{"id":"34f45ea0-8e72-4546-ae67-3c29046683dc","arxiv_id":"2506.08739","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"high","formal_verification":"none","parameter_count":3,"one_line_summary":"A joint satellite-user EKF estimates timing advance and Doppler shift, but the Doppler model omits line-of-sight projection and the validation is a single uncompared simulation.","lead":"An extended Kalman filter that jointly tracks a low Earth orbit satellite and a fast-moving ground device is used to compute timing advance and Doppler shift for 6G satellite access. A generalist should read it because initial access in LEO networks needs timing and frequency corrections when both the satellite and the user are moving.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Doppler model in Eq. (27) uses full relative velocity instead of its line-of-sight projection; Fig. 2(c) matches the non-physical total-speed model, so the claimed Doppler accuracy is not established.","rationale":"The reader's weakest assumption correctly identifies the most load-bearing issue. The paper claims an accurate Doppler-shift estimate as a headline contribution, and the Doppler law in Section IV-B is the quantity being validated. Since the simulation in Fig. 2(c) uses values numerically consistent with f_T ||v_rel||/c rather than the radial projection, the validation is circular with respect to a physically wrong model. This is not a disagreement with consensus; it is a correctness failure of the core estimation target. The TA computation and the EKF machinery are more standard and could be salvaged, but the paper's stated scope includes Doppler estimation, so the concern is decisive. The proposed check is simple and settles the matter without relying on access to the proprietary simulation code beyond recomputing a physical formula. I found no need to invent a second, independent objection; the reader's stated assumption is the same load-bearing one.","tokens_in":9823,"tokens_out":6299,"duration_ms":75081,"concrete_test":"Recompute the Doppler trace for the Fig. 2(c) scenario from the same ephemeris using the radial-projection formula Delta_f = (f_T/c) * [(p_l-p_u) . (v_l-v_u)] / ||p_l-p_u||. If the 10.9 GHz trace is no longer nearly flat at 263.36 kHz but varies by tens of kilohertz and passes near zero at closest approach, Eq. (27) and the simulation use the full-speed model, settling the concern. Also verify whether the EKF measurement update uses this Doppler term; if so, rerun the filter with the corrected formula and compare the reported position error to [1.8166%, 0.5595%, 0.7725%].","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that the framework accurately computes timing advance and Doppler shift. The TA part is standard geometry, but the Doppler part rests on Eq. (27), which writes Delta_f = f_T (v_l - v_u)/c, i.e. the full relative velocity vector without projection onto the satellite-UE line of sight. The physical non-relativistic Doppler is Delta_f = (f_T/c) d(d_lu)/dt = (f_T/c) [(p_l-p_u) . (v_l-v_u)] / ||p_l-p_u||. Eq. (27) also equates a scalar time-difference derivative with a vector difference divided by c, so it is internally inconsistent. The simulation confirms the non-physical interpretation: Fig. 2(c) shows Doppler staying in the narrow bands 263.358-263.368 kHz at 10.9 GHz and 689.165-689.190 kHz at 28 GHz over 0-40 degrees elevation, which matches f_T ||v_rel||/c and not a radial Doppler curve that should vary strongly with geometry and pass near zero near closest approach. Because the reported 'true' Doppler is generated by this wrong law, the claimed agreement between estimated and true Doppler does not validate estimation of physical Doppler. If Doppler is used as an EKF measurement, the error also enters the measurement model and its Jacobian, so the position estimates could be affected.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes an extended Kalman filter (EKF) based recursive Bayesian framework to jointly estimate the position and velocity of a LEO satellite and a mobile user equipment (UE), and then to compute the timing advance (TA) and Doppler shift experienced by the UE. The satellite motion model includes the gravitational acceleration term and its Jacobian, while range and elevation angle are used as measurements. The authors report mean percentage errors of [1.8166% 0.5595% 0.7725%] in satellite position coordinates, show TA variation with time and elevation, and study Doppler shift at 10.9 GHz and 28 GHz. A short clock-drift analysis is also included for unsynchronized satellite and UE