Pith. sign in

REVIEW 5 major objections 4 minor 35 references

RiskNet: Interaction-Aware Risk Forecasting for Autonomous Driving in Long-Tail Scenarios

T0 review · 5 major / 4 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read RiskNet claims a kinetic-energy interaction field with Doppler weighting forecasts driving risk better than TTC, THW, RSS, and NC Field in long-tail scenarios.

desk verdict RiskNet's field+trajectory-prediction combination is a plausible incremental idea, but the paper's headline claim of significant outperformance over TTC/THW/RSS/NC Field is unsupported by the purely qualitative evaluation. read the letter →

arxiv 2504.15541 v1 pith:Z4SNWW5A submitted 2025-04-22 cs.RO cs.LG

Qichao Liu , Heye Huang , Shiyue Zhao , Lei Shi , Soyoung Ahn , Xiaopeng Li This is my paper · ORCID
classification cs.ROcs.LG
keywords riskforecastinginteractionfieldautonomousdrivinggraphneuralnetworktrajectorypredictionlong-tailscenariosuncertaintymodelingsafetyassessment
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

RiskNet tries to establish that driving risk, including rare long-tail events, can be forecast as an interaction field that combines a deterministic physics-inspired term with multimodal trajectory probabilities from a graph neural network. The paper argues that classical kinematic metrics like TTC and THW miss lateral and rear threats, and that the field-based model detects conflicts earlier and with directional sensitivity. On the strength of the proposed field equations and the GNN predictor, the paper claims consistent outperformance over TTC, THW, RSS, and NC Field on highway, intersection, and roundabout benchmarks. If true, this would give autonomous driving systems a single risk representation that is interpretable, real-time, and scenario-adaptive.

What carries the argument

The load-bearing object is the interaction field of Section 4.1, culminating in Eq. (11): a sum over participants of the interaction energy E_i = (1/2) k_j C_j (m_i m_j/(m_i+m_j)) ||v_i - v_j||^2 divided by the Euclidean distance, multiplied by a longitudinal Doppler factor $\alpha$^lon and a lateral attenuation factor $\alpha$^lat = exp(-$\beta$ $sin^{2}$ $\theta$). The energy term converts relative speed and mass into hazard; the distance division spreads it spatially; the Doppler weights concentrate it in the direction of motion. The graph-neural-network predictor, inspired by MTP-GO, supplies multimodal future trajectories with probabilities, which the field then averages to produce an expected risk intensity F̃_i(p) and a time-weighted cumulative risk R_i^total. This machinery converts raw trajectory forecasts into a scalar safety signal that can be compared to classical metrics.

What would settle it

Run RiskNet with its published coefficients on a dataset that contains actual collision or near-miss events with timestamps, and check whether the risk peaks before each conflict with a higher hit rate and earlier lead time than TTC or RSS; if it does not, the claim that the field outperforms classical metrics in responsiveness is falsified.

Watch

Extended reading notes

Core claim

The paper's central claim is that the risk an autonomous vehicle faces is a spatial interaction field, not a scalar time-to-collision or headway value. The field is built from the relative kinetic energy of the ego vehicle and each surrounding agent divided by their distance, weighted by an interaction indicator, a participant danger coefficient k_j, an environmental factor C_j, and two Doppler-derived directional factors that emphasize forward threats and attenuate lateral ones. This field can be evaluated at each predicted future time step, and when the agent's future positions are replaced by a multimodal distribution from a GNN-based trajectory predictor, the expected risk becomes a probabilistic map. The paper reports that this map identifies lane-change, cut-in, and intersection conflicts earlier and more continuously than TTC, THW, RSS, or NC Field in the scenarios it displays.

Load-bearing premise

The entire framework inherits its truth from an uncalibrated hand-set formula for interaction energy, so if that formula does not match real driver risk perception or collision statistics, the claimed improvements over TTC and RSS are measuring a self-defined quantity rather than safety.

