{"id":"ff2c30c7-61ee-4df0-9b37-5664b2f3f2f7","arxiv_id":"2507.02613","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"Simulations of a 5G PRS multistatic sensing chain report roughly five-meter position errors, sub-meter-per-second speed errors, and successful Kalman tracking of linear and curved targets.","lead":"Using ordinary 5G positioning signals as a multistatic radar, this paper simulates a full chain that detects a moving target, estimates its position and velocity from several receivers, and tracks it over time. It is worth reading because it shows how existing cellular signals could provide sensing without dedicated radar hardware, if the idealized simulations hold up in practice.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The CAF Doppler grid is capped at ±400 Hz (≈±48 m/s radial speed), while Eq. (2) permits radial speeds up to 60 m/s for the stated 20–30 m/s target speeds, so valid simulated scenarios are outside the searched grid.","rationale":"The reader's CONDITIONAL verdict is appropriate, and my analysis does not change it. The estimator chain (CAF, nonlinear least-squares trilateration, ridge-regularized velocity inversion, and EKF tracking) is standard and the reported errors are plausible for the clean simulated signal model. However, the paper's own numerical setup contains a concrete internal inconsistency: the Doppler search range of ±400 Hz corresponds to radial speeds of only ±48 m/s, while Eq. (2) and the stated target speed range of 20–30 m/s together allow radial speeds up to 60 m/s, i.e., Doppler shifts up to 500 Hz. A single valid target geometry yields 430 Hz, outside the grid. This means the global-maximum CAF detection cannot work for the entire claimed parameter space, which directly undermines the headline 'high-fidelity detection' for the stated random scenario. The issue is not fatal—widening the grid to ±500 Hz or narrowing the claimed speed range would restore consistency—so the appropriate verdict remains CONDITIONAL rather than REJECT. The reader's direct-path cancellation concern is also valid and would require an over-the-air or more realistic simulation test, but it is harder to settle from the text alone. The Doppler-grid concern, by contrast, can be settled by rerunning the paper's own scenario at one specific target/velocity draw. Since no code was released, this test is the fastest check of whether the published averages are complete.","tokens_in":9389,"tokens_out":14181,"duration_ms":181734,"concrete_test":"Rerun the Section II.D single-target experiment with the target at p = (400, 0) m and v = (30, 0) m/s, using the exact same CAF implementation and Doppler grid of Section II.B. Verify whether the global maximum of |CAF_m(d, f)| for receiver 1 occurs at the true delay with f ≈ 430 Hz, or instead at a grid edge. If the true peak is not found or the estimated speed/position error is far above the Table IV averages, the Doppler grid must be widened to at least ±500 Hz, after which the same scenario should recover the reported ~5 m position error and ~0.4 m/s speed error. A sweep over target positions with radial speed just above 48 m/s would map the failure region.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing concern is an internal mismatch in the paper's own simulation setup. Section II.B sets the CAF Doppler grid to f ∈ [−400, 400] Hz with 401 bins, but the received Doppler in Eq. (3) is f_d,m = v_m f_c / c, and Eq. (2) defines v_m = v^T((p−t)/||p−t|| + (p−r_m)/||p−r_m||). The summed unit vectors have norm up to 2, so for target speeds of 20–30 m/s, v_m can reach 60 m/s, producing Doppler shifts up to 500 Hz at f_c = 2.5 GHz. Concretely, with t = (250, 144) m, r1 = (0, 0) m, p = (400, 0) m, and v = (30, 0) m/s, receiver 1 sees v_1 ≈ 51.6 m/s, i.e., f_d ≈ 430 Hz, outside the [−400, 400] Hz grid. Under the global-maximum CAF rule, the true target peak is not even searched, so the three-receiver detection and velocity chain fails or produces a large bias for this valid draw. The reported averages over random positions and speeds therefore either quietly exclude such draws or are contaminated by large errors, and the central claim that high-fidelity detection holds for speeds uniformly sampled from 20–30 m/s is not supported. This issue is independent of the direct-path cancellation concern raised by the reader; it is a mismatch between the stated parameter range and the implemented Doppler search.