clocks.","tokens_in":10092,"tokens_out":4693,"duration_ms":59296,"significance":"If the framework were correct, it would provide a useful single recursive estimator for initial-access parameters in LEO non-terrestrial networks with a mobile UE, an application of current interest. The paper has real strengths: the satellite acceleration model in Eq. (5)-(11) is physically motivated, the TA geometry in Eq. (25)-(26) is standard and correct, and the central position-estimation claim is not circular because no parameters are fitted to the target outputs. The recursive Bayesian formulation and the visibility-aware geometry are appropriate for the problem. However, the Doppler model in Eq. (27) is physically incorrect, and the validation of Doppler accuracy uses exactly this incorrect model. Since accurate Doppler estimation is one of the two headline contributions, the significance of the paper is currently contingent on a substantial correction of the Doppler formulation and a re-run of the corresponding simulations.","major_comments":[{"comment":"The Doppler shift formula is physically incorrect and dimensionally inconsistent. The paper writes Δf(t)=f_T (v_l(t)-v_u(t))/c, where v_l and v_u are three-dimensional vectors, so the right-hand side is a vector while the left-hand side is a scalar. The correct non-relativistic Doppler shift is the derivative of the slant range: Δf(t) = (f_T/c) d(d_lu(t))/dt = (f_T/c) [(p_l(t)-p_u(t)) · (v_l(t)-v_u(t))] / ||p_l(t)-p_u(t)||. The chain of equalities in Eq. (27) also treats the time derivative of the slant-range difference as if it were the full relative velocity vector, which is only the radial component. This error is load-bearing because the paper's claimed accurate Doppler estimation rests on this equation.","section":"Section IV-B, Eq. (27)"},{"comment":"The simulation confirms that the Doppler model used is the non-physical full-velocity formula rather than the radial projection. At 10.9 GHz the plotted Doppler stays in the narrow band 263.358-263.368 kHz over 0-40 degrees elevation, close to f_T ||v_rel||/c, whereas the true radial Doppler should vary strongly with geometry and pass through zero near closest approach. Consequently, the agreement between estimated and 'true' Doppler in Fig. 2(c) only validates the wrong model; it does not establish that the EKF estimates the physical Doppler shift. The Doppler simulation must be regenerated with the radial Doppler law, and the resulting curves should be shown as a function of elevation and time, including the expected zero crossing.","section":"Section V, Fig. 2(c)"},{"comment":"The steady-state Riccati expression P∞ = (R/H^2)(F + sqrt(F^2 + H^2 Q/R)) is not the algebraic Riccati equation for the EKF. For the matrix-valued F in Eq. (13), the expression is formally undefined, and no scalar-matrix conversion is provided. Since the paper advertises a stability analysis through the Riccati equation, either the correct continuous- or discrete-time algebraic Riccati equation should be stated and solved, or the stability discussion should be removed or substantially revised. This issue does not directly affect the simulation results, but it is presented as a contribution.","section":"Section III-B, Eq. (22)"},{"comment":"The measurement model h(x_t) is nonlinear (range and elevation in Eqs. (14) and the definition of d_lu), but the EKF update equations (17)-(19) use a matrix H_t without ever providing the Jacobian of h with respect to the state vector. Replacing A_t by F_t in the linear Kalman equations does not supply the measurement Jacobian. Without H_t, the EKF implementation used to produce the results is not fully specified, and the claim that the framework is a standard EKF cannot be verified from the manuscript alone. The measurement Jacobian should be derived and reported.","section":"Section II-A / Section III-B"},{"comment":"The clock-drift analysis is a definitional identity rather than an estimation method. Equation (30) shows that Δτ̂_1 - Δτ_1 = t_12 ε_2 - t_11 ε_1 = α t + β, which is just algebra from the definitions in Eq. (28). The paper does not provide a procedure to estimate ε_1 and ε_2, nor does it connect the EKF output to these quantities. The statement that the proposed framework 'computes the clock drift' is therefore unsupported. This is an advertised contribution in the abstract and Section I-C, even though it is not central to the TA estimation.","section":"Section IV-B, Eqs. (28)-(30)"}],"minor_comments":[{"comment":"The text states that the UE is assumed to move with constant acceleration, but Eq. (12) shows v_u_i(t+1)=v_u_i(t), i.e., constant velocity. Please reconcile