD, and power, plus the resulting delay spread and angular spreads, against the pre-conversion stochastic draws. Use a two-sample Kolmogorov-Smirnov test at the 5% level and quantify shifts in median delay spread and angular spread. If any environment shows a statistically significant or practically large (e.g., >10%) distributional change, the 'full compatibility' claim is not established for communication metrics beyond the aggregate BER/capacity shown.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In Section III-B (Step 7S.a), a subset of TR38.901 stochastic clusters in the target channel is converted to deterministic clusters by solving for a 3D position from cluster delay and one angle set (e.g., AA) via the law of cosines, then recomputing the complementary departure angles. This imposes an exact joint relation between delay, position, and angles that TR38.901's stochastic draws do not possess: in TR38.901, delays and angles are drawn from calibrated marginal distributions with only cross-cluster LSP correlations. After conversion, every deterministic cluster is forced onto a geometric ellipse consistent with its delay and chosen angles, so the joint distribution of delay and angles is altered whenever the original draw did not already lie exactly on that ellipse. The paper's evidence for 'full compatibility' is Fig. 4(a) BER and Fig. 7 ergodic capacity—both aggregate over many subcarriers/symbols and are insensitive to cluster-level delay-angle correlations and angular-spread distortions. The measurement comparison (Fig. 4(b)) is also only BER and is internally inconsistent (at 26 GHz/60° the measured channel requires less SNR than the simulated one at BER=10^-3), so it does not independently certify small-scale statistics. If the conversion biases delay spread, angular spreads, or delay-angle correlations, the drop-in replacement claim fails for communication modes that rely on spatial/temporal structure (beamforming, MIMO rank, positioning), even while BER parity holds.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a two-component GBSM for bistatic ISAC: a background channel (propagation not involving the sensing target) following TR38.901, and a target channel (Tx–target–Rx paths) characterized by target RCS, scattering points, and a hybrid mix of stochastic and deterministic clusters. A step-by-step generation procedure is given (general parameters, small-scale parameters, coefficients) for all four bistatic modes and the four Tx-target/target-Rx LoS/NLoS cases. The claims are that the model preserves communication performance parity with TR38.901 (BER, ergodic capacity) while enabling sensing evaluation (range error, ROC), supported by simulations in UMa/UMi/InF and by 26/28 GHz measurement-based BER comparisons.","tokens_in":23735,"tokens_out":12265,"duration_ms":121405,"significance":"If the model equations are correct and the validation is strengthened, the framework would be a useful, concrete input to 3GPP Release 19 ISAC channel modeling: it provides absolute-delay alignment, explicit target RCS/SP treatment, deterministic cluster mapping, and coverage of all bistatic modes. The measurement-based BER comparison is a real strength, as is the detailed step-by-step procedural structure. However, the central claim of 'full compatibility with TR38.901' is only as strong as the validation of the small-scale statistical properties, and the manuscript currently has several load-bearing equation-level issues that must be resolved before the results can be relied upon.","major_comments":[{"comment":"The tap delays for the concatenated target paths are written as differences. For example, τ1 = τ_{l,s} − τ_{n2,m2}, τ3 = τ_{n1,m1} − τ_{u,l}, and τ5 = τ_{n1,m1} − τ_{n2,m2}; similarly τ2, τ4, τ6 are differences of deterministic segment delays. A path that traverses the Tx-target link and then the target-Rx cluster link has a total delay equal to the sum of the segment delays, not their difference. As written, secondary target paths can have non-physical (negative or smaller-than-direct) delays, which directly corrupts the CIR tap placement and the subsequent range-error results. Unless these symbols are explicitly redefined as excess delays relative to some reference (which is not stated), this is a sign error that must be corrected and the sensing simulations re-run.","section":"Section III-C, following Eq. (5)"},{"comment":"The K-factor weights for the mixed LoS/NLoS cases are internally inconsistent with the definitions of H_NLoS1–H_NLoS3. When the Tx-target link is LoS and the target-Rx link is NLoS, the text sets γ1=0, γ2=η1, γ3=0, γ4=η̃2; when the roles are reversed it sets γ3=η2, γ4=η̃1. But H_NLoS3 is defined as the case with NLoS on both links, so γ4 should be 0 in both mixed cases, not η̃2 or η̃1. This introduces an unphysical both-NLoS contribution whenever exactly one link is NLoS. The correct weights should follow the same