{"id":"9e3de7d0-1755-4558-ab93-a6a79a97ca98","arxiv_id":"2411.17334","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A semi-implicit Euler discretization of the single-track vehicle model gives explicit lateral-velocity and yaw-rate updates whose denominators no longer vanish at zero speed.","lead":"This paper proposes a discrete-time vehicle model that stays stable at very low speeds, where standard dynamic models blow up. The model keeps the explicit update form that control algorithms like MPC need, and it is tested in simulation and on a real vehicle.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The stability theorem is only as strong as the unproven grid check of Condition 1; the proof's strict-inequality handling of ||A*||=1 must be clarified before the guarantee can be called assured.","rationale":"The reader's weakest assumption already identified Condition 1 as the load-bearing unproven step and flagged the ||A*|| = 1 boundary flaw. My independent read of equations (16)-(28) confirms both issues, and adds the concrete observation that the denominators in (11) are linear in U and vanish at positive U for the negative stiffness values used in the paper, so the domain of validity of the explicit model is not the all-U-away-from-zero set implied by the abstract. This makes the 'assured numerical stability' claim conditional on an explicit operating region U in [0, U_max] excluding the zero-crossing speeds, plus an additional condition that ||Â_k(τ_k)|| < 1, not <= 1. These are fixable with a clarified theorem statement and a denser numerical check; the construction itself (explicit update avoiding U in the denominator alone) is sound and the empirical comparisons are informative. Hence the verdict remains CONDITIONAL, in agreement with the reader.","tokens_in":24824,"tokens_out":1837,"duration_ms":15233,"concrete_test":"Recompute ||Â_k(τ_k)|| on a fine grid of U in [0, 25] m/s for the Table 1 prototype at T_s = 0.1 s, and also compute the smallest positive U where the denominators in (11), i.e., mU - T_s(k_f + k_r) and I_z U - T_s(l_f^2 k_f + l_r^2 k_r), are zero. If any such zero-crossing U lies inside a regime where the model is claimed to be stable, then Condition 1 cannot hold there and the proposition must be restated with an explicit domain restriction. Then re-run Proposition 5.3 with ||A*|| = 1 and check whether the accumulated perturbation bound grows like k ||b*||; if it does, the theorem holds only for ||A*|| < 1.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim is 'assured numerical stability.' In the proof of Proposition 5.3, equation (24) bounds the accumulated error block term by ||b*|| (1 - ||A*||^k)/(1 - ||A*||). The geometric series manipulation assumes ||A*|| < 1; the paper then separately handles ||A*|| = 1 by asserting ||b_{k-1..0}|| <= ||b*||. This assertion is non-sequitur: the correct bound from (23) is ||b_{k-1..0}|| <= k ||b*||, which grows linearly with k and is not a uniform bound. The proof itself explicitly flags this as an unaddressed scenario, so Proposition 5.3 is only established for strict inequality ||Â_k(τ_k)|| < 1, not for Condition 1 as stated. Separately, Condition 1 is never derived from the model parameters; it is only checked numerically for one CarSim prototype (Table 1), three step sizes, and speeds 0-25 m/s (Figure 3). The paper's own Equation (12) shows the denominator scales of the V,omega updates are mU - T_s(k_f + k_r) and I_z U - T_s(l_f^2 k_f + l_r^2 k_r); because k_f, k_r are negative for a conventional vehicle, these denominators are strictly negative at U = 0 but pass through zero at positive U = T_s(k_f + k_r)/m and U = T_s(l_f^2 k_f + l_r^2 k_r)/I_z. At those singular speeds the update itself is undefined regardless of the norm check. And the Jacobian Â_k entry denominators in (18)-(19) also vanish there. The numerical check deliberately restricts speeds to 0-25 m/s and does not report whether these zero-crossing speeds lie inside the claimed operating domain. If the domain is meant to exclude them, the operating envelope must be stated explicitly in Condition 1 and Proposition 5.3.