the wording with the model.","section":"Section II-A, Eq. (12)"},{"comment":"The initial condition is written as x0 ~ N(0, P_{1|0}); this notation is confusing because the initial mean is typically nonzero and the initial covariance is P_0, not P_{1|0}. Please clarify.","section":"Section III-A, Eq. (15)"},{"comment":"The vertical axis appears to list Doppler values in kHz with values such as 263.358, but the units label says 'KHz'; please use consistent SI notation and verify the scaling.","section":"Section V, Fig. 2(c)"},{"comment":"The mean percentage error is reported as [1.8166% 0.5595% 0.7725%], but the definition of 'percentage error' per coordinate is not stated. Please specify how the percentage is computed relative to the true coordinate magnitude or another reference.","section":"Section V, Fig. 2(a)"},{"comment":"The notation t11 and t12 is introduced without clear definition of the two time instants; the TDoA expression should be stated with explicit times, for example d_lu(t+Δt)-d_lu(t), to avoid ambiguity.","section":"Section IV-B, Eq. (27)"},{"comment":"There are several typographical issues, including 'Algebraic Riccatti' instead of 'Riccati', and a missing space in 'Constellation Technologies & Operations' in the author affiliation. These do not affect the technical content.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The paper has a clear and correct TA geometry and a standard EKF structure, but the Doppler model is the central weakness. I recommend revision rather than outright rejection because the radial Doppler correction is a localized change to Eq. (27) and the simulation section, and the position-estimation contribution may still be salvageable. However, the authors should be aware that if the corrected Doppler simulation does not show the expected strong elevation dependence and zero crossing near closest approach, the Doppler contribution would no longer be credible. The clock-drift and Riccati sections also need substantial tightening or removal."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The useful part of this paper is the state-augmentation: it takes the satellite-only ephemeris EKF from [7] and adds UE states and the gravitational acceleration in the satellite motion model. Constructing the Jacobian for that nonlinear acceleration is correct, and the timing-advance geometry in Eq. (26) is standard but clean. If the paper only claimed the joint EKF for position/velocity estimation, I'd call it a competent, incremental contribution.\n\nBut the Doppler half is not sound. Eq. (27) writes Delta_f = f_T (v_l - v_u)/c, using the full relative velocity vector rather than its projection onto the satellite-UE line of sight. A scalar time-derivative is also being equated to a vector difference divided by c, so the equation is internally inconsistent. The physical Doppler is (f_T/c) d(d_lu)/dt, which depends on the radial component and passes through zero near closest approach. Fig. 2(c) is the smoking gun: the reported Doppler stays in narrow bands around 263.36 kHz and 689.17 kHz over 0-40 degrees elevation, matching f_T ||v_rel||/c and not a radial projection. That means the \"true\" Doppler in the simulation is generated by the wrong law, so the claimed estimation accuracy for Doppler is not established.\n\nOther soft spots are smaller. Eq. (22) is an asserted steady-state Riccati expression with no derivation and unclear notation. The measurement Jacobian is never written out. The clock-drift analysis in Eq. (30) reduces to the definitional identity Delta_tau_hat - Delta_tau = t12*eps2 - t11*eps1 = alpha t + beta, which is not much of a result. Validation is a single in-house simulation with no baseline, no error bars, and no comparison to a standard EKF without UE states, so the reported mean percentage errors are hard to interpret.\n\nI don't think the reader's soundness score is unfair. The Doppler error is load-bearing because Doppler accuracy is half of the advertised contribution. But the paper is not incoherent; the TA part and the joint-state formulation are legitimate and the fix is identifiable: replace the Doppler model with the radial projection and rerun the simulation. If this crosses my desk, I'd send it to review with the expectation of major revision rather than desk-reject, because the topic is timely and the core EKF framework is otherwise usable. For your own work, I wouldn't cite the Doppler results until the model is corrected.","headline":"A reasonable EKF state-augmentation for joint LEO satellite-UE tracking, but the Doppler model is physically wrong and the validation can't