{"id":"5c2064cf-55fa-49b1-9550-53da171a27de","arxiv_id":"2505.16350","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A joint handover criterion that triggers when either RSRP or ISAC-based distance estimates exceed thresholds reduces the handover region length by 75.20% and increases activation probability by 76.31% in the paper's model.","lead":"This paper proposes adding radar-like distance measurements from ISAC signals to the usual signal-strength rule for deciding when a drone should switch base stations. A combined rule triggers a handover earlier and in a narrower geographic region, which could reduce failed or unnecessary handovers for cellular-connected drones.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The independence assumption in Eq. (10) is not merely optimistic; it is inconsistent with the paper's own model, where the same shadow-fading term enters both the RSRP event and the sensing CRLB, so the 76.31% improvement is not derived.","rationale":"The reader's weakest assumption correctly identifies independence of the RSRP and distance events as a key vulnerability. I agree and sharpen it: the paper's own signal model ties the two events through a shared shadow-fading term, making the independence assumption not just unvalidated but internally inconsistent. This is more load-bearing than the Gaussian CRLB-achieving error model, because even a perfect Gaussian estimator would not restore Eq. (10) if the events share common randomness. The suggested Monte Carlo test would settle whether the correlation materially changes the reported 76.31% improvement. Since the reader's verdict is already CONDITIONAL and this concern reinforces that conditionality without overturning the qualitative contribution, no change in verdict is needed. The CRLB factor-of-two issue is a separate correctness flaw, but it is secondary because correcting it would likely make the distance estimate sharper and thus would not weaken the central qualitative claim; the independence issue is the one that could invalidate the headline number.","tokens_in":1046,"tokens_out":1074,"duration_ms":108419,"concrete_test":"Run a Monte Carlo simulation that samples shadow fading values, generates RSRP per Eq. (8), and generates distance estimates using the CRLB-based Gaussian model with the same shadow fading that enters the SNR, over the same grid and altitudes as Figs. 5 and 6. Compute the empirical union probability P[(PT - PS > Gamma) OR (d_hat_S - d_hat_T > d_th)] and the average improvement over the RSRP-only criterion. If the empirical improvement differs materially from the reported 76.31%, or if Eq. (10) fails at representative points, the independence assumption is load-bearing and the headline percentages are not validated.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing step is the factorization in Eq. (10), P_joint = P_RSRP + P_dist - P_RSRP * P_dist, which requires the RSRP-activation event and the distance-activation event to be independent. Footnote 3 justifies this by citing distinct physical origins, but the model itself contradicts it. In Section II.B, the sensing channel gain and the communication gain share the same shadow-fading term (epsilon). The RSRP event in Eq. (8) is driven by epsilon_T - epsilon_S, while the distance-estimation variance CRLB(d) in Eq. (7) is inversely proportional to the per-subcarrier SNR gamma, which depends on the same epsilon_S and epsilon_T through the sensing path gain. Thus the distance-estimate error and the RSRP fluctuation are statistically dependent, both through the common geometry and through the shared shadowing. The product form in Eq. (10) therefore does not follow from the stated model; it is an independent assumption that directly controls the reported 76.31% activation-probability improvement. If the two events are positively correlated, the true union probability is smaller than Eq. (10) and the headline improvement shrinks. The CRLB in Eq. (7) also has a factor-of-two error (the likelihood in Eq. (5) omits the factor 2 for complex Gaussian noise), so the quantitative results in Eqs. (8)-(10) are not self-consistent even before correlation is considered.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes a sensing-enhanced handover (HO) activation criterion for cellular-connected drones, fusing the standard RSRP-based A3 event with an ISAC-based distance comparison. The authors derive a Cramér-Rao lower bound (CRLB) for round-trip distance estimation in an OFDM ISAC link, define the joint HO probability as the union of the two triggering events under an independence assumption, and evaluate the resulting HO region length and activation probability through numerical evaluation of the analytic formulas. For SNR ≥ 0 dB and 20% sensing pilot ratio, they report a 75.20% reduction in average HO region length and a 76.31% improvement in activation probability relative to the RSRP-only baseline.","tokens_in":9523,"tokens_out":10835,"duration_ms":86941,"significance":"The