Pith. sign in

REVIEW 4 major objections 4 minor 39 references

Towards Full-Scenario Safety Evaluation of Automated Vehicles: A Volume-Based Method

T0 review · 4 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Volume of dangerous scenarios ranks AV safety without crash data

desk verdict A genuinely different volume-based AV safety metric with a correct convexity theorem, but the key equivalence to probability-based ranking is assumed, not proven, so it needs serious but skeptical revision. read the letter →

arxiv 2506.09182 v2 pith:KO7L65Y7 submitted 2025-06-10 cs.RO cs.ET

Hang Zhou , Chengyuan Ma , Shiyu Shen , Zhaohui Liang , Xiaopeng Li This is my paper
classification cs.ROcs.ET
keywords automatedvehiclesafetyevaluationvolume-basedmetricscenariospacecrashratetimetocollisionconvexpolytopeMonteCarlovolumeestimationcar-followingmodel
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper proposes a way to evaluate the safety of automated vehicles by measuring the geometric volume of dangerous scenarios in the space of all possible driving situations, rather than estimating crash rates from naturalistic driving data. It argues that for high-performing automated vehicles, ranking by this volume gives approximately the same safety comparison as ranking by crash probability, while avoiding the need for large and complete behavior datasets. The paper also proves that for car-following with a linear control model, the set of safe scenarios is a convex polytope, so its volume can be computed exactly. This would make full-scenario safety evaluation more data-efficient and scalable.

What carries the argument

The central object is the dangerous-scenario set $D = \{x \in \Omega \mid SM(x) \le \eta\}$, defined by a surrogate safety metric, Time to Collision (TTC), over a bounded scenario space $\Omega$. The method's engine is the ratio $\mathrm{vol}(D)/\mathrm{vol}(\Omega)$, computed either by Monte Carlo sampling or, when Theorem 2 applies, by exact polytope volume from a half-space representation. Theorem 2's convexity result comes from rewriting the TTC condition as affine half-spaces under a linear car-following model, making the safe set an intersection of half-spaces.

What would settle it

Compare the volume-based and probability-based safety rankings of two AV controllers on a dataset whose scenario probabilities are strongly non-uniform, for example with most scenarios concentrated in a narrow speed range; if the ranking orders the vehicles differently, the equivalence in Theorem 1 fails in that regime.

Watch

Extended reading notes

Core claim

The paper's central claim is that the safety of an automated vehicle can be quantified by the proportion of the scenario space that is dangerous, vol(D)/vol(Ω), and that this metric is a valid basis for comparing vehicles when their safety performance is strong. Theorem 1 states that in this regime the volume-based and probability-based methods produce approximately equivalent safety rankings, because for well-performing vehicles the scenarios on which two systems disagree are borderline, low-probability edge cases. Theorem 2 states that when the car-following model is linear, the safe scenario set is convex for any TTC threshold, which allows exact volume computation via polytope algorithms. The paper further provides a Monte Carlo algorithm for general scenarios and validates the framework on synthetic controllers and six production ACC-equipped vehicles calibrated from trajectory data.

Load-bearing premise

The approach rests on the assumption that, for two vehicles being compared, the driving scenarios where exactly one of them is dangerous are roughly equal in probability, so that weighting these scenarios by real-world frequency would not change the safety ranking.