product structure as the all-LoS case (e.g., γ2 = η1η̃2, γ3 = η̃1η2) with the absent combinations set to zero. Please correct the weight table or explicitly justify a different decomposition.","section":"Section III-C, Eq. (6) and weight definitions"},{"comment":"The conversion of a subset of stochastic clusters into deterministic clusters fixes the 3D position from delay and one angle set and then recomputes the complementary angles. This enforces an exact delay-angle coupling that the original TR38.901 random draws do not possess. The paper does not statistically validate that delay spread, angular spreads, or delay-angle correlations of the remaining stochastic clusters and the overall channel are preserved. The BER and capacity comparisons in Figs. 4 and 7 are aggregate metrics that are largely insensitive to such cluster-level distortions, and since the background channel is TR38.901 by construction, equal-or-better BER is expected. Please provide quantitative comparisons of small-scale statistics (delay/angular spreads, MIMO eigenstructure, delay-angle joint distributions) or state explicitly that compatibility is claimed only for aggregate","section":"Section III-B, Step 7S.a; Section IV-A validation"},{"comment":"No Monte-Carlo counts, confidence intervals, or error bars are reported for any BER or capacity curve. The parity claim rests on small differences (e.g., 1 dB or less at BER=10^-3), and without uncertainty quantification these differences are not meaningful. Also, the comparison does not state whether total received channel power is normalized between the proposed ISAC channel and the TR38.901 baseline. Because the proposed model adds target paths and deterministic clusters, at the same transmit SNR the extra received energy alone can explain the marginal BER/capacity improvement. Please specify the power normalization and report the number of channel realizations/bits used.","section":"Section IV-A, Figs. 4(a), 7"},{"comment":"The measurement-based validation is limited to BER, and the text's generalization is not supported: the authors state that the measured channel exhibits 'marginally elevated BER' relative to simulation, but at 26 GHz/60° the measured curve requires 37.7 dB versus 40.75 dB for the simulated curve at BER=10^-3, i.e., the measured channel is better, not worse. No confidence intervals are given for the measured BER. Furthermore, the sensing claims (range error, ROC) are not corroborated by the measurements; the real-world comparison validates only the communication component. Please provide a more precise characterization of measurement uncertainty and either remove or qualify the sensing-related validation claim.","section":"Section IV-A.2, Fig. 4(b)"}],"minor_comments":[{"comment":"The law-of-cosines expression should be derived explicitly. The sign convention for â_p1 (from target to cluster vs. from cluster to target) and the definition of d in the denominator are easy to misread; a short derivation would remove ambiguity.","section":"Section III-B, formula for |a_p1|"},{"comment":"Several typographical errors remain, e.g., 'require SNR pf46dB' in Section IV-A.1 and inconsistent use of 'AA'/'AD' vs. 'AoA'/'AoD'. A careful proofread is needed.","section":"Throughout"},{"comment":"The phrase 'measured ISAC channels' is imprecise: the measurements provide a measured CIR that is then used in an OFDM BER simulation. The degree to which the reported BER is a property of the measured channel should be clarified.","section":"Section IV-A.2"},{"comment":"The number of deterministic clusters (5) and their RCS (0.1 m^2) are chosen without sensitivity analysis. Since these are free parameters of the framework, their influence on both communication parity and sensing results should be discussed.","section":"Table II and Section IV"}],"recommendation":"major_revision","confidential_remarks":"The manuscript overlaps substantially with prior ISAC channel-modeling work (e.g., [12]-[15]) in its hybrid deterministic/stochastic approach; the editor may want the authors to sharpen the novelty statement. The apparent sign/weight inconsistencies in Section III-C are the main technical blockers; they appear fixable, but the corrected equations will require re-running the sensing and CIR validation."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know three things. First, this is a genuine engineering contribution: a TR38.901-compatible framework that splits the ISAC channel into target and background components, adds geometrically anchored deterministic clusters to the standard stochastic ones, and gives explicit channel coefficients for all four bistatic LoS/NLoS cases. Second, the central promise — communication parity plus sensing capability — is plausible but not actually proven by the evidence shown. Third, the soft spots are fixable, so this deserves referee time.