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper proposes an explicit discrete-time dynamic single-track vehicle model. The discretization uses forward Euler for position, heading, and longitudinal speed, and a backward-Euler-inspired but algebraically solved update for lateral speed and yaw rate, yielding the closed-form map (12). The main claims are (i) the model avoids the low-speed singularity that afflicts standard dynamic models because the denominators of the lateral updates contain cornering-stiffness terms in addition to U, and (ii) the model is numerically stable in the sense of bounded error propagation, under a sufficient condition (Condition 1) on the norm of a Jacobian submatrix. The paper proves that condition is sufficient, verifies it numerically for one CarSim prototype over 0–25 m/s and three step sizes, compares accuracy against a kinematic model and stability against forward-Euler discretization, and reports open-loop real-vehicle experiments.","tokens_in":25266,"tokens_out":10775,"duration_ms":99699,"significance":"If the stability guarantee were fully established, this would be a practically valuable contribution to model-based control at low speed and in stop-start scenarios. The paper has clear strengths: the explicit form is genuinely useful for MPC and rollout, the avoidance of a U-only denominator is a concrete design idea, the algebraic derivation of (12) is transparent, and the empirical comparison and real-vehicle tests are extensive. However, the central 'assured numerical stability' claim is currently only a conditional theorem, and the proof has a gap at the boundary of Condition 1. The numerical validation covers a single parameter set and only verifies the non-strict inequality, so the theoretical guarantee is not as strong as advertised.","major_comments":[{"comment":"In the proof of Proposition 5.3, the bound in (24) is obtained by summing a geometric series and is valid only when ||Â*|| < 1. The text then handles the case ||Â*|| = 1 via (26), asserting ||b_{k−1..0}|| ≤ ||b*||, but this does not follow from (23). The correct bound is ||b_{k−1..0}|| ≤ k||b*||, which is unbounded as k→∞. Consequently, Proposition 5.3 is not proved under Condition 1 as stated (≤ 1); it is proved only under the strict inequality ||Â_k(τ_k)|| < 1. Since the paper's title and abstract claim 'assured numerical stability', this gap is load-bearing and must be repaired, for example by strengthening Condition 1 to strict contraction with a margin and verifying that margin, or by supplying a refined argument for the equality case that rules out linear growth of the accumulated b-term.","section":"§5.3, Eq. (26)"},{"comment":"The empirical verification in Section 6.2 checks ||Â_k(τ_k)|| ≤ 1 for a single CarSim parameter set (Table 1), step sizes 0.001, 0.01, and 0.1 s, and longitudinal speeds in [0, 25] m/s. This does not close the gap identified in Proposition 5.3, because the proof requires the strict inequality < 1, while the verification reports only non-strict ≤ 1. Moreover, because only one vehicle/parameter set is tested, the general claim of 'assured numerical stability' for the proposed model is not established across parameter variations. The authors should either prove an analytic condition ensuring ||Â|| < 1 for a parameterized family, or explicitly restate the contribution as a model whose stability is verified conditionally/numerically, with the strict-inequality issue resolved.","section":"§6.2, Figure 3"}],"minor_comments":[{"comment":"In Lemma 5.2, the statement writes g'(m) = (f(d)−f(c))/(d−c); the right-hand side should use g, not f.","section":"§2.3 / Lemma 5.2"},{"comment":"The phrase 'strongly evidents' should read 'strongly evidences'.","section":"§1, Contributions (4)"},{"comment":"The description 'substituting local (vehicle-attach) coordinate for ground coordinate system' is unclear; please spell out why setting a→U̇ is justified.","section":"§4.1"},{"comment":"The sentence 'using the nature of the inverse proposition' is vague; the argument that bounded norms imply no infinite elements should be stated formally.","section":"§5.3"},{"comment":"Definition 5.1 uses lim_{k→∞} ||x_k−y_k|| ≤ C; if the sequence does not converge, the limit may not exist. Consider replacing the limit with limsup or explicitly require convergence.","section":"§5.1"}],"recommendation":"major_revision","confidential_remarks":"The manuscript's central idea is promising, but the discrepancy between the stated Condition 1 and the proof's actual requirement must be resolved before the stability claim is credible. I also recommend that the authors either broaden the numerical verification to multiple parameter sets or tone down the 'assured' language in the title and abstract. The real-vehicle experiments are open-loop and for a single vehicle; they are supportive but not a stability proof."