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes MULTI-SCOUT, a signal-processing chain for multistatic integrated sensing and communications (ISAC) using 5G Positioning Reference Signals. It estimates bistatic range and radial velocity at each receiver via a coherent cross-ambiguity function (CAF), fuses the ranges by nonlinear least-squares trilateration, recovers the target velocity by regularized linear inversion, extends the formulation to clock-bias estimation, 3D geometry, and multi-target association, and feeds the position/velocity estimates to standard and extended Kalman filters for tracking. Performance is evaluated through simulations in 2D and 3D, reporting position errors of about 5–10 m, speed errors of about 0.2–0.5 m/s, and angle errors of about 1–3 degrees, with KF/EKF tracking for linear and circular trajectories.","tokens_in":9718,"tokens_out":8176,"duration_ms":95396,"significance":"The paper is a useful system-level integration of standard estimation and tracking techniques, and it demonstrates a complete pipeline from PRS waveform generation through detection, positioning, association, and tracking. Its strengths include an explicit treatment of receiver clock bias, a 3D extension, and a combinatorial association mechanism for multiple targets. If the numerical claims are validated, the work could serve as a reference design for PRS-based multistatic ISAC. However, the evidence is entirely synthetic and generated from the same signal model that the estimator assumes; there are no comparisons with baselines or Cramér–Rao bounds, no variance measures across simulation runs, and the Doppler-grid mismatch described below directly affects the reported averages. These issues need to be addressed before the central claims can be accepted.","major_comments":[{"comment":"The Doppler search grid is limited to f in [-400, 400] Hz, but the radial velocity in Eq. (2) can be as large as twice the target speed because the two unit vectors are summed. Since target speeds are sampled uniformly from 20 to 30 m/s, radial speeds up to about 60 m/s are possible, corresponding to Doppler shifts up to about 500 Hz at f_c = 2.5 GHz. For example, with t = (250, 144) m, r1 = (0, 0) m, p = (400, 0) m, and v = (30, 0) m/s, receiver 1 sees v_1 approximately 51.6 m/s, i.e., f_d approximately 430 Hz, which is outside the searched grid. Under the global-maximum CAF rule, such valid draws cannot be detected, so the averaged results in Tables IV and XV either exclude these draws or are contaminated by large errors. The Doppler grid should be enlarged (e.g., to at least +/-500 Hz) or the admissible speed range should be justified.","section":"II.B and II.D, Eqs. (2)–(3)"},{"comment":"The entire detection chain assumes that after direct-path cancellation, the global maximum of the CAF is the target echo. The direct-path cancellation is described in one sentence (matching the received samples with the PRS and subtracting a scaled, time-shifted replica) with no implementation details, no model of residual cancellation error, and no sensitivity analysis. In a real 5G downlink, the direct-path signal is orders of magnitude stronger than the target echo, and any residual leakage or multipath will shift the CAF global maximum and break the subsequent trilateration, velocity, association, and tracking stages. Since all simulations assume ideal cancellation, the central claim of high-fidelity detection with 5G PRS is not yet supported. The authors should provide a direct-path cancellation model, quantify the residual error, or include a scenario with direct-path leakage.","section":"II.A"},{"comment":"The echo amplitude alpha_m in Eq. (3) is never defined. Section II.D lists transmit power, antenna gains, radar cross section, and noise variance, but no path-loss or radar equation links these parameters to alpha_m or to the signal-to-noise ratio. Consequently, the simulated CAF peaks and all reported errors depend on an unspecified constant, and the experiments are not reproducible or comparable with other work. Please provide the propagation model, the resulting SNR, and the mapping from system parameters to alpha_m.","section":"II.A and II.D"},{"comment":"The multi-target experiments place the two targets far apart in bistatic range and Doppler (e.g., receiver 1 sees 290 m and 653 m, and -5.3 m/s and -38.1 m/s). Thus the association results do not exercise the claimed capability, stated in the Introduction, to preserve high-fidelity estimates 'even when echoes overlap.' The non-maximum-suppression procedure that selects exactly N_T peaks will fail when two targets appear in nearby delay and Doppler bins, but no such scenario is simulated or analyzed. Please either add an overlapping-echo case or temper the claim accordingly.","section":"V, Tables XII–XV"}],"minor_comments":[{"comment":"The number of random initializations for the Trust-Region Reflective solver is not specified; please