{"id":"df14731c-a6be-4862-b712-c68c2740090b","arxiv_id":"2506.15321","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"high","formal_verification":"none","parameter_count":3,"one_line_summary":"Across nine pedestrian datasets, the maximum angular velocity scales with walking speed as v^{-0.8}, bounding the speed-turning phase space.","lead":"This paper analyzes pedestrian trajectories from nine experiments and reports a power-law ceiling on how fast a walker can turn at a given speed: the maximum angular velocity falls as speed to the -0.8 power. If the pattern holds, it would give crowd modelers a simple kinematic bound for trajectory prediction and simulation.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The reported exponents (-1.8 for κ, -0.8 for ω) are within the noise floor of the Eq. 2 estimator for straight walking with fixed tracking noise, so the claimed kinematic bound may be an artifact; a straight-walk null test can settle it.","rationale":"The stress-test pass should not manufacture doubt. The paper is honest about its method: Section 3.2 explicitly labels the trend curves as 'subjectively defined', and the open data are a real asset. However, the central claim is quantitative and universal, and the single most load-bearing assumption is that the upper envelope in the scatter plots is a real kinematic boundary. The reader flagged this as an artifact risk; I agree, and the mechanism can be made precise from Eq. 2: with fixed positional noise, the curvature estimator's noise floor scales as v^{-2}, so the reported κ exponent (-1.8) and its ω restatement (-0.8) are close to what pure noise would produce. This is not an internal inconsistency or a disagreement with consensus; it is a correctness risk in the inference from data to constraint. The proposed straight-walk test is decisive because straight walking has no turning constraint, so any surviving v^{-1.8} envelope would be entirely attributable to the estimator. If the test instead shows a flat or different envelope on straight segments, the kinematic-constraint interpretation is supported. Given the present manuscript contains no such control, the reader's REJECT verdict is unchanged.","tokens_in":11727,"tokens_out":6023,"duration_ms":65982,"concrete_test":"Use the shared open data (Data Availability). In a rectilinear dataset such as Experiment 8 single-file walking, select windows with essentially no turning (e.g., heading change < 1° over 1 s). Compute v, κ, and ω exactly as in §3.1 for both sampling intervals, then fit the upper envelope with the same subjective fixed-exponent procedure. If these no-turning segments still produce κ_max ~ v^{-1.8} (ω_max ~ v^{-0.8}), the envelope is a measurement-noise artifact, not a kinematic bound. A complementary simulation: add Gaussian position noise of tracker-residual amplitude to synthetic straight trajectories and check whether the fitted envelope exponent approaches -2 (-1 for ω).","verdict_should_be":"UNCHANGED","load_bearing_attack":"Equation 2 estimates κ = α/s from position samples. If the tracker has perpendicular noise ε, the heading error is ~ε/s, so the error in κ is ~ε/s^2 = ε/(v^2 Δt^2). Thus, for perfectly straight walking, the largest estimated κ values in a speed bin decline as v^{-2}, and the largest estimated ω = κ v decline as v^{-1}, purely from the estimator's noise floor. The manuscript reports κ_crit ∝ v^{-1.8} and ω_crit ∝ v^{-0.8} (Eq. 3): both are close to these null scalings. Section 3.2 states the trend curves are 'subjectively defined' with a 'fixed exponent of -1.8', so the exponent is imposed rather than estimated with uncertainty, and no null model, goodness-of-fit, or error analysis is given. Moreover, because ω = κv (Eq. 2), Eq. 3's ω∝v^{-0.8} is just the algebraic restatement of the κ∝v^{-1.8} fit, not an independent result. Finite-sample maxima strengthen the artifact: with more samples, the upper envelope of a noisy cloud rises, so a declining envelope can appear without any kinematic constraint. The claim that pedestrians obey a universal turning limit therefore rests on the untested assumption that the envelope is not dominated by this estimator noise.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper analyzes pedestrian trajectories from nine experiments, computes curvature and speed from position samples via Eq. (2), and plots scatter diagrams. The authors draw a \"subjectively defined\" trend curve with a fixed exponent of -1.8 for the upper envelope of curvature versus speed, and then state that this implies a power-law scaling between critical angular velocity and speed, ωcrit ∝ v^{-0.8} (Eq. 3). They interpret this as a universal kinematic constraint bounding the speed–angular-velocity phase space of pedestrian motion. The manuscript provides no quantitative fitting procedure, no goodness-of-fit measures, and no control for measurement noise.","tokens_in":12062,"tokens_out":5694,"duration_ms":58695,"significance":"If the claimed scaling were established, it would offer a simple kinematic constraint useful for pedestrian modeling and trajectory prediction. The paper compiles a large set of heterogeneous experimental data, which is a strength. However, the central evidence is an eyeballed envelope with a fixed exponent, and the angular-velocity