\n\nWhat is actually new: the target/background decomposition follows 3GPP discussions, and the law-of-cosines cluster positioning is similar to QuaDRiGa, but the integration into a full TR38.901-style procedural framework, the complete coefficient equations for the four propagation cases, and the support for TRP-TRP / TRP-UE / UE-TRP / UE-UE modes make this a useful toolbox for 6G ISAC simulation. That is a real service to the community.\n\nThe main soft spot is overclaiming. The communication parity check is largely by construction: the background channel is TR38.901 plus extra target paths, so BER and capacity parity at low RCS is expected. The stress-test note about Step 7S.a is right: converting stochastic clusters to deterministic clusters via the law of cosines forces an exact delay-angle relation that TR38.901's statistical draws do not have, and the paper never checks that this preserves delay spread, angular spreads, or delay-angle correlations. Aggregate BER and capacity are insensitive to those statistics, so the drop-in-replacement claim is only established for the specific simulated metrics, not for beamforming or MIMO rank. That is a load-bearing assumption, but it is testable and the authors should test it.\n\nOther issues are smaller. There are no error bars or Monte Carlo counts on any simulation curve. The measurement comparison is internally inconsistent: at 26 GHz and 60° the measured channel needs about 3 dB less SNR than simulated at BER 10^-3, contradicting the claim that measured channels show marginally elevated BER. The RCS model comes from the authors' own prior paper, which is fine, but it is fitted, not independently verified. None of these are fatal; they are about matching the claims to the evidence.\n\nBottom line: this is a serious paper with a clear citation record and a framework people in the ISAC channel modeling subfield will want to use. It is not a new physical principle and the validation is softer than the prose. I would send it to peer review, with a strong request for Monte Carlo statistics, a statistical comparison of delay/angle distributions after cluster conversion, and a careful re-examination of the measurement numbers.\n\nYes, I would bring this to a reading group for the methodology discussion, and I would cite it if I were working on ISAC channel modeling.","headline":"Useful 3GPP-aligned ISAC channel modeling framework, but the 'full compatibility' claim rests on validation that is partly by construction and statistically thin.","tokens_in":24263,"tokens_out":2013,"would_cite":true,"duration_ms":21665,"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 hybrid extension of the standard TR38.901 channel model claims to preserve communication performance while adding bistatic sensing evaluation.","keywords":["ISAC","geometry-based stochastic channel model","TR38.901","bistatic sensing","radar cross-section","hybrid clustering","spatiotemporal consistency","6G"],"falsifier":"Run the proposed hybrid model and TR38.901 with identical scenario parameters over many random seeds and compare the empirical distributions of root-mean-square delay spread, angular spreads, and per-cluster powers; if the deterministic cluster conversion shifts these distributions beyond the standard's calibration variance, the communication-parity and statistical-fidelity claims would be contradicted.","tokens_in":23243,"feed_emoji":"📡","tokens_out":5002,"duration_ms":45105,"temperature":0.7,"pith_summary":"The paper aims to give 6G designers a single channel model that works for both communication and sensing. It splits the radio channel into a target channel (paths that bounce off the object being sensed) and a background channel (everything else), then extends the stochastic cluster model of the TR38.901 standard with deterministic clusters anchored to explicit 3D coordinates and absolute delays. The central claim is that this hybrid model reproduces TR38.901's communication metrics (bit-error rate, channel capacity) across urban macro, urban micro, and indoor factory scenarios while also producing realistic bistatic sensing metrics such as target range error and detection probability. If correct, it would let system-level ISAC simulations run without giving up standardization compatibility.","feed_headline":"Bistatic radar joins the 3GPP channel model without losing comms","feed_subtitle":"Hybrid deterministic-stochastic clusters let one simulation evaluate both bit-error rate and target detection.","key_machinery":"The key machinery is hybrid stochastic-deterministic clustering: a subset of the standard stochastic clusters is converted into deterministic clusters by fixing their 3D coordinates from arrival angles and absolute delay using the law of cosines, then recomputing departure angles to restore consistency. This anchors a subset of clusters to explicit positions so delays and angles are physically coherent for sensing, while the remaining stochastic clusters preserve