topic is timely given the growing interest in low-altitude wireless networks and ISAC, and the paper provides a transparent analytic framework with closed-form performance metrics and 3GPP-aligned parameter settings. The main strength is the explicit connection between sensing accuracy (CRLB) and handover decision statistics, together with a complexity assessment for the additional distance estimation. However, the headline quantitative claims are not robust: the activation-probability improvement is structurally guaranteed by defining the joint criterion as a union, and the reported magnitude rests on an independence assumption that is inconsistent with the paper's own shadow-fading model and on a CRLB formula that appears to contain a factor-of-two error. If these derivation issues are corrected, the framework could be a useful starting point for sensing-assisted HO design, but the current numerical claims are not supported as stated.","major_comments":[{"comment":"The factorization P_joint = P_RSRP + P_dist - P_RSRP*P_dist in Eq. (10) requires the RSRP-triggering event and the distance-triggering event to be independent. In the model of Section II.B, the sensing path gain √β_S shares the same shadow-fading term ε as the communication gain, and γ in Eq. (7) depends on β_S. Consequently, the CRLB, and therefore the distribution of the distance estimate in Eq. (9), depends on the same random shadowing that drives the RSRP difference in Eq. (8). The two events are statistically dependent, so the footnote's appeal to 'distinct physical origins' is not supported by the model. For positive correlation the true union probability is smaller than Eq. (10), and the reported 76.31% improvement is not a consequence of the stated model. Please re-derive the joint probability from the full model (averaging over ε) or provide an explicitly justified independence model.","section":"Section III, Eq. (10) and Footnote 3"},{"comment":"The likelihood in Eq. (5) is written for a real Gaussian observation: the exponent lacks the complex modulus and the normalization corresponds to a scalar real Gaussian. The received signal r_{m,n} is complex, and for circularly symmetric complex Gaussian noise the Fisher information for a deterministic complex signal contains a factor of 2 that is missing in Eq. (6). As a result, the CRLB in Eq. (7) is a factor of two larger than the correct bound. Since Eq. (9) uses CRLB(d_S)+CRLB(d_T) as the distance-error variance, this error directly inflates P_dist and distorts all reported improvements and HO-region lengths. Please correct the likelihood and Fisher information derivation and recompute the numerical results.","section":"Section II.C, Eqs. (5)-(7)"},{"comment":"The statement that the joint criterion improves the HO activation probability by 76.31% is, by construction, a consequence of taking the union of two events: P_joint ≥ P_RSRP whenever P_dist > 0, regardless of any physical mechanism. The existence of an improvement is therefore not a simulation finding but a mathematical identity. The manuscript should not present it as a surprising result; instead, the paper should clearly state that the contribution lies in the magnitude and parameter sensitivity of the improvement, and should focus the analysis on the assumptions (independence, CRLB-achieving estimator) that control that magnitude.","section":"Section IV.B.5 / Fig. 6 and Abstract"}],"minor_comments":[{"comment":"The exponent in the likelihood should use the squared modulus |r_{m,n} - α a_{m,n} e^{-jA_n τ_S}|^2, and the normalization should be that of a complex Gaussian (e.g., (πσ^2)^{-1}); the current expression is not valid for complex data.","section":"Eq. (5)"},{"comment":"The reported D_HO ranges in Fig. 2 are not accompanied by the y-coordinate or the averaging procedure; please specify whether these are for y=0 or averaged over y.","section":"Section IV.B.1 / Fig. 2"},{"comment":"Reference [8] appears before [15] in the introduction, and the bibliography ordering is scrambled (for example, [15] is placed after [7] and before [8]). Please reorder the references so that citations appear in numerical order.","section":"References"},{"comment":"The statement that the LoS probability approaches 100% at 100-300 m altitude is asserted without a specific reference; if TR 36.777 is the source, please cite the relevant section or annex.","section":"Section II.B"},{"comment":"The effective data rate R_eff is defined using P_HO as a function of position, but the spatial map in Fig. 7 is not accompanied by an explanation of how the 'maximum data rate difference' is computed or whether it is averaged over a drone trajectory or user distribution; please clarify.","section":"Section IV.B.6 / Eq. (11)"},{"comment":"The paper uses the term 'simulation' to describe what appears to be numerical evaluation of the closed-form expressions (8)-(10); if no Monte Carlo simulation is performed, please state explicitly that the curves are analytic evaluations.