state the number of seeds and the stopping criterion used.","section":"II.C"},{"comment":"The 'total' measurement and filtering errors in Table XVI should be defined explicitly as cumulative over the 25 s track; without this clarification, the hundreds-of-meters values appear inconsistent with the per-snapshot errors of a few meters in Tables IV, VII, X, and XV.","section":"VI, Table XVI"},{"comment":"The guard window sizes used in the 2D non-maximum suppression are not given; these values affect whether close peaks are resolved, so they should be reported and their effect discussed.","section":"V"},{"comment":"The KF/EKF covariance matrices Q and R are stated without justification or sensitivity analysis; please explain the chosen values and their influence on the results.","section":"VI"},{"comment":"All results are based on synthetic data generated from the same signal model the estimator assumes; a comparison with a Cramér–Rao bound or with a baseline estimator would significantly strengthen the significance of the reported accuracies.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The paper fits an applied systems-oriented venue. The Doppler-grid inconsistency in Section II.B is a concrete, fixable error that undermines the reported averages, and the missing propagation model and ideal direct-path cancellation assumption are additional load-bearing gaps for the 'high-fidelity' claim. I do not recommend rejection because the methodology is standard and the pipeline is coherent; a careful revision with corrected simulations and a clearly stated propagation/cancellation model could make the claims credible."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague, quick take on arXiv:2507.02613 (MULTI-SCOUT). It's a clean simulation study that wires standard 5G PRS processing blocks—CAF matched filtering, nonlinear least-squares trilateration, ridge-regularized velocity inversion, a permutation-based association scheme, and KF/EKF tracking—into one multistatic sensing chain. The pieces are not new; the specific chain and its evaluation on 20–30 m/s targets with clock bias and multi-target cases is. The paper is readable and the equations are standard. I think the numbers are internally consistent, but everything rests on synthetic data generated from the same model the estimator assumes, with no error bars, no baseline, and no Cramer-Rao comparison. That keeps it at conditional.\n\nThe bigger technical worry is the CAF Doppler grid. The paper searches f in [-400, 400] Hz (401 bins), but Eq. (2) lets radial speed reach twice the target speed. With speeds 20–30 m/s, v_m can be 60 m/s, which at f_c=2.5 GHz is 500 Hz. I checked a concrete draw: t=(250,144), r1=(0,0), p=(400,0), v=(30,0) gives v_1≈51.6 m/s, fd≈430 Hz, outside the grid. Under the global-maximum CAF rule, the true peak is not searched, so the velocity chain degrades for valid draws. The reported small averages suggest either those draws were excluded or the effect is hidden. This is an internal mismatch, not an experimental nitpick.\n\nThe reader's direct-path cancellation concern is real but more a practical challenge; an over-the-air test would tell. The lack of released code/data is a reproducibility problem.\n\nWho is this for? People working on 5G/6G ISAC who want a concrete baseline chain for multistatic PRS sensing. It deserves a serious referee, but I would want the Doppler grid fixed, a CRB comparison, and code released before trusting the numbers.","headline":"Tidy multistatic 5G PRS sensing chain, standard pieces and synthetic-only validation; the Doppler-grid mismatch is a real internal flaw, but the paper is worth a serious referee.","tokens_in":10298,"tokens_out":7090,"would_cite":false,"duration_ms":83260,"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 standard 5G Positioning Reference Signal, observed by multiple distributed receivers, can be turned into a multistatic radar-like sensing chain that detects, locates, and tracks moving targets with position errors…","keywords":["integrated sensing and communications","multistatic radar","5G positioning reference signal","cross-ambiguity function","range-Doppler estimation","target tracking","Kalman filter","trilateration"],"falsifier":"A controlled field test with the paper's three-receiver equilateral geometry, a strong line-of-sight path, and a target at known position would settle it: if the cross-ambiguity global maximum frequently lands on a non-echo bin after direct-path subtraction, the reported ~5 m positioning and sub-0.5 m/s velocity errors will not reproduce.","tokens_in":9178,"feed_emoji":"📡","tokens_out":5802,"duration_ms":60199,"temperature":0.7,"pith_summary":"This