Editorial extensions

If this is right

  • If the field equations represent real hazard, autonomous vehicles can replace multiple ad hoc safety metrics with one continuous, direction-aware risk map that covers longitudinal, lateral, and rearward threats.
  • The GNN predictor's multimodal outputs turn a point-prediction safety check into a probabilistic risk map, so planning modules can reason about 'what if the other vehicle merges now' rather than only about the most likely trajectory.
  • Time-weighted cumulative risk explicitly accounts for the growth of prediction uncertainty, which could improve braking and evasive decisions in the few seconds before a conflict.
  • Because the field is computed from relative states rather than scene-specific rules, the same equations apply to highways, intersections, and roundabouts without retuning per scenario.

Reading between the lines

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

  • A direct test of the field's validity would be to calibrate k_j, C_j, and beta against labeled near-collision or collision events; the paper leaves these coefficients hand-set, so fitting them to real outcome data is the natural next step.
  • The Doppler anisotropy could be extended to non-vehicular agents such as pedestrians and cyclists by modeling their effective speed and direction, which the current lateral attenuation treats only through the angle theta.
  • Because RiskNet outputs a full probabilistic risk map, it could be coupled to an optimization-based planner that penalizes high expected risk along candidate trajectories, effectively turning risk forecasting into a planning cost.
  • The three evaluation datasets all come from German drone-recorded traffic; whether the coefficients generalize to other countries, road rules, and driving cultures is an open empirical question the paper does not address.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 4 minor

Summary. RiskNet combines a deterministic, field-theoretic risk model with a GNN-based multimodal trajectory prediction module to forecast driving risk in long-tail scenarios. The deterministic part defines interaction energy and force between the ego vehicle and surrounding participants, applies Doppler-inspired directional weights, and is extended probabilistically by weighting risk fields by predicted trajectory modes. The paper evaluates the framework on highD, inD, and rounD datasets and claims that RiskNet significantly outperforms TTC, THW, RSS, and NC Field in accuracy, responsiveness, and directional sensitivity. The evaluation, however, is almost entirely qualitative: three hand-selected scenarios are compared via color-coded high/moderate/low risk time series, and no numerical risk-comparison metrics, ground-truth risk labels, statistical tests, or parameter values are reported.

Significance. If the central claims were established, a unified interaction-aware risk field that propagates probabilistic trajectory uncertainty would be a useful component for autonomous driving safety assessment. The conceptual combination of a physics-inspired field model with learned multimodal prediction is reasonable, and the qualitative scenarios suggest the framework can produce interpretable risk visualizations. The paper does not, however, provide the quantitative evidence needed to substantiate the empirical superiority claims in the abstract and conclusions: there are no ground-truth risk labels, no numerical comparison metrics, no baselines with error bars, and no disclosure of the free parameters that enter the risk equations. The directional-sensitivity advantage is also partly built into the model by construction. The paper does not ship machine-checked proofs, reproducible code, or a parameter-free derivation, so the contribution is limited to a conceptually interesting but unvalidated modeling proposal.