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The paper presents a natural and incremental idea—combining RSRP and distance-based triggers for handover—and the quantitative claims are not currently reliable because of the independence issue and the CRLB error. The improvement in activation probability is structurally guaranteed, so the paper's real contribution would be a careful, self-consistent quantification of the improvement under realistic dependencies. If the authors can correct the derivations and re-run the evaluations, the paper could become suitable for publication. I saw no sign of misconduct; the citation list is heavily self-referential, but that is common in this area."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThis is a legitimate but modest extension: an OR-rule handover trigger that fires when either the RSRP A3 condition or an ISAC-based distance difference exceeds its threshold. The paper derives analytic activation probabilities and shows the OR rule shortens the handover region. That is a real, useful idea for drone mobility, and the authors lay out the geometry and signal model clearly. Eqs (8) and (9) follow from the log-normal shadowing and Gaussian distance-error assumptions, and the complexity discussion is honest.\n\nThe problems are in the load-bearing assumptions. Eq (10) factorizes the joint activation probability as P_RSRP + P_dist - P_RSRP*P_dist, which requires independence between the RSRP event and the distance event. Footnote 3 justifies this by 'distinct physical origins,' but the model itself shares the same shadow-fading term ε between the communication and sensing links, and the distance CRLB in Eq (7) depends on the sensing SNR, which depends on ε. So the two events are statistically dependent. The product form is an extra assumption, not a derivation. If the events are positively correlated, the union probability is overestimated and the 76.31% improvement shrinks. The paper does not provide a Monte Carlo check or a sensitivity analysis around this, so the headline number is not established.\n\nThere is also a concrete error: Eq (5) writes the likelihood as a real Gaussian for complex observations. The correct complex Gaussian density leads to a Fisher information that is twice as large, so the CRLB in Eq (7) is high by a factor of 2. That is fixable and actually conservative—a corrected CRLB makes the distance trigger sharper—but it needs to be stated.\n\nThe figures are numerical evaluations of the closed-form formulas, not simulation in the sense of drawing random realizations. That is fine for validating the math, but it does not validate the model assumptions.\n\nThis paper is for readers working on ISAC-assisted mobility management. It is a plausible first step, not a breakthrough. The idea is worth referee time, but the corrected version should include a proper CRLB, a sensitivity analysis of the independence assumption, and at least one Monte Carlo validation before the percentages are taken at face value.\n\nRecommendation: send to peer review, with major revision required.","headline":"A sensible OR-rule handover idea with clean derivations, but the headline improvements rest on an independence assumption that the paper's own model contradicts, plus a factor-of-two CRLB error.","tokens_in":10082,"tokens_out":7247,"would_cite":false,"duration_ms":56921,"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 joint handover criterion that combines RSRP with ISAC distance estimates reduces the handover region length by 75.20% and improves activation probability by 76.31% in drone networks.","keywords":["low-altitude wireless networks","handover","integrated sensing and communication","RSRP","Cramér–Rao lower bound","OFDM sensing","drone cellular networks","handover activation probability"],"falsifier":"Run the paper's geometry in a Monte Carlo simulation where the shadow-fading term and the distance-estimation error are drawn from a joint Gaussian with a positive correlation coefficient, and compare the empirical handover activation probability with the formula $P_{\\mathrm{RSRP}} + P_{\\mathrm{dist}} - P_{\\mathrm{RSRP}} P_{\\mathrm{dist}}$. A correlation of even 0.3 across the handover region would reduce the joint probability enough to erase a substantial part of the claimed 76.31% improvement; a field test with a real ISAC base station and a drone at 100–300 m altitude, using measured distance-error statistics instead of the CRLB, would settle which regime holds.","tokens_in":8976,"feed_emoji":"📡","tokens_out":7902,"duration_ms":60064,"temperature":0.7,"pith_summary":"The paper asks whether a cellular drone network can make handovers more reliable by letting the network's own sensing measurements, not just the received signal strength, decide when to switch base stations. It models an ISAC base station that estimates the drone's distance from its