Editorial extensions

If this is right

  • Safety rankings of production AVs can be produced without collecting naturalistic driving data to estimate scenario probabilities.
  • For linear car-following controllers, the volume of the safe region can be computed exactly and quickly, eliminating sampling variance in car-following evaluations.
  • The Monte Carlo algorithm scales to multi-lane and other complex scenarios, avoiding both the curse of dimensionality and the curse of rarity.
  • The volume-based metric can rank AVs even when crash probability data is incomplete, as long as the scenario space is consistently bounded.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The equivalence in Theorem 1 could be tested empirically by comparing volume rankings with probability-based rankings on a large naturalistic dataset; divergence would mark the boundary of the method's validity.
  • The convexity result might extend to other safety metrics that are affine in the state variables, such as deceleration-based surrogates, enabling exact volume computation beyond car-following.
  • The volume metric's dependence on the chosen bounds for gap, speed, and acceleration means rankings may shift if the scenario box changes; a standardized scenario box would be needed for regulatory use.
  • Designing controllers to minimize vol(D)/vol(Ω) would directly shrink the geometric failure region, turning the evaluation metric into an optimization objective.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. The paper proposes a volume-based framework for automated-vehicle safety evaluation. It introduces a unified scenario representation that reduces diverse driving situations to at most three lanes and six surrounding vehicles, defines dangerous scenarios through a TTC threshold, and proposes vol(D)/vol(Omega) as a safety metric that avoids estimating naturalistic scenario probabilities. The paper claims that for highly automated AVs the volume-based ranking approximately matches a probability-based ranking (Theorem 1), and that for linear car-following models the safe scenario set is convex (Theorem 2), enabling exact polytope-volume computation. The framework is assessed with Monte Carlo, vertex enumeration, and randomized polytope-volume algorithms in single-lane and multi-lane simulations, and is applied to six production AV car-following controllers calibrated from the Ultra-AV dataset.

Significance. If the equivalence between volume-based and probability-based rankings were established, the proposed metric would be a genuinely useful, data-efficient complement to crash-rate estimation, and the convexity theorem would allow exact volume computation in an important special case. The paper has real strengths: the convexity argument is essentially correct once the missing relative-speed case in Eq. (18) is supplied by constraints (23)--(24); the Monte Carlo method is validated against exact vertex enumeration with small errors; the use of field-calibrated production AV models is a valuable step beyond purely synthetic controllers; and the proposed metric is not fitted to produce a desired ranking. However, the central theoretical claim that volume can replace probability is not actually proved, and no experiment tests it against a nonuniform probability baseline. As a result, the paper currently establishes a tractable volume-measurement tool, but not the safety-ranking justification that motivates it.