paper argues that a standard 5G Positioning Reference Signal can do double duty as a multistatic radar: one transmitter and several geographically separated receivers turn the same OFDM-PRS waveform into a sensing system that detects, localizes, and tracks moving targets without additional radar hardware. The proposed chain extracts bistatic delay and Doppler from each receiver's cross-ambiguity function, fuses the delays by nonlinear least-squares trilateration to get position, inverts the radial-speed equations to get velocity, resolves which peaks belong to which target in multi-target scenes, and feeds the estimates into Kalman filters for smooth tracks. In simulation, the chain reports average position errors of roughly 5-10 m, speed errors of 0.2-0.5 m/s, and angle errors of a few degrees, with an extended Kalman filter tracking both linear and circular motion. If these numbers hold in real deployments, cellular networks would gain a passive sensing capability for traffic monitoring, security, and situational awareness at little extra spectrum cost.","feed_headline":"5G positioning signals can track moving targets to ~5 m","feed_subtitle":"Ordinary 5G positioning signals, seen by several receivers, become a radar for locating and tracking targets.","key_machinery":"The load-bearing object is the coherent cross-ambiguity function (CAF): matched-filtering each receiver's samples against the known PRS waveform over a delay grid and a Doppler grid, with parabolic interpolation around the peak, converts the communication signal into bistatic range and radial velocity measurements. The downstream machinery is nonlinear least-squares trilateration (trust-region reflective solver with random restarts) to fuse bistatic range ellipses into a position, ridge-regularized inversion of the radial-speed equations to recover velocity, 2D non-maximum suppression plus a residual-minimizing assignment to associate peaks in multi-target scenes, and standard and extended Kalman filters with a polar-coordinate state to track the resulting position-velocity sequence.","core_discovery":"The central claim is that multistatic ISAC can be built entirely on existing 5G PRS: a single gNB transmits the standard periodic OFDM-PRS waveform while three or four receivers process the same signal via a coherent cross-ambiguity function, extracting bistatic ranges and radial velocities with sub-bin accuracy through parabolic interpolation. For one target, three non-collinear receivers in 2D (or four non-coplanar in 3D) give a unique position by nonlinear least-squares trilateration, and a ridge-regularized inversion of the Doppler equations gives the full velocity vector; a fourth receiver additionally lets the system estimate and remove a common clock bias. For multiple targets, 2D non-maximum suppression picks candidate peaks and the assignment that minimizes the sum of trilateration residuals pairs them across receivers. Feeding the position-velocity snapshots into a standard Kalman filter tracks linear motion, while an extended Kalman filter with a polar-coordinate state tracks linear and circular motion. The paper's simulations report average position errors around 5-10 m and speed errors around 0.2-0.5 m/s across the 2D, 3D, synchronized, and multi-target settings, with the EKF reducing circular-motion tracking error from 818.76 m to 63.89 m.","pith_inferences":["Beyond the paper's simulations: real 5G deployments have stronger direct-path leakage and multipath than the model's subtraction step assumes, so the reported errors are likely to degrade first at the CAF peak-selection stage; a hardware-in-the-loop test would quantify this.","The paper fixes PRS numerology 0 and 15.36 MHz sampling; increasing bandwidth (higher subcarrier spacing) should sharpen range resolution roughly proportionally, a trade-off the paper leaves implicit.","The association search over (N_T!)^M permutations grows factorially with target count; branch-and-bound is mentioned but untested, so scaling beyond two or three targets is an open question.","The two symmetric low-cost assignments in Table XI show an inherent pairing ambiguity in the permutation search; using velocity consistency across time rather than a single snapshot is a natural error-correction step the paper does not explore."],"forward_implications":["If the approach is correct, any 5G network with PRS coverage can serve as a passive radar system, with no additional sensing hardware or spectrum.","Three synchronized receivers suffice for 2D positioning of a single target; a fourth receiver adds robustness to common clock bias, improving the practicality of distributed deployments.","In 3D, four