scaling is an algebraic restatement of the curvature scaling. Moreover, the reported exponents are close to those expected from the estimator's noise floor for straight walking, and the paper provides no null test. As it stands, the central claim is not supported by the presented analysis.","major_comments":[{"comment":"The trend curve is described as \"subjectively defined\" with a \"fixed exponent of -1.8\" in curvature. No fitting procedure, confidence intervals, goodness-of-fit statistics, or comparison with alternative exponents is reported. The exponent is therefore imposed rather than estimated from the data, and this is the sole evidence for the claimed universal power law.","section":"Section 3.2, Eq. (3)"},{"comment":"Because Eq. (2) defines ω = κ v, the statement ωcrit ∝ v^{-0.8} is a direct algebraic rearrangement of κcrit ∝ v^{-1.8}. The angular-velocity scaling is not an independent empirical finding; the paper's headline claim about ω carries no information beyond the κ–v envelope.","section":"Eq. (3)"},{"comment":"The upper envelope of the scatter clouds is interpreted as a kinematic boundary, but no test rules out the null hypothesis that it arises from the estimator's noise floor. For straight-line walking with perpendicular tracking noise ε, the error in curvature estimated by Eq. (2) is of order ε/(v^2 Δt^2), so the largest estimated κ in a speed bin scales as v^{-2} and the largest estimated ω = κ v scales as v^{-1}. The reported exponents, -1.8 and -0.8, are close to these null scalings, yet no straight-walk control, synthetic-noise test, or sample-size analysis is provided to show that the envelope is a genuine kinematic constraint.","section":"Section 3.2, Figs. 12–20"},{"comment":"The paper does not report how many samples fall in each speed bin or whether the upper envelope is stable under subsampling. Since the maximum value of a noisy sample cloud increases with sample size, a declining envelope can appear simply from heterogeneous sample counts across speed bins, even with no physical turning limit. This alternative explanation is not addressed.","section":"Section 3.2, sampling configuration"}],"minor_comments":[{"comment":"In the paragraph on trajectory errors, \"will introducing\" should be \"will introduce.\"","section":"Section 2.1"},{"comment":"The Sampling Interval column lists two values for each dataset; the caption should clarify which value corresponds to the frame-by-frame sampling and which to the decimated sampling, and how these relate to the Frame Rate column.","section":"Table 1"},{"comment":"The use of the plus-minus sign in κcrit ∝ ±v^{-1.8} is confusing; the scaling presumably applies to the absolute value of curvature, and this should be stated explicitly.","section":"Eq. (3)"},{"comment":"The captions do not identify the axes or their units clearly. The figures should state explicitly that the vertical axis is curvature (or angular velocity) and the horizontal axis is speed, with units.","section":"Figures 12–20"},{"comment":"The data are hosted on a Google Drive folder without a DOI or permanent identifier; a stable repository DOI would improve reproducibility.","section":"Data availability"}],"recommendation":"reject","confidential_remarks":"The stress-test concern about the estimator noise floor is compelling, and the paper's own admission of a subjectively defined trend curve makes the central claim unsupported as submitted. The reported exponents are close to the null expectations for pure tracking noise, and without a null test the paper does not demonstrate a real kinematic bound. A complete re-analysis with rigorous envelope estimation and noise controls would be needed before the claim could be evaluated; this goes beyond a minor revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe headline claim here — a universal ωcrit ∝ v^{-0.8} bound across nine pedestrian datasets — is not supported by the evidence as presented. The exponent is imposed, not measured. Section 3.2 states the trend curve is 'subjectively defined' with a 'fixed exponent of -1.8' in curvature; the ω exponent is then just the algebraic restatement (ω = κv). No fit statistics, confidence intervals, or null models appear anywhere. The stress-test worry lands hard: for straight walking with fixed tracking noise, the estimator in Eq. 2 gives κ errors that scale as v^{-2} and ω errors as v^{-1}, so the largest estimated values in a speed bin decline with speed even in the complete absence of a kinematic constraint. The reported exponents (-1.8, -0.8) are close enough to those null scalings that the envelope may be nothing more than the noise floor of the estimator.\n\nWhat the paper does well: it assembles a genuinely broad set of independent trajectory datasets, applies a variety of preprocessing filters, and examines two sampling rates. That is real work, and the data are open. The question — does a speed-dependent turning bound exist? — is a good one. The bounded-region picture is plausible and consistent with the known speed-curvature trade-off in the literature.\n\nThe soft spots are not minor; they are load-bearing. The envelope is fit by eye. The exponent is fixed a priori, which is circular when the same scatter is said to confirm it. The decimation analysis shows changes in the constant coefficient but no robustness check on the exponent. And the paper does not report any negative control, such as computing the same envelopes from straight-line walking.