major comments (4)
  1. [2.3, Theorem 1 (Eq. 14)] Theorem 1 is not proved; the crucial uniformity condition in Eq. (14) is assumed rather than derived. The proof reduces the difference between probability-based and volume-based comparisons to the disagreement set D(Veh1,Veh2), which is correct, but then it simply asserts that P(x1) approximately equals P(x2) over that set. The informal reasoning that follows Eq. (14) does not establish this from any property of 'highly automated' controllers or from driving statistics, and the term 'highly automated' is never given a quantitative definition. A nonuniform scenario density concentrated on a small-volume subset of D(Veh1,Veh2) can make the signs of the probability difference and the volume difference disagree, so Eq. (14) is essentially the theorem's conclusion in disguise. The paper should either state Theorem 1 under an explicit quantitative condition (for example, a bounded density ratio over the disagreement set) with an error bound, or reframe the claim as an empirical hypothesis and test it; the current statement overclaims.
  2. [4.2--4.4] No experiment checks the claimed volume--probability equivalence against a nonuniform probability baseline. The production-AV ranking in Section 4.4 is computed purely from volumes, and the conclusion explicitly defers a comparison with probability-based results to future work. Because Theorem 1 is the only justification for using volume as a proxy for crash risk, the experimental section should include at least one setting in which the same controllers are ranked under a plausible nonuniform scenario distribution estimated from the Ultra-AV dataset, and the two rankings should be compared. Without such a comparison, the paper's central methodological claim is unsupported by the evidence it presents.
  3. [3.1, Algorithm 1 and Eq. (15)] The Monte Carlo estimator in Eq. (15) equals vol(D)/vol(Omega) only if the samples are drawn uniformly from the scenario space Omega, but Algorithm 1 specifies only that the initial state and BV behaviors are 'randomly sampled' without stating the distribution. If the BV behavior draws use naturalistic, calibrated, or otherwise nonuniform conditional distributions, the reported percentages are not volume proportions under the Lebesgue measure on Omega. The authors should state explicitly that all scenario variables are sampled uniformly from their declared boxes, or otherwise show that the estimator is unbiased for the volume ratio intended by Eq. (15).
  4. [2.3 and 4.1] The volume ratio is sensitive to the choice of the scenario bounds Omega, specified in Section 4.1 as dmin=5 m, dmax=100 m, vmin=0 m/s, vmax=40 m/s, amin=-4 m/s^2, amax=2 m/s^2. These bounds are partly a modeling choice, and changing them changes the relative weight given to different regions of the scenario space; no sensitivity analysis is reported. A controller that fails in a small but high-probability region can appear better or worse depending on how the box is chosen. The paper should report the ranking under several plausible box choices, or justify why a single fixed box is a canonical test envelope independent of the evaluation goal.
minor comments (4)
  1. [3.2, Eq. (18)] The equivalence in Eq. (18) omits the case vf_t - vl_t <= 0, for which TTC is infinite and the scenario is safe regardless of the gap. The later constraints (23)--(24) correctly cover this case, so the convexity theorem survives, but the proof should state the piecewise definition explicitly.
  2. [3.2] The text at the end of the theorem says 'the feasible set D forms a convex polyhedron' even though D was defined as the dangerous set in Eq. (6); the convex set is the safe set S. The same confusion appears earlier as 'the safe scenario set D' before Theorem 2.
  3. [2.1, Eq. (1)] Equation (1) lists 'vf0, vf0' twice in the argument of the car-following function; the second occurrence should presumably be vl0, the initial lead-vehicle speed.
  4. [Title and Section 2.2] The term 'full-scenario' is stronger than what is demonstrated: the experimental evaluation in Sections 4.2--4.4 covers single-lane car-following and a MOBIL-based multi-lane setting, but does not exercise the full set of scenario transformations described in Section 2.2, such as intersections, merging, work zones, or pedestrians.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the volume metric and convexity theorem follow from stated definitions and model assumptions, and the production-AV ranking is an output of external field-data calibration; Theorem 1's uniformity condition is an unproven sufficient condition rather than a circular reduction.

full rationale

The paper's derivation chain is not circular. The volume metric vol(D)/vol(Omega) is defined directly from the TTC threshold (Eqs. 5-7) and the bounded scenario box (Eqs. 25-29), with no parameter fitted to any target safety ranking. Theorem 2's intended derivation is a genuine consequence of the linear car-following model: each TTC condition (Eq. 18) becomes an affine half-space, so the safe set is an intersection of half-spaces; this does not presuppose the ranking. (A separate completeness concern: Eq. 18 rewrites TTC>eta only for v_f-v_l>0 and omits the v_f-v_l<=0 region where TTC is +infinity; this is a correctness issue, not a circularity.) The production-AV results use controller parameters calibrated from the external Ultra-AV field dataset (Zhou et al. 2024); the calibration target is longitudinal acceleration behavior, not dangerous-scenario volume, so the safety ranking is an output rather than a fitted input. The main caveat is Theorem 1: the proof shows that probability- and volume-based rankings agree when scenario probabilities are approximately uniform over the disagreement set (Eq. 14), and then asserts without proof that highly automated AVs satisfy this condition. That assertion is a soundness gap, not a circular reduction, because Eq. 14 is a stated sufficient condition rather than the theorem's conclusion itself, and the volume metric can be computed without relying on Eq. 14. The paper's own conclusion defers an empirical comparison with probability-based evaluation to future work, confirming that Theorem 1 is an unvalidated premise rather than a manufactured equivalence. Self-citations (the Ultra-AV dataset and prior scenario-ranking work) serve as data sources or benchmarks, not as authorities that forbid alternative choices, so they are not load-bearing circularity. The low score reflects only these non-circular caveats.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central framework rests on several domain assumptions about scenario representation and safety metrics. The most fragile is the uniformity condition in Theorem 1. The volume metric also depends on arbitrarily chosen scenario bounds. The calibrated controller parameters are empirically fitted, reducing inventiveness.