major comments (5)
  1. [§5.2.3, Figs. 12–14] The central claim of the abstract and conclusions—that RiskNet significantly outperforms TTC, THW, RSS, and NC Field in accuracy, responsiveness, and directional sensitivity—is not supported by the reported comparison. The comparative evaluation consists of color-coded high/moderate/low risk time series for three hand-selected episodes, with no numerical risk values, detection-time measurements, thresholds, area-under-curve statistics, confidence intervals, or statistical tests, and no ground-truth risk labels against which accuracy is defined. Because the parameters entering Eqs. (2)–(11) and (30) are not reported, the displayed risk levels are consistent with arbitrary rescaling, so the claimed advantage over the baselines cannot be assessed.
  2. [§4.1.2, Eqs. (7)–(10)] The directional-sensitivity claim is partly enforced by construction. Equations (8)–(10) define alpha_lon and alpha_lat, and Eq. (11) multiplies every interaction-field term by these directional weights, so the risk field is larger along the ego heading by design. Demonstrating this property in selected scenarios is therefore a restatement of the model definition rather than independent evidence of empirical directional sensitivity. To support the claimed advantage, the authors should compare against an isotropic version of the same field or calibrate the directional weights against observed conflicts.
  3. [§4.1.2, Eqs. (7)–(8) and §4.2.3, Eq. (27)] There is an internal inconsistency in the definition of theta. Equations (7) and (27) define theta_ij as the angle between the velocity vectors of vehicle i and participant j, so a participant moving in the same direction ahead of the ego corresponds to theta = 0, not theta = 180 as stated in the text after Eq. (7). The claim that alpha becomes larger when theta approaches 180 degrees is therefore incorrect under the stated definition, and the direction of the Doppler-based weighting needs to be clarified before the directional behavior of the model can be interpreted.
  4. [§4.1–§4.2] The free parameters k_j, C_j, beta, v_j0, and omega_r are never assigned values, calibrated, or analyzed for sensitivity. Without these values the risk field in Eq. (11) is not fully specified, and the qualitative comparisons in Section 5.2.3 cannot be reproduced. The authors should report all parameter values, any fitting procedure, and a sensitivity analysis, or prove that the conclusions are invariant to their choices.
  5. [§5.2.2, Table 1] Table 1 reports ADE, FDE, APDE, ANLL, and FNLL for the trajectory prediction module, but only for the four displayed cases, with no test-set size, no baselines, and no error bars. The claim that the predictor provides reliable and coherent probabilistic outputs is therefore not quantitatively established.
minor comments (4)
  1. [§4.2.2, Eqs. (22)–(23)] The control input u^s in Eq. (22) is introduced but never formally defined, and the state transition function f and process noise covariance structure are not specified; this makes the EKF formulation difficult to reproduce.
  2. [§5.2.3, Figs. 12–14] The color-coded risk time series in Figs. 12–14 lack axis labels, threshold definitions, and any quantitative scale, so the reader cannot determine what high/moderate/low risk means in physical units.
  3. [Throughout] The abbreviation NC Field is used repeatedly but never defined or referenced; it should be spelled out and the corresponding method should be identified precisely.
  4. [§1.1 and §5.1] The selected scenarios are asserted to be long-tail, but no frequency or rarity analysis is provided to demonstrate that the three displayed episodes are representative of long-tail conditions rather than common traffic interactions.

Circularity Check

2 steps flagged · score 6.0 of 10

Directional-sensitivity and broader-risk claims are baked into Eqs. (11) and (28), making those 'demonstrated advantages' restatements of the construction; the core superiority comparison otherwise lacks quantitative ground truth.

  1. self definitional [Section 4.1.2, Eqs. (7)-(11); Section 5.2.1 (Fig. 7) and Section 5.2.3 (Figs. 12-14) directional-sensitivity claims]
    "Considering both longitudinal and lateral directional adjustments, the risk field formulation is updated as: F̃_ij = Σ_j I_ij · α_ij^lon · α_ij^lat E_i / sqrt(||x_i(t)-x_j(t)||^2 + ||y_i(t)-y_j(t)||^2) (11). This enhanced expression captures the directional sensitivity of risk, emphasizing forward threats while attenuating lateral ones. ... The model accurately captures both longitudinal and lateral interaction risks, achieving high directional sensitivity."

    Eqs. (8)-(10) define α_ij^lon as a max-clamped Doppler ratio in cosθ and α_ij^lat as exp(-β sin^2θ), and Eq. (11) multiplies every interaction term by their product. The anisotropy—stronger forward, weaker lateral—is therefore an input to the risk model, not an empirical finding. When Section 5 reports that RiskNet 'achieves high directional sensitivity' and displays color-coded risk maps, it is exhibiting the definition in Eqs. (7)-(11) rather than testing it against any independent ground-truth directional-risk label or numerical metric.

