{"id":"1f95da11-4344-405d-8e8d-9a4493249157","arxiv_id":"2508.16980","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A finite-difference kinematic observer turns low-rate position samples into high-rate predicted positions, letting a reconfigurable intelligent surface steer its beam ahead of a moving user.","lead":"The paper shows that a simple physics-based predictor, fed by low-rate position fixes, can compute a fast-moving user's future location and keep a reconfigurable smart surface aimed correctly despite control delays. It matters because it offers a low-complexity alternative to high-rate localization or machine learning for real-time RIS beamforming in mobile wireless systems.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Eq. (18) as printed sums per-element powers with an unexplained e^{2j(...)} factor and no outer |.|^2, so the dB-loss numbers in Section IV may be artifacts of an incorrect power metric.","rationale":"The reader identified the smoothness/local-linearization assumption as the weakest point. That is a legitimate secondary concern, and the paper's own Gauss-Markov and high-noise results show degradation when the premise weakens. However, the single most load-bearing concern is more fundamental: the numerical evidence for the central claim is produced by the power model in Eq. (18), and that equation as written is not the correct coherent-combining expression for an RIS array. It omits the magnitude-squared of the total field and instead appears to sum per-element powers with a doubled phase factor. If this is what the simulations computed, then all of the dB-loss comparisons in Section IV, including the 0.14-0.60 dB versus 6.32-7.39 dB numbers in Table IV, are unreliable regardless of how well the predictor estimates the UE position. This concern does not require assuming any particular motion model; it attacks the measurement instrument used to validate the entire approach. Because the preprint provides no code, a reader cannot determine whether the simulator used the printed expression or a corrected coherent one. The proposed concrete check, re-running the key scenarios with the standard coherent array formula, would settle the issue. If the corrected formula preserves the reported losses within about 1 dB, then the reader's CONDITIONAL verdict stands on the weaker grounds of missing error bars and code. If not, the quantitative claims are unsupported. Hence I recommend keeping CONDITIONAL but with a new, more specific requirement: correct or justify Eq. (18) and recompute the headline numbers.","tokens_in":16374,"tokens_out":11509,"duration_ms":124932,"concrete_test":"Re-derive the received power for a single RIS cell from the bistatic radar equation with RCS \\sigma = 4\\pi D^4\\lambda^{-2}\\cos^2(\\theta^t)\\mathrm{sinc}^2(...), then write the full-array received power as P_r = P_t G_t G_r \\lambda^2 / ((4\\pi)^3 |\\mathbf{r}^t|^2|\\mathbf{r}^r|^2) \\times |\\sum_{m,n} \\sqrt{\\sigma_{m,n}} \\Gamma_{m,n} e^{-jk_0(|\\mathbf{r}^t_{m,n}|+|\\mathbf{r}^r_{m,n}|)}|^2. Re-run the UE1 scenario (Fig. 6) and the car scenario (Table IV) with this corrected metric and also with Eq. (18) as printed. If the average power losses or the ranking of OPS/OPA versus naive change by more than 1 dB, the published quantitative claims are an artifact of the metric; if they do not change materially, Eq. (18) is a typographical error and the concern is minor.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The received-power expression in Appendix A, Eq. (18), is the metric on which every numerical claim in Section IV rests. As printed, the summand contains |\\Gamma_{m,n}|^2 e^{2j(\\angle\\Gamma_{m,n} - k_0(|\\mathbf{r}^t_{m,n}|+|\\mathbf{r}^r_{m,n}|))} and the whole sum is not squared in magnitude. For an array whose elements are aligned according to Eq. (3), this expression yields a received power proportional to \\sum|\\Gamma|^2, i.e., an incoherent sum over M\\times N cells. Physical coherent reflection, by contrast, gives a received power proportional to |\\sum \\Gamma e^{-jk_0(...)}|^2, which for equal magnitudes scales as (MN)^2. The printed formula is also complex-valued, not a real power, unless an implicit real part or magnitude is taken. If the simulator literally implements Eq. (18), then the headline numbers (e.g., 0.14-0.60 dB average loss for OPS/OPA versus 6.32-7.39 dB for the naive scheme in Table IV) are computed under a metric that cannot correctly represent RIS beamforming gain, and they do not support the central claim. The paper provides no code or data artifact that would let a reader verify whether the simulations used a corrected coherent formula or the printed one.