the statistical character of the standard. The target channel is then built by cascading Tx-target and target-Rx links with RCS-weighted path gains, and the background channel is the standard stochastic model; the two are summed wit","core_discovery":"The central claim is that a geometry-based statistical channel model can serve both communication and sensing by decomposing the ISAC channel into target and background components, and by converting a subset of stochastic clusters into deterministic ones whose 3D positions, delays, and departure angles are recomputed from geometry via the law of cosines. The authors derive channel coefficients for all four bistatic propagation cases (LoS or NLoS on each leg of the Tx-target and target-Rx links), including Doppler contributions from target and deterministic-cluster motion. They validate that the resulting model stays within roughly a decibel of standard TR38.901 bit-error-rate performance in","pith_inferences":["A natural testable extension would be to compare the delay spread, angular spread, and K-factor distributions of the hybrid model against TR38.901; aggregate BER alone is unlikely to reveal subtle distortions introduced by the deterministic cluster conversion.","If the parity result generalizes, the same hybrid-cluster idea could be applied to monostatic ISAC (target co-located with the transmitter) by treating the target channel as a dedicated deterministic cluster inside a unified background channel.","The framework assumes a known RCS per scattering point, but in practice RCS is angle- and frequency-dependent; integrating empirical RCS statistics, as the paper does for a UAV target, is likely essential for realistic sensing evaluation.","The joint identification problem the authors flag — labeling measured paths as target vs. background — is the practical bottleneck; synthetic CIRs from this model could directly support supervised learning approaches to that labeling task."],"forward_implications":["If correct, ISAC system simulators can use one standardized model for both link-level communication metrics and sensing KPIs, avoiding separate ray-tracing and stochastic simulation chains.","The framework's support for all four 3GPP bistatic modes (TRP-TRP, TRP-UE, UE-TRP, UE-UE) would allow network-level sensing studies under a single modeling framework.","Because deterministic clusters carry explicit RCS and velocity parameters, the model can evaluate target-dependent sensing performance (range error, detection probability) for different target types and distances.","The observed BER parity in UMa, UMi, and InF suggests the model could serve as a drop-in TR38.901 replacement for communication simulations, preserving backward compatibility.","The measured-channel BER comparisons indicate the model tracks real indoor-factory channels within a few dB, supporting use in measurement-calibrated studies."],"fun_headline_variants":["Bistatic ISAC channel model keeps comms while adding radar","Hybrid clusters let one channel model do radar and comms","Geometry-based ISAC model: bistatic radar without losing link","3GPP channel model gains sensing, keeps bit-error rate","One channel model for both sensing and comms in bistatic systems"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The load-bearing premise is that converting a subset of stochastic TR38.901 clusters into deterministic clusters via the law of cosines and recomputed departure angles preserves the small-scale statistical properties (delay, angle, power distributions) that TR38.901 was calibrated to reproduce, and the paper only checks aggregate BER, which is insensitive to such distortions.","fun_headline_variants_meta":{"raw":{"variants":["Bistatic ISAC channel model keeps comms while adding radar","Hybrid clusters let one channel model do radar and comms","Geometry-based ISAC model: bistatic radar without losing link","3GPP channel model gains sensing, keeps bit-error rate","One channel model for both sensing and comms in bistatic systems"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000274,"raw_usage":{"total_tokens":1444,"prompt_tokens":682,"completion_tokens":762,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":426,"completion_tokens_details":{"reasoning_tokens":674}},"tokens_in":426,"tokens_out":762,"duration_ms":6659,"temperature":1.0,"reasoning_tokens":674,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T19:34:39.593384+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the proposed hybrid model and TR38.901 with identical scenario parameters over many random seeds and compare the empirical distributions of root-mean-square delay spread, angular spreads, and per-cluster powers; if the deterministic cluster conversion shifts these distributions beyond the standard's calibration variance, the communication-parity and statistical-fidelity claims would be contradicted.","supporting_citations":[],"review_version":1}