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here is the short version: this is a workmanlike engineering paper with a genuinely convenient result — an explicit discrete-time single-track model with linear tires whose V and omega updates have denominators mU - Ts(kf+kr) and IzU - Ts(lf^2 kf + lr^2 kr), so they are finite at U=0. The trick is to apply backward-Euler-like updates to V and omega, solve them in closed form, and use forward Euler for X, Y, phi, U. The algebra checks out, and the simulations against CarSim show much better accuracy than the kinematic model and better stability than forward Euler at larger step sizes. That part is solid and useful for MPC and learning-based control in stop-start scenarios.\n\nThe soft spots are in the stability guarantee. Proposition 5.3's proof has a genuine gap at the boundary ||A*|| = 1. Equation (24)'s geometric series only works for ||A*|| < 1; the paper's handling of the equality case in (26) simply asserts ||b_{k-1..0}|| <= ||b*||, but the correct bound is k||b*||, which grows without bound. So the theorem, as proved, establishes bounded disturbance amplification only under the strict inequality. That's probably the operative case in practice — their Figure 3 shows the norm well below 1 for the tested range — but the stated Condition 1 says <= 1. The fix is easy: state Condition 1 with '< 1', or give a separate argument for the boundary. This matters because 'assured numerical stability' is the paper's headline claim.\n\nCondition 1 itself is not derived from the parameters; it's verified numerically for one CarSim prototype, three step sizes, and U in 0-25 m/s. That's honest and the empirical coverage is reasonable, but it falls short of 'assured' for a general vehicle. One stress-test note claims the denominators also vanish at positive speeds; that's wrong — with negative cornering stiffnesses the zero-crossing speeds are negative, so they're outside the stated operating domain. No point holding that against the paper. The other stress-test points land.\n\nThe real-vehicle experiments are open-loop with RMS errors in the 0.3-2.6 m range. They show the model tracks trajectories, but there's no baseline comparison, so they don't demonstrate superiority — they demonstrate adequacy. Fine for a practical note, weaker than the simulation claims.\n\nOverall: this deserves a serious referee and likely acceptance after a moderate revision. The core model is useful and correct; the proof needs the inequality fix and the operating domain needs to be stated explicitly in Condition 1. I'd take it to reading group and would cite it if I were working on predictive control for urban driving.","headline":"A useful semi-implicit discretization of the single-track model that stays finite at zero speed, with a stability theorem that needs a strict-inequality fix and a Condition 1 that is only empirically checked.","tokens_in":25772,"tokens_out":3745,"would_cite":true,"duration_ms":33148,"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":"An explicit discrete-time dynamic vehicle model stays numerically stable at zero speed.","keywords":["dynamic vehicle model","numerical stability","low-speed singularity","explicit discretization","backward Euler","linear tire model","model predictive control","stop-start driving"],"falsifier":"Run two trajectories from nearby initial states under identical control for a vehicle parameter set or a step size where the norm of the two-by-two error-amplification submatrix exceeds 1, and observe the state difference growing without bound; alternatively, construct a trajectory where the norm equals 1 at every step and show the accumulated error grows linearly, which would violate the boundedness that Proposition 5.3 claims.","tokens_in":29,"feed_emoji":"🚗","tokens_out":11553,"duration_ms":216191,"temperature":0.7,"pith_summary":"This paper aims to establish that a six-state single-track dynamic vehicle model can be written in an explicit discrete-time form—the next state is a direct function of the current state and control input—without experiencing the low-speed singularity that makes conventional dynamic models explode near standstill. The singularity comes from tire slip-angle formulas whose denominator is the longitudinal speed $U$; the proposed scheme removes that source of blow-up while retaining the explicit form that model-based controllers prefer. The