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a physics-based predictive beamforming framework for RIS-aided wireless systems. From low-rate noisy UE position measurements, a kinematic observer estimates velocity (OPS) or velocity and acceleration (OPA); a predictor then extrapolates the UE position ahead by an accumulated delay, and the RIS reflection phases are set from the predicted position using Eq. (3). The scheme is evaluated against ideal and naive benchmarks in four scenarios: a constant-acceleration trajectory (UE1), real car acceleration profiles (UE2), an aggressively piloted drone dataset (UE3), and a Gauss-Markov mobility model (UE4). The central claim is that OPS/OPA keep the average power loss close to ideal, e.g., 0.14-0.60 dB versus 6.32-7.39 dB for the naive scheme in Table IV.","tokens_in":16628,"tokens_out":9349,"duration_ms":103464,"significance":"The conceptual idea is simple, low-complexity, and parameter-free, and the use of external car and drone datasets is a genuine strength that goes beyond synthetic trajectories. If the numerical results are confirmed after correcting the power metric, the approach would be a useful practical contribution to RIS beamforming under localization and control latency. However, the printed received-power expression in Appendix A is not a valid coherent power formula, and the paper provides no code or data artifact. The numerical support for the central claim therefore cannot be accepted as it stands. The approach also inherits the standard constant-speed/constant-acceleration smoothness assumption, and the paper honestly shows degradation in high-randomness and high-noise regimes.","major_comments":[{"comment":"Equation (18) is not a physically correct received-power expression. As printed, the sum is complex-valued, contains |Γ_mn|^2 e^{2j(∠Γ_mn - k0(|r_t_mn|+|r_r_mn|))}, and has no outer magnitude squared. A coherent RIS model requires a magnitude-squared coherent sum of element contributions, essentially Pr ∝ |Σ sqrt(Gt Gr σ) Γ_mn e^{-jk0(|r_t_mn|+|r_r_mn|)} / (|r_t_mn||r_r_mn|)|^2. With the printed formula, perfectly aligned phases produce an incoherent sum scaling as MN rather than (MN)^2, and misalignment losses are governed by cos(2Δ) instead of |Σ e^{jΔ}|^2. Since every Section IV numerical claim, including Figs. 5c-10 and Table IV, rests on this backscattering block, the authors must correct Eq. (18) and rerun the simulations, or clearly state and justify the exact metric actually implemented.","section":"Appendix A, Eq. (18)"},{"comment":"The reported OPA loss of 3.9e-15 dB in the noise-free delay sweep is a circular result: UE1 is generated from Eq. (6) with constant acceleration, which is exactly the model embedded in the OPA observer and predictor. This should be presented as an implementation consistency check, not as evidence of predictive accuracy. The independent evidence for the approach comes from UE2-UE4, and those scenarios should be the basis for the accuracy claims.","section":"Section IV-B5, Fig. 7(b)"},{"comment":"The headline average losses (e.g., 0.14 vs 0.35 dB and 0.15 vs 0.60 dB for OPS vs OPA) are reported without confidence intervals or per-realization dispersion. With only 10 realizations for UE2 and 10-50 realizations in other scenarios, the OPS/OPA ordering and the 'close to ideal' claims are not statistically supported. Please report error bars, box plots, or per-realization distributions for the average-loss figures.","section":"Section IV-C, Table IV"}],"minor_comments":[{"comment":"The caption contains a redundant phrase: 'magnitude and phase of the reflection coefficient magnitude' should be 'magnitude and phase of the reflection coefficient'.","section":"Section II-A, Fig. 2(b)"},{"comment":"The relationship q = floor(Jk) with J = T/T_M is correct, but the dependence q = q(k) is implicit; writing q(k) explicitly in Eq. (12) would improve readability.","section":"Section III-B, Eq. (12)"},{"comment":"The paper cites the public drone dataset but does not provide simulation code or generated data. Given the ambiguity in Eq. (18), releasing the simulator would be essential for reproducibility and verification.","section":"Section IV"}],"recommendation":"major_revision","confidential_remarks":"The central idea is plausible and the use of external real-world datasets is commendable. My main concern is the received-power metric in Eq. (18), which appears to be an incorrect incoherent formula and is the basis for all numerical results. This is fixable in revision, but the simulations must be rerun with a coherent magnitude-squared formula and the manuscript updated accordingly. I would be supportive after that correction and after the circular UE1 evidence is reframed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the paper's core idea—use finite-difference kinematic estimates from low-rate position fixes to predict UE position ahead of the latency budget and point the RIS accordingly—is simple, practical, and plausibly useful. The multi-scenario validation, including real car and drone datasets, is a genuine plus. But the received-power model in Appendix A, Eq. (18), is not a coherent power expression, and since no code or data is released, the headline numbers in Section IV cannot be trusted as printed.