  2. self definitional [Section 4.2.3, Eq. (28); Section 5.2.2(2), Fig. 11]
    "𝔼[F̃_ij(p)] = Σ_l π_l · F̃_ij^l(p) (28). ... The probabilistic risk map (a-1), informed by trajectory uncertainty, captures a broader and more continuous risk region that reflects the threat posed by multiple potential paths of the high-speed vehicle."

    Eq. (28) defines the probabilistic risk impact as the probability-weighted sum of the deterministic modal risk fields F̃_ij^l(p). Whenever two predicted modes are spatially separated and have positive probability, the mixture must cover a wider and more continuous region than either single-mode deterministic map; this is a mathematical property of averaging, not a measured safety improvement. The Fig. 11 comparison therefore restates the construction in Eq. (28). Because no ground-truth collision or near-miss labels are used, 'captures a broader ... risk region' is a consequence of the expectation formula rather than an empirically validated advantage.

full rationale

RiskNet's deterministic field (Eqs. 2-11) is a new construction, and the GNN trajectory predictor is trained with an NLL loss; these components have independent content and do not depend on a self-citation chain. MTP-GO is cited as external inspiration. However, two claimed advantages are circular: (1) directional sensitivity is guaranteed by the α_lon/α_lat weights in Eq. (11), then reported as a demonstrated result; (2) the probabilistic map's broader risk region is guaranteed by the mixture definition in Eq. (28), then reported as a comparison result. The remaining 'significantly outperforms TTC/THW/RSS/NC Field' claim lacks ground-truth risk labels, numeric metrics, and statistical tests; that is a severe evidence deficiency, but on the stated rubric an unsupported empirical claim is not itself circularity. The θ=180° direction statement in Section 4.1.2 is internally inconsistent with Eq. (27), but that is a correctness issue. Overall, partial self-definitional circularity in two prominent advantages yields a score of 6.

Assumptions & free parameters 5 free parameters · 5 assumptions · 1 invented entities

The central model rests on an uncalibrated field construction with several unspecified or hand-set coefficients (k_j, C_j, v_j0, beta, omega_r) and on borrowed analogies from acoustics. The trajectory prediction module is a standard learned component, but the risk measure itself has no external grounding. This ledger shows that the paper's contribution is a proposed mechanism, not an empirically anchored one.

free parameters (5)
  • k_j = unspecified
    Inherent danger coefficient of participant j in Eq. (2). Values for pedestrians, cyclists, cars, and trucks are not reported.
  • C_j = unspecified
    Environmental constraint factor in Eq. (2) intended to capture speed limits, signals, and lane boundaries. No calibration procedure or values are given.
  • beta = 0.5 to 2 (range stated, actual value not pinned)
    Lateral decay coefficient in Eq. (9), described as 'based on empirical observations' with no fitting procedure or chosen value reported. The directional risk results depend on this choice.
  • v_j0 = unspecified
    Wave speed in the Doppler equations (6)-(8). No value, units, or physical interpretation is given, and the directional weights depend on it.
  • omega_r = unspecified
    Time-step risk weights in Eq. (30) for cumulative risk. The paper says they can be configured but does not specify values used in experiments.
assumptions (5)
  • domain assumption Traffic risk can be represented as an additive field of interaction energies and forces (Eqs. 2-5).
    No derivation from collision statistics, human risk perception, or safety theory is provided; the field is postulated as a modeling choice.
  • ad hoc to paper The acoustic Doppler effect provides a valid analogy for directional risk weighting (Eqs. 6-8).
    Risk perception is not a wave phenomenon, and the Doppler formula is borrowed heuristically without physical justification or empirical calibration.
  • ad hoc to paper Lateral risk decays exponentially with sin^2 of the angle (Eq. 9).
    The functional form and the beta range are chosen by hand; no empirical fit or theoretical argument is given.
  • domain assumption Surrounding agent futures are adequately modeled as a Gaussian mixture over a fixed number of modes via GNN/Neural ODE (Eqs. 12-25).
    This requires sufficient training data and calibrated uncertainty, but no evidence on mode count, training set size, or calibration is provided beyond four selected cases.
  • domain assumption highD, inD, and rounD contain representative long-tail scenarios for validation.
    These are naturalistic driving datasets, not specifically curated for long-tail events; no long-tail filtering or rarity analysis is described.
invented entities (1)
  • Interaction field / interaction force (F_ij)
    purpose: Quantify collision risk as a spatial field proportional to relative kinetic energy divided by distance and directionally weighted by a Doppler analogy.
    No falsifiable prediction outside this paper is provided. The field is a heuristic construction whose parameters are not calibrated against ground-truth risk labels, collisions, or human judgments.