central guarantee is that the update for lateral velocity $V$ and yaw rate $\\omega$ is obtained through a backward-Euler-style linear solve that becomes algebraic under the linear tire model, and the paper proves a sufficient condition for numerical stability. Simulation comparisons with a commercial full-vehicle simulator and real-vehicle stop-start experiments are presented as evidence that the model is accurate and stable in exactly the low-speed regime where earlier explicit models fail.","feed_headline":"Explicit vehicle model stays stable down to zero speed","feed_subtitle":"A mixed forward-backward Euler rule keeps the explicit form controllers need and ends the low-speed singularity.","key_machinery":"The load-bearing mechanism is a mixed discretization: $X$, $Y$, $\\varphi$, and $U$ are advanced by forward Euler, while $V_{k+1}$ and $\\omega_{k+1}$ are written as backward-Euler-style updates in which the unsolved variable is the one being updated. With the linear tire model, this pair of equations collapses to a two-by-two linear solve, so the model stays explicit. That solving step moves the cornering stiffnesses into the denominators of the lateral-velocity and yaw-rate updates, which is what removes the pure speed dependence and kills the singularity. The error-propagation analysis then factors each step into a block matrix whose two-by-two submatrix is $\\hat{A}_k$, and Condition 1 demands $\\|\\hat{A}_k\\| \\le 1$, a sufficient bound for the product of these matrices to remain bounded.","core_discovery":"The paper's central claim is that its update rule is an explicit discrete-time dynamic vehicle model whose lateral-velocity and yaw-rate updates have denominators $mU_k - T_s(k_f + k_r)$ and $I_z U_k - T_s(l_f^2 k_f + l_r^2 k_r)$, rather than just $U_k$. Because these denominators do not vanish at $U=0$, the slip-angle blow-up that characterizes the low-speed singularity is avoided, and the model remains explicit, so no fixed-point iteration is needed at a step. The paper states that if Condition 1 holds—the two-by-two error-amplification submatrix of the Jacobian has norm at most 1 at every intermediate point—then Proposition 5.3 guarantees numerical stability in the sense that two trajectories driven by the same control sequence stay within a bounded distance forever. The supporting evidence is numerical verification of the condition for speeds from 0 to 25 m/s at step sizes up to 0.1 s, together with accuracy comparisons and real-vehicle trajectories.","pith_inferences":["Beyond the paper: the stability condition is verified numerically, not derived; a useful extension would be an explicit inequality relating $T_s$, mass, inertia, and cornering stiffness that guarantees the norm bound without simulation.","Beyond the paper: since the discrete map is explicit and smooth away from the singular limit, differentiating through it for gradient-based control or learned dynamics is straightforward, but whether the stability margin survives learned parameter changes is untested.","Beyond the paper: at $U=0$ the update for $V$ and $\\omega$ becomes a fixed linear map independent of the control; analyzing that map's eigenvalues would reveal whether a fully stopped vehicle can accumulate spurious yaw rate over many idle steps."],"forward_implications":["The model works where earlier explicit dynamic models break down: urban stop-start maneuvers with $U$ crossing zero no longer produce exploding lateral states.","Because the update is explicit and algebraic, predictive controllers and model-based learning rollouts can call it without iterative solves at each step.","If Condition 1 holds, initial-state errors stay bounded over an arbitrarily long horizon, so small sensing or localization errors do not compound without limit.","The full-vehicle-simulator comparisons show trajectory errors typically an order of magnitude smaller than the kinematic model's, so the added dynamic fidelity comes without a stability penalty.","Stability at step sizes up to $0.1$ s means longer prediction horizons are available under fixed computational budgets."],"supporting_citations":[{"why":"Documents that numerical stability in vehicle system dynamics is classically obtained with implicit methods, motivating the mixed discretization.","marker":"[7]"},{"why":"Provides the forward/backward Euler background and the explicit-implicit trade-off the proposed scheme targets.","marker":"[8]"},{"why":"Shows the tire slip-angle