\n\nWhat's new: the observer equations are textbook finite differences, but assembling them into an OPS/OPA predictor for proactive RIS control, with a complexity breakdown and validation across diverse trajectories, is a legitimate engineering contribution. The complexity table is useful, and the external car/drone datasets are good evidence that the predictor works under real acceleration and aggressive maneuvering.\n\nThe soft spots, in order of severity. First, Eq. (18) sums terms of the form |Γ|² e^(2j(...)) with no outer magnitude squared. Coherent reflection should give power proportional to |Σ ...|², which for equal magnitudes scales as (MN)²; the printed formula gives an incoherent sum. If the simulator implements Eq. (18), the phase alignment has no array gain, and the reported dB losses (e.g., 0.14–0.60 dB vs 6.32–7.39 dB in Table IV) are artifacts of an incorrect metric. If the simulation uses a corrected coherent formula, the paper should say so. As it stands, the absence of code/data makes the quantitative claims unverifiable. This is a load-bearing flaw, not a typesetting nit.\n\nSecond, the reported averages have no confidence intervals or per-realization spread. Ten or one hundred realizations are fine, but the reader needs to see the variance, especially for the OPS-vs-OPA comparisons.\n\nThird, UE1 is generated using exactly the constant-acceleration model embedded in OPA, so the near-zero loss of OPA there is circular; the paper calls it a baseline, but it should be flagged as a self-consistency check rather than evidence.\n\nFourth, the smoothness assumption is the load-bearing premise, and the Gauss-Markov results honestly show degradation when it fails. That is a fair limitation, not a flaw.\n\nIf the power metric is corrected, the central idea may well hold; the predictor itself is sound. But as a preprint, the results need a major revision before I'd rely on them.\n\nRecommendation: send it to peer review. A serious referee should have caught Eq. (18). The idea is practical and the fix is likely straightforward. I would not cite the numbers until the formula is corrected and the simulation artifacts are released.\n\nWho it's for: researchers working on RIS control under mobility, especially those using GNSS-rate localization.","headline":"A simple kinematic predictor for RIS beamforming is worthwhile, but the printed received-power formula in Eq. (18) is not a coherent power metric, so the headline numbers are unverified.","tokens_in":17173,"tokens_out":5207,"would_cite":false,"duration_ms":53963,"reading_group":"maybe","serious_thinker":"unclear","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper claims a kinematic observer-predictor can convert low-rate noisy position measurements into accurate ahead-of-time user positions, keeping RIS beamforming losses close to those of ideal instantaneous control.","keywords":["reconfigurable intelligent surface","RIS beamforming control","kinematic observer","position prediction","low-rate localization","latency compensation","UE mobility","passive beamforming"],"falsifier":"Run the same RIS setup with a user that makes a sharp turn, sudden stop, or direction reversal within the localization interval (for example, during $T_M = 100$ ms), and measure received-power loss relative to ideal control. If the OPS/OPA loss rises to within a few dB of the naive benchmark in such a trajectory, the local-linearization premise is violated and the central claim overreaches.","tokens_in":16189,"feed_emoji":"📡","tokens_out":10026,"duration_ms":84634,"temperature":0.7,"pith_summary":"The paper proposes a physics-based kinematic observer-predictor to make reconfigurable-intelligent-surface (RIS) beamforming proactive rather than reactive. Given only low-rate and noisy position fixes of a mobile user, the observer estimates the user's speed and, optionally, acceleration, and the predictor extrapolates the user's position ahead by the accumulated localization and control delay. The RIS then steers its reflection phases using the predicted position instead of the outdated measured one. The central claim is that this nearly restores ideal instantaneous beamforming: in the tested car scenario, average received-power losses drop to about 0.14-0.60 dB, while a naive delayed update loses 6.32-7.39 dB. This matters because it offers a low-complexity, scalable