free parameters (4)
  • Scenario bounds dmin, dmax, vmin, vmax, amin, amax = 5, 100 m; 0, 40 m/s; -4, 2 m/s^2
    Chosen by hand to define the bounded scenario space; the volume metric is sensitive to these bounds, so they act as free choices that affect the reported proportions.
  • TTC threshold η = 1 s (and 0.5 s intervals up to 5 s)
    Threshold separating safe from dangerous scenarios; the dangerous volume depends on it.
  • Time horizon T and step Δt = T from 1 to 25, Δt=0.2 s
    Longer horizons include more dangerous scenarios; volume proportion grows with T.
  • Calibrated car-following parameters (k1, k2, thw) for VehA-F = From Table 1, e.g., VehA k1=0.018, k2=0.156, thw=1.378
    Fitted to Ultra-AV trajectory data; the safety ranking is a function of these fitted values.
assumptions (5)
  • domain assumption Linear car-following models adequately represent production ACC controllers
    Used in Theorem 2 and all experiments (Section 3.2, Section 4.1); if controllers are nonlinear or time-varying, convexity may fail.
  • domain assumption TTC is a valid surrogate safety metric for ranking hazards
    The dangerous set is defined by TTC thresholds; other surrogate metrics may give different orderings.
  • domain assumption Any driving scenario can be represented with at most three lanes and six surrounding vehicles without loss of safety-relevant information
    The unified model in Section 2.2 abstracts away exact lateral positions and ignores agents beyond the six nearest; this is essential to the 'full-scenario' claim.
  • domain assumption The scenario space Ω is a bounded box and uniform sampling over it defines a meaningful measure
    Section 3.1 uses uniform i.i.d. sampling and may not cover the BV behavior distributions explicitly; volume is not invariant under rescaling of the box.
  • ad hoc to paper For highly automated AVs, scenario probabilities are approximately uniform over the disagreement set in Theorem 1
    Section 2.3, Eq. (14); this condition is asserted, not proven, and is effectively the theorem's conclusion.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Towards Full-Scenario Safety Evaluation of Automated Vehicles: A Volume-Based Method." pith.science (2026). https://pith.science/paper/KO7L65Y7

@misc{pith2026250609182,
  author       = {Pith},
  title        = {Pith review of: Towards Full-Scenario Safety Evaluation of Automated Vehicles: A Volume-Based Method},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KO7L65Y7}},
  note         = {Machine review of arXiv:2506.09182}
}
read the original abstract

With the rapid development of automated vehicles (AVs) in recent years, commercially available AVs are increasingly demonstrating high-level automation capabilities. However, most existing AV safety evaluation methods are primarily designed for simple maneuvers such as car-following and lane-changing. While suitable for basic tests, these methods are insufficient for assessing high-level automation functions deployed in more complex environments. First, these methods typically use crash rate as the evaluation metric, whose accuracy heavily depends on the quality and completeness of naturalistic driving environment data used to estimate scenario probabilities. Such data is often difficult and expensive to collect. Second, when applied to diverse scenarios, these methods suffer from the curse of dimensionality, making large-scale evaluation computationally intractable. To address these challenges, this paper proposes a novel framework for full-scenario AV safety evaluation. A unified model is first introduced to standardize the representation of diverse driving scenarios. This modeling approach constrains the dimension of most scenarios to a regular highway setting with three lanes and six surrounding background vehicles, significantly reducing dimensionality. To further avoid the limitations of probability-based method, we propose a volume-based evaluation method that quantifies the proportion of risky scenarios within the entire scenario space. For car-following scenarios, we prove that the set of safe scenarios is convex under specific settings, enabling exact volume computation. Experimental results validate the effectiveness of the proposed volume-based method using both AV behavior models from existing literature and six production AV models calibrated from field-test trajectory data in the Ultra-AV dataset. Code and data will be made publicly available upon acceptance of this paper.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

39 extracted references · 37 canonical work pages

  1. [1]

    , " * write output.state after.block = add.period write newline

    ENTRY address author booktitle chapter doi edition editor eid howpublished institution isbn issn journal key month note number organization pages publisher school series title type url volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.all := #1...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in "" FUNCTION format.date year ...