support the central Doppler claim.","tokens_in":10662,"tokens_out":1993,"would_cite":false,"duration_ms":27286,"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":"The paper claims that one extended Kalman filter over the joint satellite-and-user state can estimate LEO satellite position to within about 1.8 percent mean error and compute timing advance and Doppler shift even when the user moves at…","keywords":["LEO satellite networks","timing advance","Doppler shift","extended Kalman filter","recursive Bayesian estimation","user mobility","non-terrestrial networks","satellite visibility"],"falsifier":"Compare the Doppler-versus-elevation curve from a real LEO pass with the paper's simulated curve: physical Doppler must cross zero near the moment of closest approach, whereas the paper's model stays near $f_T \\|v_l-v_u\\|/c$ and does not drop to zero, so a measured null would settle which formula is right.","tokens_in":9565,"feed_emoji":"🛰️","tokens_out":9284,"duration_ms":97880,"temperature":0.7,"pith_summary":"The paper argues that timing advance and Doppler shift in a LEO satellite network can be estimated together by modelling the satellite and the ground user as a single joint system and filtering that system with an extended Kalman filter. The satellite motion includes the gravitational acceleration of a central force, the user moves with a constant velocity, and the measurements are slant range and elevation angle. From the filtered state the paper computes timing advance as twice the slant range divided by the speed of light, $\\text{TA}(t)=2 d_{lu}(t)/c$, and Doppler shift from the satellite-user relative velocity scaled by the carrier frequency. Simulations show the filter tracking the satellite position with mean percentage error around [1.8166% 0.5595% 0.7725%] even at airplane-speed user motion. If the approach holds, it gives one recursive estimator for the initial-access parameters that LEO networks currently have to compute through separate steps.","feed_headline":"One filter tracks LEO satellite plus fast user for timing and Doppler","feed_subtitle":"One recursive estimator holds satellite position error near 1.8 percent at airplane speeds and derives both parameters.","key_machinery":"The central object is the joint discrete-time state $x_t = [p_l, v_l, p_u, v_u]$, propagated with gravitational acceleration $a_l = -\\mu p_l/\\|p_l\\|^3$ for the satellite and constant velocity for the user. The argument is carried by the Jacobian $F(t)$, a block matrix with identity and $\\Delta t$ blocks plus the gravity-gradient block $A$, which linearizes the nonlinear motion for the extended Kalman filter. Slant range and elevation angle form the measurement model, and the formulas $\\text{TA}(t)=2 d_{lu}(t)/c$ and $\\Delta f(t)=f_T (v_l-v_u)/c$ translate the filtered state into the two communication parameters. The time-difference-of-arrival relations then turn the same state estimate into a clock-drift estimate when the clocks differ.","core_discovery":"The paper's central claim is that the position and velocity of a LEO satellite and a mobile user can be concatenated into one state and estimated recursively, with the satellite acceleration linearized through the Jacobian of the inverse-square gravitational field. From that state, the timing advance follows directly as the round-trip light time to the estimated slant range, and the Doppler shift follows as $f_T (v_l-v_u)/c$, the transmitted frequency times the relative velocity divided by the speed of light. The paper reports mean percentage position errors of [1.8166% 0.5595% 0.7725%] for the satellite coordinates in simulation, and shows the estimates staying close to the true values across the satellite visibility window. It further claims that when the satellite and user clocks are not synchronized, the clock drift takes the form $\\alpha t + \\beta$, so the same filter can separate clock offset from the time difference of arrivals.","pith_inferences":["A natural extension the paper does not pursue is to replace the full-relative-velocity Doppler expression with its line-of-sight projection, which would let the filter track the Doppler null near closest approach and be checked directly against real satellite beacon measurements.","The same joint-state filter could be extended to multiple visible satellites, allowing the user position to be estimated from several range and elevation measurements instead of assuming a known user trajectory.","The linear clock-drift model suggests a cheap synchronization scheme: estimate $\\alpha$ and $\\beta$ online from successive TDoA measurements and feed