way to compensate RIS control latency without requiring faster localization or a heavier control channel.","feed_headline":"Kinematic predictor trims smart-surface beamforming loss to 0.6 dB","feed_subtitle":"Low-rate, noisy position fixes plus a physics-based observer keep a smart reflector aimed at a moving user.","key_machinery":"The central machinery is a two-stage kinematic observer-predictor. The observer converts sparse noisy position samples $y[q]$ into velocity estimates, using $\\hat{\\mathbf{v}}[q]=(y[q]-y[q-1])/T_M$ for OPS or a three-point backward difference for OPA, plus the acceleration estimate $\\hat{\\mathbf{a}}[q]=(y[q]-2y[q-1]+y[q-2])/T_M^2$. The predictor extrapolates the user position ahead by the estimated accumulated delay $\\hat{T}_A[k]$: $\\hat{\\mathbf{r}}_{\\mathrm{rx}}[k]=y[q]+\\hat{\\mathbf{v}}[q]\\hat{T}_A[k]$ for OPS, with an added $\\tfrac{1}{2}\\hat{\\mathbf{a}}[q]\\hat{T}_A[k]^2$ term for OPA. The load-bearing physical assumption is that any nonlinear trajectory can be locally linearized over the localization interval $T_M$, making constant-speed or constant-acceleration kinematics valid between samples.","core_discovery":"The central claim is that the accumulated delay in RIS control, from low-rate localization sampling plus processing and RIS setup time, can be compensated by predicting where the user will be when the beamforming update actually takes effect. The paper demonstrates this with two estimators: OPS, which assumes locally constant speed, and OPA, which also estimates acceleration from a few consecutive position fixes. The predicted position is inserted into the phase rule $\\angle\\Gamma^*_{m,n} = \\mathrm{mod}(k_0(|\\mathbf{r}^t_{m,n}|+|\\mathbf{r}^r_{m,n}|)+\\pi, 2\\pi)-\\pi$, and the corresponding varactor voltages are read from a lookup table. Across constant-acceleration, realistic car-acceleration, aggressive-drone, and Gauss-Markov mobility scenarios, average power losses stay close to the ideal baseline under moderate noise, with the simpler speed-only OPS becoming more robust than OPA as localization noise grows.","pith_inferences":["We infer a natural closed-loop extension: periodic beam scans or channel-quality feedback could correct drift when the kinematic model's constant-acceleration assumption is violated by abrupt maneuvers.","The same delay-compensation logic could transfer to other latency-limited beam-steering problems, such as millimeter-wave or terahertz beam tracking with sparse position or angle updates.","The paper's performance surfaces suggest a joint design rule for the localization interval $T_M$ and the trajectory's smoothness, and possibly an analytic bound on the largest $T_M$ before predicted-position error exceeds a fraction of the beamwidth.","We infer that the noise-robustness of the speed-only mode might make the approach useful beyond the centimetre-level precise-point-positioning accuracy emphasized in the paper, extending to cheaper noisier localizers in low-mobility regimes."],"forward_implications":["Low-rate localization systems such as 1–20 Hz satellite positioning become usable for real-time RIS beamforming, because the predictor generates high-rate position references between samples.","The scheme compensates both the localization sampling delay and the RIS setup/control delay, so beamforming is computed for the moment the update truly takes effect.","Under localization noise, the speed-only OPS variant is more robust than the acceleration-based OPA variant, so the two modes offer a noise-dependent trade-off.","The observer and predictor stages add negligible computational cost compared with per-element phase computation, and they localize naturally at the RIS controller, supporting time-division multiplexing of multiple users.","Errors of tens of milliseconds in estimating the accumulated delay cost only a few tenths of a dB, so exact delay knowledge is not required."],"supporting_citations":[{"why":"Supplies the nonlinear systems result that arbitrary dynamics can be locally linearized over short intervals, which justifies the kinematic observer design.","marker":"[17]"},{"why":"Provides the transmission-line circuit model of a RIS cell that underlies the reflection coefficient and the phase rule in Eq. (3).","marker":"[18]"},{"why":"Provides the electromagnetic backscattering model used to compute received power at the UE in the simulations.","marker":"[19]"},{"why":"Supplies the varactor junction-capacitance model and device parameters used to build the voltage-to-phase lookup table.","marker":"[20]"},{"why":"Provides the RIS phase-design principle from which Eq. (3) is taken.","marker":"[21]"},{"why":"Provides the three-point backward finite-difference formula used by the OPA observer for velocity