how reviews work

0 comments
Cite this review

Pith. "Pith review of RiskNet: Interaction-Aware Risk Forecasting for Autonomous Driving in Long-Tail Scenarios." pith.science (2026). https://pith.science/paper/Z4SNWW5A

@misc{pith2026250415541,
  author       = {Pith},
  title        = {Pith review of: RiskNet: Interaction-Aware Risk Forecasting for Autonomous Driving in Long-Tail Scenarios},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Z4SNWW5A}},
  note         = {Machine review of arXiv:2504.15541}
}
read the original abstract

Ensuring the safety of autonomous vehicles (AVs) in long-tail scenarios remains a critical challenge, particularly under high uncertainty and complex multi-agent interactions. To address this, we propose RiskNet, an interaction-aware risk forecasting framework, which integrates deterministic risk modeling with probabilistic behavior prediction for comprehensive risk assessment. At its core, RiskNet employs a field-theoretic model that captures interactions among ego vehicle, surrounding agents, and infrastructure via interaction fields and force. This model supports multidimensional risk evaluation across diverse scenarios (highways, intersections, and roundabouts), and shows robustness under high-risk and long-tail settings. To capture the behavioral uncertainty, we incorporate a graph neural network (GNN)-based trajectory prediction module, which learns multi-modal future motion distributions. Coupled with the deterministic risk field, it enables dynamic, probabilistic risk inference across time, enabling proactive safety assessment under uncertainty. Evaluations on the highD, inD, and rounD datasets, spanning lane changes, turns, and complex merges, demonstrate that our method significantly outperforms traditional approaches (e.g., TTC, THW, RSS, NC Field) in terms of accuracy, responsiveness, and directional sensitivity, while maintaining strong generalization across scenarios. This framework supports real-time, scenario-adaptive risk forecasting and demonstrates strong generalization across uncertain driving environments. It offers a unified foundation for safety-critical decision-making in long-tail scenarios.

Figures

Figures reproduced from arXiv: 2504.15541 by the authors.

Figure 1
Figure 1. presents the joint distributions of key risk factors. As shown in [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 4
Figure 4. Predicted interaction risk between ego (white) and surrounding vehicles (blue) The dynamic and uncertain behavior of traffic participants, along with their interaction with AVs during planning and decision-making, plays a critical role in risk assessment [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

35 extracted references · 7 canonical work pages

  1. [3]

    Presented at the 2017 IEEE Intelligent Vehicles Symposium (IV), IEEE, Los Angeles, CA, USA, pp

    Modified artificial potential field method for online path planning applications, in: 2017 IEEE Intelligent Vehicles Symposium (IV). Presented at the 2017 IEEE Intelligent Vehicles Symposium (IV), IEEE, Los Angeles, CA, USA, pp. 180 –185. https://doi.org/10.1109/IVS.2017.7995717 Brown, T.L.,

    arXiv 2017

  2. [7]

    Presented at the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), pp

    Convolutional Social Pool ing for Vehicle Trajectory Prediction, in: 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW). Presented at the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), pp. 1549–15498. https://doi.org/10.1109/CVPRW.2018.00196 Fatality Analysis Reporting System (FARS...