  3. [3]

    Journal of Intelligent Transportation Systems 25(2):170--187

    Arvin R, Khattak AJ, Kamrani M, Rio-Torres J (2020) Safety evaluation of connected and automated vehicles in mixed traffic with conventional vehicles at intersections. Journal of Intelligent Transportation Systems 25(2):170--187

    work page 2020

  4. [4]

    Mathematical Programming 35(1):32--41

    Byrd RH, Schnabel RB (1986) Continuity of the null space basis and constrained optimization. Mathematical Programming 35(1):32--41

    work page 1986

  5. [5]

    Volesti: Volume Approximation and Sampling for Convex Polytopes in R

    Chalkis A, Fisikopoulos V (2020) volesti: Volume approximation and sampling for convex polytopes in r. arXiv preprint arXiv:2007.01578

    work page Pith review arXiv 2020

  6. [6]

    Journal of the ACM (JACM) 26(3):401--414

    Cohen J, Hickey T (1979) Two algorithms for determining volumes of convex polyhedra. Journal of the ACM (JACM) 26(3):401--414

    work page 1979

  7. [7]

    IEEE Transactions on Intelligent Transportation Systems 24(7):6971--6988

    Ding W, Xu C, Arief M, Lin H, Li B, Zhao D (2023) A survey on safety-critical driving scenario generation—a methodological perspective. IEEE Transactions on Intelligent Transportation Systems 24(7):6971--6988

    work page 2023

  8. [8]

    Conference on robot learning, 1--16 (PMLR)

    Dosovitskiy A, Ros G, Codevilla F, Lopez A, Koltun V (2017) Carla: An open urban driving simulator. Conference on robot learning, 1--16 (PMLR)

    work page 2017

Show all 39 references
  1. [9]

    Mathematics of Operations Research 8(3):381--402

    Dyer ME (1983) The complexity of vertex enumeration methods. Mathematics of Operations Research 8(3):381--402

    1983

  2. [10]

    ACM Transactions on Mathematical Software (TOMS) 44(4):1--21

    Emiris IZ, Fisikopoulos V (2018) Practical polytope volume approximation. ACM Transactions on Mathematical Software (TOMS) 44(4):1--21

    2018

  3. [11]

    IEEE Transactions on Intelligent Transportation Systems 22(9):5635--5647

    Feng S, Feng Y, Sun H, Bao S, Zhang Y, Liu HX (2020 a ) Testing scenario library generation for connected and automated vehicles, part ii: Case studies. IEEE Transactions on Intelligent Transportation Systems 22(9):5635--5647

    2020

  4. [12]

    IEEE Transactions on Intelligent Transportation Systems 23(2):1213--1222

    Feng S, Feng Y, Sun H, Zhang Y, Liu HX (2020 b ) Testing scenario library generation for connected and automated vehicles: An adaptive framework. IEEE Transactions on Intelligent Transportation Systems 23(2):1213--1222

    2020

  5. [13]

    IEEE Transactions on Intelligent Transportation Systems 22(3):1573--1582

    Feng S, Feng Y, Yu C, Zhang Y, Liu HX (2020 c ) Testing scenario library generation for connected and automated vehicles, part i: Methodology. IEEE Transactions on Intelligent Transportation Systems 22(3):1573--1582

    2020

  6. [14]