estimation.","marker":"[23]"},{"why":"Provides the acceleration estimation formula used by the OPA observer.","marker":"[24]"},{"why":"Supplies the real 0-100 km/h acceleration profiles of combustion and electric cars used in the UE2 scenario.","marker":"[25]"},{"why":"Supplies the aggressively piloted racing-drone trajectory dataset used as the UE3 stress scenario.","marker":"[27]"},{"why":"Supplies the Gauss-Markov mobility model used as the UE4 random-motion scenario.","marker":"[28]"}],"fun_headline_variants":["Kinematic predictor keeps RIS beamforming loss to 0.6 dB on the move","Predictive physics-based observer aims smart surfaces at fast movers","Low-rate location fixes plus kinematic predictor keep RIS on target","Predictive control for RIS: beamforming ahead of the user's motion","Predictive observer for RIS: low-rate fixes, high-rate aiming"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole scheme rests on the user's trajectory being smooth enough over the time between two position fixes that constant-speed or constant-acceleration extrapolation stays accurate; if the user turns sharply or jerks between samples, the predicted position points the beam to the wrong place and the power loss climbs toward the naive baseline.","fun_headline_variants_meta":{"raw":{"variants":["Kinematic predictor keeps RIS beamforming loss to 0.6 dB on the move","Predictive physics-based observer aims smart surfaces at fast movers","Low-rate location fixes plus kinematic predictor keep RIS on target","Predictive control for RIS: beamforming ahead of the user's motion","Predictive observer for RIS: low-rate fixes, high-rate aiming"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000698,"raw_usage":{"total_tokens":3163,"prompt_tokens":964,"completion_tokens":2199,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":580,"completion_tokens_details":{"reasoning_tokens":2106}},"tokens_in":580,"tokens_out":2199,"duration_ms":17174,"temperature":1.0,"reasoning_tokens":2106,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T17:09:25.228443+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same RIS setup with a user that makes a sharp turn, sudden stop, or direction reversal within the localization interval (for example, during $T_M = 100$ ms), and measure received-power loss relative to ideal control. If the OPS/OPA loss rises to within a few dB of the naive benchmark in such a trajectory, the local-linearization premise is violated and the central claim overreaches.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the nonlinear systems result that arbitrary dynamics can be locally linearized over short intervals, which justifies the kinematic observer design."},{"cited_title":"Circuit modelling of reflecting intelligent sur- faces,","cited_arxiv_id":null,"evidence_quote":"Provides the transmission-line circuit model of a RIS cell that underlies the reflection coefficient and the phase rule in Eq. (3)."},{"cited_title":"Electromagnetic model of reflective intelligent surfaces,","cited_arxiv_id":null,"evidence_quote":"Provides the electromagnetic backscattering model used to compute received power at the UE in the simulations."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the varactor junction-capacitance model and device parameters used to build the voltage-to-phase lookup table."},{"cited_title":"Reconfigurable intelligent surfaces: Principles and opportunities,","cited_arxiv_id":null,"evidence_quote":"Provides the RIS phase-design principle from which Eq. (3) is taken."},{"cited_title":"Three-point backward finite- difference method for solving a system of mixed hyperbolic—parabolic partial differential equations,","cited_arxiv_id":null,"evidence_quote":"Provides the three-point backward finite-difference formula used by the OPA observer for velocity estimation."},{"cited_title":"On the influence of velocity and acceleration estimators on a servopneumatic system behaviour,","cited_arxiv_id":null,"evidence_quote":"Provides the acceleration estimation formula used by the OPA observer."},{"cited_title":"The kinematic advantage of electric cars,","cited_arxiv_id":null,"evidence_quote":"Supplies the real 0-100 km/h acceleration profiles of combustion and electric cars used in the UE2 scenario."},{"cited_title":"Race against the machine: A fully-annotated, open-design dataset of autonomous and piloted high-speed flight,","cited_arxiv_id":null,"evidence_quote":"Supplies the aggressively piloted racing-drone trajectory dataset used as the UE3 stress scenario."},{"cited_title":"Betweenness centrality based dynamic source routing for flying ad hoc networks in marching formation,","cited_arxiv_id":null,"evidence_quote":"Supplies the Gauss-Markov mobility model used as the UE4 random-motion scenario."}],"review_version":2}