estimate with longitudinal speed in the denominator, the source of the low-speed singularity.","marker":"[9]"},{"why":"Establishes the kinematic-versus-dynamic model distinction and why dynamic models are preferred for high-accuracy control.","marker":"[6]"},{"why":"Supplies the kinematic bicycle model used as the accuracy baseline in the trajectory comparisons.","marker":"[14]"},{"why":"Describes the hysteresis-slip plus speed-lower-bound workaround for low-speed tire modeling that the paper contrasts with its explicit approach.","marker":"[17]"},{"why":"Motivates the need for an explicit discrete model in model predictive control.","marker":"[12]"},{"why":"Motivates the need for explicit transition models in model-based learning.","marker":"[13]"}],"fun_headline_variants":["Explicit car model tames low-speed singularity blow-up","No more infinity at zero speed: explicit model stable","Mixed Euler rule kills low-speed divergence in cars","Zero-speed stability: explicit model with finite slip angles"],"cache_read_input_tokens":27776,"weakest_assumption_plain":"The stability theorem assumes Condition 1, that the two-by-two error-amplification submatrix has norm at most 1 over the whole operating domain; the paper only checks this numerically for one vehicle prototype, three step sizes, and speeds up to 25 m/s, and the proof's bounding step leaves the exactly-equal-to-1 case unresolved.","fun_headline_variants_meta":{"raw":{"variants":["Explicit car model tames low-speed singularity blow-up","No more infinity at zero speed: explicit model stable","Mixed Euler rule kills low-speed divergence in cars","Zero-speed stability: explicit model with finite slip angles"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000176,"raw_usage":{"total_tokens":1318,"prompt_tokens":999,"completion_tokens":319,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":615,"completion_tokens_details":{"reasoning_tokens":256}},"tokens_in":615,"tokens_out":319,"duration_ms":4250,"temperature":1.0,"reasoning_tokens":256,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T12:13:27.652903+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run two trajectories from nearby initial states under identical control for a vehicle parameter set or a step size where the norm of the two-by-two error-amplification submatrix exceeds 1, and observe the state difference growing without bound; alternatively, construct a trajectory where the norm equals 1 at every step and show the accumulated error grows linearly, which would violate the boundedness that Proposition 5.3 claims.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the kinematic bicycle model used as the accuracy baseline in the trajectory comparisons."},{"cited_title":"Nu- merical methods in vehicle system dynamics: state of the art and current developments","cited_arxiv_id":null,"evidence_quote":"Documents that numerical stability in vehicle system dynamics is classically obtained with implicit methods, motivating the mixed discretization."},{"cited_title":"A discussion on euler method: A review","cited_arxiv_id":null,"evidence_quote":"Provides the forward/backward Euler background and the explicit-implicit trade-off the proposed scheme targets."},{"cited_title":"Recent advances in tyre models and testing procedures","cited_arxiv_id":null,"evidence_quote":"Shows the tire slip-angle estimate with longitudinal speed in the denominator, the source of the low-speed singularity."},{"cited_title":"Kinematic and dynamic ve- hicle models for autonomous driving control design","cited_arxiv_id":null,"evidence_quote":"Establishes the kinematic-versus-dynamic model distinction and why dynamic models are preferred for high-accuracy control."},{"cited_title":"Tire modeling for low-speed and high-speed calculations.SAE Transactions, pages 474–483, 1995","cited_arxiv_id":null,"evidence_quote":"Describes the hysteresis-slip plus speed-lower-bound workaround for low-speed tire modeling that the paper contrasts with its explicit approach."},{"cited_title":"Model predictive multi- objective vehicular adaptive cruise control","cited_arxiv_id":null,"evidence_quote":"Motivates the need for an explicit discrete model in model predictive control."},{"cited_title":"Improve generalization of driving policy at signalized intersections with adver- sarial learning","cited_arxiv_id":null,"evidence_quote":"Motivates the need for explicit transition models in model-based learning."}],"review_version":1}