    Nature 615(7953):620--627

    Feng S, Sun H, Yan X, Zhu H, Zou Z, Shen S, Liu HX (2023) Dense reinforcement learning for safety validation of autonomous vehicles. Nature 615(7953):620--627

    2023

  7. [15]

    Nature communications 12(1):748

    Feng S, Yan X, Sun H, Feng Y, Liu HX (2021) Intelligent driving intelligence test for autonomous vehicles with naturalistic and adversarial environment. Nature communications 12(1):748

    2021

  8. [16]

    (2014) Concept and development of a unified ontology for generating test and use-case catalogues for assisted and automated vehicle guidance

    Geyer S, Baltzer M, Franz B, Hakuli S, Kauer M, Kienle M, Meier S, Wei gerber T, Bengler K, Bruder R, et al. (2014) Concept and development of a unified ontology for generating test and use-case catalogues for assisted and automated vehicle guidance. IET Intelligent Transport ...

    2014

  9. [17]

    ://news.qq.com/rain/a/20250110A02ND800, accessed: 2025-03-24

    Huawei (2025) Huawei showcases 'parking-to-parking' autonomous driving feature. ://news.qq.com/rain/a/20250110A02ND800, accessed: 2025-03-24

    2025

  10. [18]

    Transportation research part B: methodological 166:1--18

    Li X (2022) Trade-off between safety, mobility and stability in automated vehicle following control: An analytical method. Transportation research part B: methodological 166:1--18

    2022

  11. [19]

    nature communications 15(1):4808

    Liu HX, Feng S (2024) Curse of rarity for autonomous vehicles. nature communications 15(1):4808

    2024

  12. [20]

    Random Structures & Algorithms 30(3):307--358

    Lov \'a sz L, Vempala S (2007) The geometry of logconcave functions and sampling algorithms. Random Structures & Algorithms 30(3):307--358

    2007

  13. [21]

    2023 IEEE 26th International Conference on Intelligent Transportation Systems (ITSC), 5649--5654 (IEEE)

    Ma C, Li X, Ma K, Zhang P, Long K, Chen S (2023) Trajectory-based performance ranking system of low-level automated vehicles. 2023 IEEE 26th International Conference on Intelligent Transportation Systems (ITSC), 5649--5654 (IEEE)

    2023

  14. [22]

    Machines 9(12):342

    Ma J, Che X, Li Y, Lai EMK (2021) Traffic scenarios for automated vehicle testing: A review of description languages and systems. Machines 9(12):342

    2021

  15. [23]

    Transportation Research Part C: Emerging Technologies 48:285--300

    Milan \'e s V, Shladover SE (2014) Modeling cooperative and autonomous adaptive cruise control dynamic responses using experimental data. Transportation Research Part C: Emerging Technologies 48:285--300

    2014

  16. [24]

    Safety science 121:634--650

    Moody J, Bailey N, Zhao J (2020) Public perceptions of autonomous vehicle safety: An international comparison. Safety science 121:634--650

    2020

  17. [25]

    2022 9th International Conference on Dependable Systems and Their Applications (DSA), 754--762 (IEEE)

    Ren H, Gao H, Chen H, Liu G (2022) A survey of autonomous driving scenarios and scenario databases. 2022 9th International Conference on Dependable Systems and Their Applications (DSA), 754--762 (IEEE)

    2022

  18. [26]

    2016 IEEE 19th international conference on intelligent transportation systems (ITSC), 1360--1365 (IEEE)

    Roesener C, Fahrenkrog F, Uhlig A, Eckstein L (2016) A scenario-based assessment approach for automated driving by using time series classification of human-driving behaviour. 2016 IEEE 19th international conference on intelligent transportation systems (ITSC), 1360--1365 (IEEE)

    2016

  19. [27]

    Accident Analysis & Prevention 176:106814

    Song Y, Chitturi MV, Noyce DA (2022) Intersection two-vehicle crash scenario specification for automated vehicle safety evaluation using sequence analysis and bayesian networks. Accident Analysis & Prevention 176:106814