    arXiv 2018

  3. [9]

    Safety Science 84, 67 –77

    On the assessment of uncertainty in risk diagrams. Safety Science 84, 67 –77. https://doi.org/10.1016/j.ssci.2015.12.001 Guo, J., Kurup, U., Shah, M.,

    doi:10.1016/j.ssci.2015.12.001 2015

  4. [11]

    Presented at the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp

    Social GAN: Socially Acceptable Trajectories with Generative Adversarial Networks, in: 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Presented at the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 2255–2264. https://doi.org/10.1109/CVPR.2018.00240 Hasuo, I.,

    arXiv 2018

  5. [12]

    https://doi.org/10.48550/arXiv.2206.03418 Huang, H., Liu, J., Zhang, B., Zhao, S., Li, B., Wang, J.,

    Responsibility -Sensitive Safety: an Introduction with an Eye to Logical Foundations and Formalization. https://doi.org/10.48550/arXiv.2206.03418 Huang, H., Liu, J., Zhang, B., Zhao, S., Li, B., Wang, J.,

    arXiv doi:10.48550/arxiv.2206.03418

  6. [13]

    IEEE Transactions on Intelligent Transportation Systems 1–15

    LEAD: Learning-Enhanced Adaptive Decision- Making for Autonomous Driving in Dynamic Environments. IEEE Transactions on Intelligent Transportation Systems 1–15. https://doi.org/10.1109/TITS.2025.3531293 Huang, H., Wang, J., Fei, C., Zheng, X., Yang, Y., Liu, J., Wu, X., Xu, Q.,

    arXiv 2025

  7. [14]

    SCIENCE CHINA Information Sciences 63, 190203

    A probabilistic risk assessment framework considering lane-changing behavior interaction. SCIENCE CHINA Information Sciences 63, 190203. https://doi.org/10.1007/s11432-019-2983-0 Karle, P., Geisslinger, M., Betz, J., Lienkamp, M.,

    doi:10.1007/s11432-019-2983-0

  8. [15]

    IEEE Transactions on Intelligent Transportation Systems 23, 16962–16982

    Scenario Understand ing and Motion Prediction for Autonomous Vehicles —Review and Comparison. IEEE Transactions on Intelligent Transportation Systems 23, 16962–16982. https://doi.org/10.1109/TITS.2022.3156011 Köhler, C.,

    arXiv 2022

Show all 35 references
  1. [18]

    IEEE Intelligent Transportation Systems Magazine 1 –1

    Threat Assessment Techniques in Intelligent Vehicles: A Comparative Survey. IEEE Intelligent Transportation Systems Magazine 1 –1. https://doi.org/10.1109/MITS.2019.2907633 Lin, H., Liu, Y., Li, S., Qu, X.,

    2019

  2. [22]

    IEEE Transactions on Knowledge and Data Engineer ing 36, 5633–5646

    Attention -Aware Social Graph Transformer Networks for Stochastic Trajectory Prediction. IEEE Transactions on Knowledge and Data Engineer ing 36, 5633–5646. https://doi.org/10.1109/TKDE.2024.3390765 Ortiz, F.M., Sammarco, M., Detyniecki, M., Costa, L.H.M.K.,

    2024

  3. [24]

    IEEE Transactions on Intelligent Transportation Sy stems 23, 17284–17300

    Recurrent Models for Lane Change Prediction and Situation Assessment. IEEE Transactions on Intelligent Transportation Sy stems 23, 17284–17300. https://doi.org/10.1109/TITS.2022.3163353 Schulz, J.C.,

    2022

  4. [25]

    IFAC -PapersOnLine 50, 15006 –15011

    Multi-objective platoon maneuvering using artificial potential fields. IFAC -PapersOnLine 50, 15006 –15011. https://doi.org/10.1016/j.ifacol.2017.08.2570 Vogel, K.,