echo signal, derives the Cramér–Rao lower bound for that distance estimate, and proposes a joint handover criterion: trigger a handover if either the RSRP of the target base station exceeds the serving base station by a hysteresis margin, or the estimated distance to the serving base station exceeds the distance to the target by a threshold. In simulation over the paper's ground-to-air geometry, the joint criterion reduces the average handover region length by 75.20% and improves the handover activation probability by 76.31% relative to RSRP-only operation at $\\mathrm{SNR}\\ge 0$ dB with a 20% sensing pilot ratio. The reason to care is that low-altitude drones move in three dimensions and suffer frequent redundant or failed handovers under RSRP-only rules, and this paper gives a concrete mechanism, with a parameter-derived error model, for why sensing distance should fix part of that problem.","feed_headline":"Sensing plus signal strength shrinks drone handover zone 75%","feed_subtitle":"Adding ISAC distance estimates to RSRP cuts handover region length by 75% and boosts activation probability by 76%.","key_machinery":"The load-bearing object is the joint handover criterion itself: an OR-combination of a signal-strength event and a sensing-distance event. The RSRP event captures the traditional A3 rule through the $Q$-function of a shadow-fading contrast; the distance event captures the spatial geometry of the drone relative to the two base stations through ISAC range estimates. What carries the argument is the Cramér–Rao lower bound $\\mathrm{CRLB}(d) = \\frac{3c^2}{8\\pi^2\\gamma \\Delta f^2 M \\rho N (\\rho N - 1)(2\\rho N - 1)}$, which turns the sensing pilot ratio, bandwidth, SNR, and OFDM resource count into a concrete error variance for the distance trigger, and the independence assumption that lets the two events combine as $P(A\\cup B)=P(A)+P(B)-P(A)P(B)$.","core_discovery":"The central claim is that a handover trigger based on the union of two events outperforms the conventional RSRP-only trigger in the low-altitude drone setting. The first event is the standard A3 condition, $P_T - P_S > \\Gamma$, whose activation probability under shadow fading is $P_{\\mathrm{HO}}^{\\mathrm{RSRP}} = Q\\!\\left(\\frac{\\Gamma + L(d_T) - L(d_S)}{\\sqrt{2}\\,\\sigma_{\\mathrm{SF}}}\\right)$. The second event uses ISAC distance estimates $\\hat d_S$, $\\hat d_T$ modeled as Gaussian with variance equal to the Cramér–Rao lower bound of Eq. (7), giving $P_{\\mathrm{HO}}^{\\mathrm{dist}} = Q\\!\\left(\\frac{d_{\\mathrm{th}} + d_T - d_S}{\\sqrt{\\mathrm{CRLB}(d_T) + \\mathrm{CRLB}(d_S)}}\\right)$. The joint criterion fires when either condition holds, so its activation probability is $P_{\\mathrm{HO}}^{\\mathrm{joint}} = P_{\\mathrm{HO}}^{\\mathrm{RSRP}} + P_{\\mathrm{HO}}^{\\mathrm{dist}} - P_{\\mathrm{HO}}^{\\mathrm{RSRP}} P_{\\mathrm{HO}}^{\\mathrm{dist}}$, under the paper's stated independence assumption. The paper reports that this union rule shortens the spatial handover region by 75.20% on average and raises activation probability by 76.31% at $\\mathrm{SNR}\\ge 0$ dB with sensing pilot ratio 20%, while retaining most of the gain at lower SNR and lower pilot ratios.","pith_inferences":["The same union-of-triggers logic could be applied to other ISAC outputs, such as Doppler or angle estimates, which would matter for non-line-of-sight or maneuvering flight where distance alone is less informative.","If real estimators do not achieve the CRLB, the distance branch contributes less than the paper's model says, so the 76.31% figure is best read as an upper bound for the activation-probability gain.","The paper's fixed-straight-line trajectory leaves ping-pong handovers out of scope; adding a time-to-trigger or hysteresis to the distance event would be a direct test of whether the joint criterion also reduces unnecessary back-and-forth handovers.","A field measurement of the correlation between RSRP shadowing and ISAC distance errors would tell whether the independence assumption is safe or whether the joint formula needs a covariance term."],"forward_implications":["At $\\mathrm{SNR}\\ge 0$ dB and a 20% sensing pilot ratio, the joint criterion reduces the average handover region length by 75.20% and raises activation probability by 76.31% compared with RSRP-only.","The joint criterion keeps its advantage at lower SNR because the RSRP branch remains active when sensing accuracy degrades, so the OR rule degrades more slowly than sensing alone.","The lower boundary of the handover region stays near $x = -13$ m as the distance threshold $d_{\\mathrm{th}}$ ranges from 0 to 100 m, making the criterion less sensitive to threshold tuning than the RSRP-only rule.","Inside the handover region the joint criterion improves the effective data rate by up to 2.67 Mbps, supporting latency-sensitive drone services such as real-time video and control signaling.","The added complexity of distance estimation, based