    2022

  20. [28]

    2021 36th IEEE/ACM International Conference on Automated Software Engineering (ASE), 1342--1346 (IEEE)

    Tang Y, Zhou Y, Zhang T, Wu F, Liu Y, Wang G (2021) Systematic testing of autonomous driving systems using map topology-based scenario classification. 2021 36th IEEE/ACM International Conference on Automated Software Engineering (ASE), 1342--1346 (IEEE)

    2021

  21. [29]

    Dispon vel Acesso Dezembro

    Treiber M, Helbing D (2016) Mobil: General lane-changing model for car-following models. Dispon vel Acesso Dezembro

    2016

  22. [30]

    https://github.com/mcmtroffaes/pycddlib, version 2.1.7, accessed May 2025

    Troffaes M (2024) pycddlib: A python wrapper for cddlib. https://github.com/mcmtroffaes/pycddlib, version 2.1.7, accessed May 2025

    2024

  23. [31]

    Transportation research record 1980(1):70--78

    Ulfarsson GF, Kim S, Lentz ET (2006) Factors affecting common vehicle-to-vehicle collision types: Road safety priorities in an aging society. Transportation research record 1980(1):70--78

    2006

  24. [32]

    (2020) Scipy 1.0: fundamental algorithms for scientific computing in python

    Virtanen P, Gommers R, Oliphant TE, Haberland M, Reddy T, Cournapeau D, Burovski E, Peterson P, Weckesser W, Bright J, et al. (2020) Scipy 1.0: fundamental algorithms for scientific computing in python. Nature methods 17(3):261--272

    2020

  25. [33]

    Accident Analysis & Prevention 157:106157

    Wang C, Xie Y, Huang H, Liu P (2021) A review of surrogate safety measures and their applications in connected and automated vehicles safety modeling. Accident Analysis & Prevention 157:106157

    2021

  26. [34]

    (2022) Finding critical scenarios for automated driving systems: A systematic mapping study

    Zhang X, Tao J, Tan K, T \"o rngren M, S \'a nchez JMG, Ramli MR, Tao X, Gyllenhammar M, Wotawa F, Mohan N, et al. (2022) Finding critical scenarios for automated driving systems: A systematic mapping study. IEEE Transactions on Software Engineering 49(3):991--1026

    2022

  27. [35]

    IEEE Transactions on Intelligent Transportation Systems 19(3):733--744

    Zhao D, Huang X, Peng H, Lam H, LeBlanc DJ (2017) Accelerated evaluation of automated vehicles in car-following maneuvers. IEEE Transactions on Intelligent Transportation Systems 19(3):733--744

    2017

  28. [36]

    IEEE Transactions on Intelligent Transportation Systems 18(3):595--607

    Zhao D, Lam H, Peng H, Bao S, LeBlanc DJ, Nobukawa K, Pan CS (2016) Accelerated evaluation of automated vehicles safety in lane-change scenarios based on importance sampling techniques. IEEE Transactions on Intelligent Transportation Systems 18(3):595--607

    2016

  29. [37]

    Accident Analysis & Prevention 218:108043

    Zhou H, Ma C, Ma K, Li X (2025) Quantile-based scenario generation for automated vehicle safety evaluation. Accident Analysis & Prevention 218:108043

    2025

  30. [38]

    Scientific Data 11(1):1123

    Zhou H, Ma K, Liang S, Li X, Qu X (2024) A unified longitudinal trajectory dataset for automated vehicle. Scientific Data 11(1):1123

    2024

  31. [39]

    China Journal of Highway and Transport 32(6):1--19

    Zhu B, Zhang Px, Zhao J, Chen H, Xu Z, Zhao X, Deng W (2019) Review of scenario-based virtual validation methods for automated vehicles. China Journal of Highway and Transport 32(6):1--19

    2019

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.