    2017 doi

  5. [27]

    IEEE Access 10, 70662 –70674

    Acceleration -Based Collision Criticality Metric for Holistic Online Safety Assessment in Automated Driving. IEEE Access 10, 70662 –70674. https://doi.org/10.1109/ACCESS.2022.3186765 Wang, J., Huang, H., Li, K., Li, J.,

    2022

  6. [29]

    Accident Analysis & Prevention 145, 105680

    Driving risk assessment based on naturalistic driving study and driver attitude questionnaire analysis. Accident Analysis & Prevention 145, 105680. https://doi.org/10.1016/j.aap.2020.105680 Wang, J., Wu, J., Li, Y .,

    2020

  7. [31]

    Engineering 33, 90 –107

    Reachability -Based Confidence -Aware Probabilistic Collision Detection in Highway Driving. Engineering 33, 90 –107. https://doi.org/10.1016/j.eng.2023.10.010 Westny, T., Oskarsson, J., Olofsson, B., Frisk, E.,

    2023 doi

  8. [32]

    IEEE Trans

    MTP -GO: Graph -Based Probabilistic Multi -Agent Trajectory Prediction With Neural ODEs. IEEE Trans. Intell. Veh. 8, 4223 –4236. https://doi.org/10.1109/TIV.2023.3282308 Yao, H., Li, X., Li, Q., Yu, C.,

    2023

  9. [33]

    Transportation Research Part E: Logistics and Transportation Review 192, 103780

    Safety aware neural network for connected and automated vehicle operations. Transportation Research Part E: Logistics and Transportation Review 192, 103780. https://doi.org/10.1016/j.tre.2024.103780 Zhou, H., Zhong, Z.,

    2024

  10. [34]

    Accident Analysis & Prevention 148, 105798

    Evasive behavior-based method for threat assessment in different scenarios: A novel framework for intelligent vehicle. Accident Analysis & Prevention 148, 105798. https://doi.org/10.1016/j.aap.2020.105798 Zhou, W., Cao, Z., Deng, N., Liu, X., Jiang, K., Yang, D.,

    2020

  11. [35]

    IEEE Transactions on Intelligent Transportation Systems 24, 3476–3488

    Dynamically Conservative Self -Driving Planner for Long-Tail Cases. IEEE Transactions on Intelligent Transportation Systems 24, 3476–3488. https://doi.org/10.1109/TITS.2022.3225538

    2022

  12. [660]

    https://doi.org/10.3390/math9060660 Li, Y., Zheng, Y., Morys, B., Pan, S., Wang, J., Li, K.,

    doi

  13. [1796]

    https://doi.org/10.1109/JAS.2023.123744 Liu, J., Luo, Y., Xiong, H., Wang, T., Huang, H., Zhong, Z.,

    2023

  14. [2003]

    Accident Analysis & Prevention 35, 427–433

    A comparison of headway and time to collision as safety ind icators. Accident Analysis & Prevention 35, 427–433. https://doi.org/10.1016/S0001-4575(02)00022-2 Wang, C., Popp, C., Winner, H.,

    doi

  15. [2013]

    Lea rning context sensitive behavior models from observations for predicting traffic situations, in: 16th International IEEE Conference on Intelligent Transportation Systems (ITSC 2013). pp. 1764–1771. https://doi.org/10.1109/ITSC.2013.6728484 Goerlandt, F., Reniers, G.,

    2013

  16. [2015]

    IEEE Transactions on Intelligent Transportation Systems 16, 2203 –2214

    The Driving Safety Field Based on Driver –Vehicle–Road Interactions. IEEE Transactions on Intelligent Transportation Systems 16, 2203 –2214. https://doi.org/10.1109/TITS.2015.2401837 Wang, X., Li, Z., Alonso -Mora, J., Wang, M.,

    2015

  17. [2016]