on a joint angle-range-velocity estimation method, is $O[(N_t + \\log(MN) + g)MN]$, which the paper argues is comparable to conventional OFDM transceiver operations."],"supporting_citations":[{"why":"Defines the 3GPP A3 RSRP handover event that is the baseline criterion the paper extends.","marker":"[11]"},{"why":"Supplies the RSRP activation probability expression in Eq. (8) that the joint criterion builds on.","marker":"[13]"},{"why":"Provides the 3GPP UMa-AV LoS path loss model used for the ground-to-air channel.","marker":"[19]"},{"why":"Supplies the radar cross section and sensing path loss relation used in the ISAC signal model.","marker":"[18]"},{"why":"Source for the Fisher information and Cramér–Rao lower bound derivation for delay estimation.","marker":"[21]"},{"why":"The joint estimation method whose complexity estimate is cited to show the added cost of distance estimation is manageable.","marker":"[22]"},{"why":"The OFDM waveform and channel estimation context that motivates the ISAC signal model.","marker":"[16]"}],"fun_headline_variants":["ISAC + RSRP handover cuts drone zone by 75%","Sensing + signal strength slashes drone handover zone 75%","Drone handover 75% shorter with ISAC-assisted RSRP","Fusing ISAC and RSRP shrinks drone handover zone 75%","ISAC sensing plus RSRP improves drone handover 76%"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the signal-strength measurement and the sensing-distance estimate fail independently, so the probability that either one triggers a handover is the simple sum minus the product; if the two errors are positively correlated, the reported activation-probability gain shrinks, and the paper's Gaussian, CRLB-achieving error model also makes the distance estimate look more accurate than real estimators usually are.","fun_headline_variants_meta":{"raw":{"variants":["ISAC + RSRP handover cuts drone zone by 75%","Sensing + signal strength slashes drone handover zone 75%","Drone handover 75% shorter with ISAC-assisted RSRP","Fusing ISAC and RSRP shrinks drone handover zone 75%","ISAC sensing plus RSRP improves drone handover 76%"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000792,"raw_usage":{"total_tokens":3586,"prompt_tokens":1136,"completion_tokens":2450,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":752,"completion_tokens_details":{"reasoning_tokens":2349}},"tokens_in":752,"tokens_out":2450,"duration_ms":14992,"temperature":1.0,"reasoning_tokens":2349,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T15:03:13.842714+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the paper's geometry in a Monte Carlo simulation where the shadow-fading term and the distance-estimation error are drawn from a joint Gaussian with a positive correlation coefficient, and compare the empirical handover activation probability with the formula $P_{\\mathrm{RSRP}} + P_{\\mathrm{dist}} - P_{\\mathrm{RSRP}} P_{\\mathrm{dist}}$. A correlation of even 0.3 across the handover region would reduce the joint probability enough to erase a substantial part of the claimed 76.31% improvement; a field test with a real ISAC base station and a drone at 100–300 m altitude, using measured distance-error statistics instead of the CRLB, would settle which regime holds.","supporting_citations":[{"cited_title":"NR; Radio Resource Con- trol (RRC); Protocol specification (Release 16),","cited_arxiv_id":null,"evidence_quote":"Defines the 3GPP A3 RSRP handover event that is the baseline criterion the paper extends."},{"cited_title":"A Trajectory Prediction based Intelligent Handover Control Method in UA V Cellular Networks,","cited_arxiv_id":null,"evidence_quote":"Supplies the RSRP activation probability expression in Eq. (8) that the joint criterion builds on."},{"cited_title":"Technical Report TR 36.777: Enhanced LTE Support for Aerial Vehicles,","cited_arxiv_id":null,"evidence_quote":"Provides the 3GPP UMa-AV LoS path loss model used for the ground-to-air channel."},{"cited_title":"A Cluster-based Statistical Channel Model for Integrated Sensing and Communication Channels,","cited_arxiv_id":null,"evidence_quote":"Supplies the radar cross section and sensing path loss relation used in the ISAC signal model."},{"cited_title":"Mensing,Location Determination in OFDM based Mobile Radio Systems","cited_arxiv_id":null,"evidence_quote":"Source for the Fisher information and Cramér–Rao lower bound derivation for delay estimation."},{"cited_title":"A Novel Joint Angle-Range-Velocity Estimation Method for MIMO-OFDM ISAC Sys- tems,","cited_arxiv_id":null,"evidence_quote":"The joint estimation method whose complexity estimate is cited to show the added cost of distance estimation is manageable."},{"cited_title":"Channel Spreading Function-Inspired Channel Transfer Function Estimation for OFDM Systems with High-Mobility,","cited_arxiv_id":null,"evidence_quote":"The OFDM waveform and channel estimation context that motivates the ISAC signal model."}],"review_version":1}