\n\nBottom line: this is a paper that deserves a serious referee because the question and data are worth engaging, but it is not publishable in its current form. A revision would need a proper envelope estimator (e.g., quantile regression or extreme-value methods), a null-model test on straight-walk data, and honest error bars on the exponent. If the exponent survives those tests, it would be a real result. I would not cite it yet, but I would bring it to a reading group as a case study in estimator noise masquerading as a scaling law.","headline":"The paper's universal turning-limit exponent is likely an artifact of the curvature estimator's noise floor, but the question and data are worth a serious revision.","tokens_in":12582,"tokens_out":2108,"would_cite":false,"duration_ms":20492,"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":"Pedestrian turning rate has a speed limit: the fastest allowable turn falls as walking speed increases.","keywords":["pedestrian dynamics","angular velocity","curvature","power-law scaling","kinematic constraint","crowd trajectories","turning behavior","phase space"],"falsifier":"Refit the upper envelope after randomly subsampling trajectories from each of the nine datasets; if the fitted exponent changes systematically with sample size, or if surrogate trajectories built from the same speed distribution and noise but with unconstrained random headings show the same $-0.8$ power-law envelope, then the claimed kinematic constraint is a statistical artifact rather than a physical boundary.","tokens_in":11532,"feed_emoji":"🚶","tokens_out":8864,"duration_ms":82502,"temperature":0.7,"pith_summary":"This paper claims that pedestrian turning is governed by a kinematic speed limit: the highest angular velocity a walker can sustain falls as $\\omega_{\\mathrm{crit}} \\propto v^{-0.8}$, equivalently critical curvature falls as $\\kappa_{\\mathrm{crit}} \\propto v^{-1.8}$. The claim is based on re-analyzing trajectories from nine controlled experiments covering crossing flows, unidirectional and bidirectional corridors, pedestrian-vehicle interaction, single-file walking, and antipodal circle crossings. In every dataset and at two sampling rates, the upper edge of the speed-angular velocity scatter cloud follows the same power law, so the authors interpret it as a universal boundary separating kinematically possible from impossible pedestrian motion. A reader should care because, if true, this is a simple quantitative constraint that any model of pedestrian motion, from crowd simulations to trajectory prediction, would need to respect.","feed_headline":"Pedestrians turn less sharply as they speed up","feed_subtitle":"Across nine trajectory datasets, the maximum turning rate scales as walking speed to the -0.8 power.","key_machinery":"The load-bearing object is the critical envelope of the scatter plot of instantaneous speed versus angular velocity, computed from discrete trajectories as $\\kappa = d\\theta/d\\ell \\simeq \\alpha/s = \\omega/v$, where $\\alpha$ is the heading change in a sampling interval and $s$ the distance traveled. On log-log axes the upper edge of each cloud is fitted, by a subjectively drawn trend curve, to a power law with a fixed exponent, giving $\\kappa_{\\mathrm{crit}} \\propto \\pm v^{-1.8}$ and hence $\\omega_{\\mathrm{crit}} \\propto \\pm v^{-0.8}$ (Eq. 3). This envelope, not any individual trajectory, is what carries the argument that a kinematic turning constraint exists and that it scales with speed.","core_discovery":"The central discovery, on the paper's own terms, is that pedestrian trajectories occupy a bounded region of the $(\\omega, v)$ phase space whose upper envelope obeys a power law with exponent approximately $-0.8$, rather than a fixed angular-velocity ceiling. Because curvature is defined as $\\kappa = \\omega/v$, this is equivalent to a critical-curvature law $\\kappa_{\\mathrm{crit}} \\propto v^{-1.8}$. The bound appears consistently across nine datasets with different geometries, densities, interaction types, filtering schemes, and sampling intervals, which the authors take as evidence that it is a kinematic constraint intrinsic to walking rather than a feature of any single experiment or measurement setup.","pith_inferences":["The paper does not derive the exponent; a biomechanical model based on step frequency and stride length might explain $-0.8$ and could be tested against these same datasets.","The paper pools all trajectories into one cloud; a stricter test would fit per-pedestrian critical curves and ask whether the exponent survives within a single walker.","If the bound is truly kinematic, it should be reproducible under forced circular walking at controlled speeds, with the tightest sustainable radius growing as $v^{1.8}$.","A surrogate-data check would separate physics from statistics: random-walk headings with identical speed and noise statistics should not reproduce the $-0.8$ envelope if the constraint is real."],"forward_implications":["At higher walking speeds, sharp turns become kinematically less available: the maximum sustainable angular velocity decreases, so