    IEEE Transactions on Intelligent Transportation Systems 17, 1538–1550

    A Combined Model - and Learning-Based Framework for Interaction -Aware Maneuver Prediction. IEEE Transactions on Intelligent Transportation Systems 17, 1538–1550. https://doi.org/10.1109/TITS.2015.2506642 Bounini, F., Gingras, D., Pollart, H., Gruyer, D.,

    2015

  18. [2017]

    An LSTM network for highway trajectory prediction, in: 2017 IEEE 20th International Conference on Intelligent Transportation Systems (ITSC). pp. 353 –359. https://doi.org/10.1109/ITSC.2017.8317913 Bahram, M., Hubmann, C., Lawitzky, A., Aeberhard, M., Wollherr, D.,

    2017

  19. [2018]

    Proc Natl Acad Sci USA 115, E2879–E2887

    Integration of speed and time for estimating time to contact. Proc Natl Acad Sci USA 115, E2879–E2887. https://doi.org/10.1073/pnas.1713316115 Cheng, H., Jiang, Y., Zhang, H., Chen, K., Huang, H., Xu, S., Wang, J., Zheng, S.,

    doi

  20. [2019]

    Presented at the 2019 IEEE Intelligent Transportation Systems Conference (ITSC), pp

    An Integrated Approach to Probabilistic Vehicle Trajectory Prediction via Driver Characteristic and Intention Estimation, in : 2019 IEEE Intelligent Transportation Systems Conference (ITSC). Presented at the 2019 IEEE Intelligent Transportation Systems Conference (ITSC), pp. 3...

    2019

  21. [2020]

    IEEE Transactions on Intelligent Transportation Systems 21, 3135–3151

    Is it Safe to Drive? An Overview of Factors, Metrics, and Datasets for Driveability Assessment in Autonomous Driving. IEEE Transactions on Intelligent Transportation Systems 21, 3135–3151. https://doi.org/10.1109/TITS.2019.2926042 Gupta, A., Johnson, J., Fei-Fei, L., Savarese,...

    2019

  22. [2021]

    Engineering 7, 1313–1325

    Towards the Unified Principles for Level 5 Autonomous Vehicles. Engineering 7, 1313–1325. https://doi.org/10.1016/j.eng.2020.10.018 Wang, J., Huang, H. , Li, Y., Zhou, H., Liu, J., Xu, Q.,

    2020 doi

  23. [2022]

    Transportation Research Part C: Emerging Technologies 138, 103656

    Trustworthy safety improvement for autonomous driving using reinforcement learning. Transportation Research Part C: Emerging Technologies 138, 103656. https://doi.org/10.1016/j.trc.2022.103656 Chang, C.-J., Jazayeri, M.,

    2022

  24. [2023]

    Accident Analysis & P revention 191, 107172

    Road traffic safety assessment in self - driving vehicles based on time -to-collision with motion orientation. Accident Analysis & P revention 191, 107172. https://doi.org/10.1016/j.aap.2023.107172 Scheel, O., Nagaraja, N.S., Schwarz, L., Navab, N., Tombari, F.,

    2023

  25. [2024]

    IEEE Transactions on Intelligent Transportation S ystems 25, 17614 – 17627

    Cooperative Longitudinal Driving and Lane Assignment Strategy for Left -Turn Connected and Autonomous Vehicles at Signalized Intersections With a Contraflow Left -Turn Lane. IEEE Transactions on Intelligent Transportation S ystems 25, 17614 – 17627. https://doi.org/10.1109/TIT...

  26. [2025]

    Transportation Research Part C: Emerging Technologies 171, 104981

    Emergency Inde x (EI): A two -dimensional surrogate safety measure considering vehicles’ interaction depth. Transportation Research Part C: Emerging Technologies 171, 104981. https://doi.org/10.1016/j.trc.2024.104981 Deo, N., Trivedi, M.M.,

    2024

  27. [4850]

    https://doi.org/10.1038/s41467-020-18353-4 Leon, F., Gavrilescu, M.,

    doi

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.