non-coplanar receivers extend the same delay-fusion chain to full spatial positioning, so altitude becomes observable.","Multi-target scenes are handled by a residual-minimizing association step, and the EKF's polar-coordinate state allows tracking turning targets, not just straight-line motion."],"supporting_citations":[{"why":"Supplies the 5G NR PRS waveform structure (OFDM numerology, Gold-sequence PRS generation, comb mapping, cyclic prefixes) that the whole sensing chain assumes.","marker":"[13]"},{"why":"Demonstrates 5G network-based passive radar, providing the CAF-style range-Doppler processing that the multistatic chain adapts.","marker":"[7]"},{"why":"Shows 5G-based passive radar using channel response estimated via reference signals, grounding the use of reference signals for sensing.","marker":"[8]"},{"why":"Introduces 5G PRS-based sensing with a sensing reference signal approach, the direct precursor for using PRS as a radar waveform.","marker":"[6]"},{"why":"Covers moving-target sensing in NextG networks and serves as the monostatic/bistatic baseline that MULTI-SCOUT extends to distributed receivers.","marker":"[5]"},{"why":"Recent multistatic ISAC localization with 5G NR signals, the work this paper positions against as the multistatic state of the art.","marker":"[12]"}],"fun_headline_variants":["5G positioning signals act as radar for moving targets","Multistatic 5G PRS turns cell signals into target tracker","One 5G transmitter, four receivers: track targets with PRS","5G ISAC: using positioning signals to detect and track","5G PRS-based radar detects, positions, and tracks targets"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"Each receiver must remove the direct transmitter-to-receiver signal well enough that the strongest peak in its range-Doppler map is the target echo, not line-of-sight leakage, multipath, or clutter.","fun_headline_variants_meta":{"raw":{"variants":["5G positioning signals act as radar for moving targets","Multistatic 5G PRS turns cell signals into target tracker","One 5G transmitter, four receivers: track targets with PRS","5G ISAC: using positioning signals to detect and track","5G PRS-based radar detects, positions, and tracks targets"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00027,"raw_usage":{"total_tokens":1669,"prompt_tokens":1035,"completion_tokens":634,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":651,"completion_tokens_details":{"reasoning_tokens":545}},"tokens_in":651,"tokens_out":634,"duration_ms":7667,"temperature":1.0,"reasoning_tokens":545,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T20:25:11.666195+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A controlled field test with the paper's three-receiver equilateral geometry, a strong line-of-sight path, and a target at known position would settle it: if the cross-ambiguity global maximum frequently lands on a non-echo bin after direct-path subtraction, the reported ~5 m positioning and sub-0.5 m/s velocity errors will not reproduce.","supporting_citations":[{"cited_title":"NR: physical channels and modulation,","cited_arxiv_id":null,"evidence_quote":"Supplies the 5G NR PRS waveform structure (OFDM numerology, Gold-sequence PRS generation, comb mapping, cyclic prefixes) that the whole sensing chain assumes."},{"cited_title":"5G network-based passive radar,","cited_arxiv_id":null,"evidence_quote":"Demonstrates 5G network-based passive radar, providing the CAF-style range-Doppler processing that the multistatic chain adapts."},{"cited_title":"5G-based passive radar utilizing channel response estimated via reference signals,","cited_arxiv_id":null,"evidence_quote":"Shows 5G-based passive radar using channel response estimated via reference signals, grounding the use of reference signals for sensing."},{"cited_title":"5G PRS-based sensing: A sensing reference signal ap- proach for joint sensing and communication system,","cited_arxiv_id":null,"evidence_quote":"Introduces 5G PRS-based sensing with a sensing reference signal approach, the direct precursor for using PRS as a radar waveform."},{"cited_title":"I-SCOUT: Integrated sensing and communications to uncover moving targets in NextG networks,","cited_arxiv_id":null,"evidence_quote":"Covers moving-target sensing in NextG networks and serves as the monostatic/bistatic baseline that MULTI-SCOUT extends to distributed receivers."},{"cited_title":"Localization accu- racy improvement in multistatic ISAC with LoS/NLoS condition using 5G NR signals,","cited_arxiv_id":null,"evidence_quote":"Recent multistatic ISAC localization with 5G NR signals, the work this paper positions against as the multistatic state of the art."}],"review_version":1}