rapid direction changes must be spread over longer distances or preceded by deceleration.","The same bound implies a minimum turning radius that grows with speed: because $\\kappa \\propto v^{-1.8}$, the radius of the tightest achievable turn scales roughly as $v^{1.8}$.","Trajectory prediction and crowd-simulation models that ignore the bound will over-generate high-speed, high-curvature motion that real pedestrians do not produce.","The speed-angular velocity phase space, not just the classical speed-density fundamental diagram, should be treated as a basic descriptor of pedestrian motion.","Because the law is reported at both frame-by-frame and decimated sampling, the constraint is claimed to be stable with respect to temporal resolution."],"supporting_citations":[{"why":"Supplies the perpendicular crossing-flow dataset (Experiment 3) with two flow geometries and a wide density range.","marker":"Cao et al. 2017"},{"why":"Supplies the bidirectional flow dataset (Experiment 2) covering individual and group walking.","marker":"Feliciani & Nishinari 2016"},{"why":"Supplies the lane-nucleation active-flow dataset (Experiment 4) with head-on, chiral, and perpendicular streams.","marker":"Bacik et al. 2023"},{"why":"Supplies the pedestrian-vehicle interaction dataset (Experiment 5), pre-filtered with a Kalman filter.","marker":"Yang et al. 2019"},{"why":"Supplies the unidirectional circular-corridor data at different walking speeds (Experiment 6).","marker":"Fujita et al. 2019"},{"why":"Supplies the single-file age-composition dataset (Experiment 8) spanning young and older adults.","marker":"Cao et al. 2016"},{"why":"Supplies the circle-antipode collision-avoidance dataset (Experiment 9), the only raw unfiltered comparison.","marker":"Xiao et al. 2019"},{"why":"Documents the crossing-motion experiment (Experiment 1) and its trajectory-extraction procedure.","marker":"Wang et al. 2023"}],"fun_headline_variants":["Max turn rate scales as walking speed to the -0.8","Pedestrian turning obeys a universal -0.8 power law","Speed-angular velocity phase space has a bounded envelope","Critical angular velocity drops with speed across 9 datasets","A kinematic bound: pedestrian turns scale with speed^-0.8"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the upper edge of the scatter cloud, which the authors themselves describe as a subjectively defined trend curve, is a genuine kinematic boundary rather than an artifact created by finite sample sizes, measurement noise, and the way maxima grow with the number of trajectories sampled.","fun_headline_variants_meta":{"raw":{"variants":["Max turn rate scales as walking speed to the -0.8","Pedestrian turning obeys a universal -0.8 power law","Speed-angular velocity phase space has a bounded envelope","Critical angular velocity drops with speed across 9 datasets","A kinematic bound: pedestrian turns scale with speed^-0.8"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000165,"raw_usage":{"total_tokens":1169,"prompt_tokens":780,"completion_tokens":389,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":396,"completion_tokens_details":{"reasoning_tokens":304}},"tokens_in":396,"tokens_out":389,"duration_ms":4362,"temperature":1.0,"reasoning_tokens":304,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T19:36:32.745873+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Refit the upper envelope after randomly subsampling trajectories from each of the nine datasets; if the fitted exponent changes systematically with sample size, or if surrogate trajectories built from the same speed distribution and noise but with unconstrained random headings show the same $-0.8$ power-law envelope, then the claimed kinematic constraint is a statistical artifact rather than a physical boundary.","supporting_citations":[{"cited_title":"2017, Journal of Statistical Mechanics: Theory and Experiment, 2017, 033404, doi: 10.1088/1742-5468/aa620d","cited_arxiv_id":null,"evidence_quote":"Supplies the perpendicular crossing-flow dataset (Experiment 3) with two flow geometries and a wide density range."},{"cited_title":"2016, Physical Review E, 94, 032304, doi: 10.1103/PhysRevE.94.032304","cited_arxiv_id":null,"evidence_quote":"Supplies the bidirectional flow dataset (Experiment 2) covering individual and group walking."},{"cited_title":"2019, Physical Review E, 99, 062307, doi: 10.1103/PhysRevE.99.062307","cited_arxiv_id":null,"evidence_quote":"Supplies the unidirectional circular-corridor data at different walking speeds (Experiment 6)."},{"cited_title":"2016, Physical Review E, 94, 012312, doi: 10.1103/PhysRevE.94.012312","cited_arxiv_id":null,"evidence_quote":"Supplies the single-file age-composition dataset (Experiment 8) spanning young and older adults."},{"cited_title":"2019, Transportation Research Part C: Emerging Technologies, 103, 174, doi: 10.1016/j.trc.2019.04.007","cited_arxiv_id":null,"evidence_quote":"Supplies the circle-antipode collision-avoidance dataset (Experiment 9), the only raw unfiltered comparison."},{"cited_title":"Exploring crowd persistent dynamism from pedestrian crossing perspective: An empirical study","cited_arxiv_id":"2311.04827","evidence_quote":"Documents the crossing-motion experiment (Experiment 1) and its trajectory-extraction procedure."}],"review_version":2}