them back to the user's local oscillator."],"forward_implications":["When the filter converges, satellite position error stays near the reported mean percentage values and timing advance can be read directly from the estimated slant range.","Timing advance varies with elevation angle, reaching a minimum at maximum elevation and a maximum near the visibility boundary, so a network can predict when to refresh the TA during a pass.","Doppler shift scales with carrier frequency, giving roughly 260 kHz at 10.9 GHz and 680 kHz at 28 GHz in the simulation, so the same state estimate supports frequency precompensation in either band.","The clock-drift expression $\\Delta t_{\\text{clock drift}} = \\alpha t + \\beta$ lets the filter estimate user-satellite clock offset from time difference of arrivals without extra hardware."],"supporting_citations":[{"why":"Supplies the EKF-based satellite ephemeris tracking approach that this work extends to joint satellite-user dynamics.","marker":"[7]"},{"why":"Provides the algebraic Riccati equation used to check steady-state stability of the linearized filter.","marker":"[10]"},{"why":"Supplies the gravitational potential model $a_l = -\\mu p_l/\\|p_l\\|^3$ used for LEO satellite acceleration.","marker":"[9]"},{"why":"Gives the geodetic-to-geocentric coordinate conversion used to place the user in earth-centered coordinates.","marker":"[8]"},{"why":"Presents a location-based timing advance baseline that motivates accounting for satellite motion.","marker":"[4]"},{"why":"Presents prior timing advance estimation that assumes one position is known, which the EKF replaces with a joint estimate.","marker":"[6]"}],"fun_headline_variants":["EKF estimates timing and Doppler for LEO satellite links","One Bayesian filter derives timing and Doppler for fast users","Joint EKF yields timing and Doppler for LEO satellite networks","Recursive filter handles satellite motion and user speed for timing and Doppler"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that Doppler shift depends on the total relative speed between satellite and user, not just the speed along the line connecting them; physical Doppler is the line-of-sight component, and that component is zero at closest approach.","fun_headline_variants_meta":{"raw":{"variants":["EKF estimates timing and Doppler for LEO satellite links","One Bayesian filter derives timing and Doppler for fast users","Joint EKF yields timing and Doppler for LEO satellite networks","Recursive filter handles satellite motion and user speed for timing and Doppler"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000755,"raw_usage":{"total_tokens":3393,"prompt_tokens":1019,"completion_tokens":2374,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":635,"completion_tokens_details":{"reasoning_tokens":2305}},"tokens_in":635,"tokens_out":2374,"duration_ms":20517,"temperature":1.0,"reasoning_tokens":2305,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T05:03:52.949052+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compare the Doppler-versus-elevation curve from a real LEO pass with the paper's simulated curve: physical Doppler must cross zero near the moment of closest approach, whereas the paper's model stays near $f_T \\|v_l-v_u\\|/c$ and does not drop to zero, so a measured null would settle which formula is right.","supporting_citations":[{"cited_title":"Ephemeris tracking and e rror prop- agation analysis of LEO satellites with application to oppo rtunistic navigation,","cited_arxiv_id":null,"evidence_quote":"Supplies the EKF-based satellite ephemeris tracking approach that this work extends to joint satellite-user dynamics."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the algebraic Riccati equation used to check steady-state stability of the linearized filter."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the gravitational potential model $a_l = -\\mu p_l/\\|p_l\\|^3$ used for LEO satellite acceleration."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the geodetic-to-geocentric coordinate conversion used to place the user in earth-centered coordinates."},{"cited_title":"Location-based timing advance estimation for 5G integrated LEO satellite communications,","cited_arxiv_id":null,"evidence_quote":"Presents a location-based timing advance baseline that motivates accounting for satellite motion."},{"cited_title":"Timing advance estimation in low earth orbit satellite networks,","cited_arxiv_id":null,"evidence_quote":"Presents prior timing advance estimation that assumes one position is known, which the EKF replaces with a joint estimate."}],"review_version":1}