pith. sign in

arxiv: 2604.02573 · v1 · submitted 2026-04-02 · 📡 eess.SY · cs.SY· eess.SP

Dynamic Risk Generation for Autonomous Driving: Naturalistic Reconstruction of Vehicle-E-Scooter Interactions

Abin Mathew , Zhitong He , Lingxi Li , Yaobin Chen This is my paper

Pith reviewed 2026-05-13 20:28 UTC · model grok-4.3

classification 📡 eess.SY cs.SYeess.SP
keywords autonomous drivinge-scootersocial force modelrisk generationcollision avoidancesimulationnaturalistic datamotion planning
0
0 comments X

The pith

Real traffic data combined with a social force model generates synthetic e-scooter interactions that reach potential collision levels for testing autonomous vehicles.

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

The paper develops a reconstruction pipeline that takes collected on-road vehicle and e-scooter trajectories and modifies the e-scooter component with a social force model to produce more dynamic and riskier movements than those observed in the original recordings. This matters because autonomous driving systems must prove reliable against uncertain micromobility behaviors that current datasets rarely capture at critical moments. The method keeps the interactions grounded in real data while allowing the risk level to be increased in a controlled way. A case study on one recorded encounter shows the pipeline can be configured to reach near-miss conditions. If successful, vehicle motion planners can be validated against harder test cases without exposing real road users to danger.

Core claim

The paper presents a pipeline that utilizes collected on-road traffic data to reconstruct naturalistic vehicle-e-scooter interactions and applies a Social Force Model to the e-scooter to generate more dynamic and potentially risky movements, thereby creating configurable synthetic scenarios that extend the original interactions to more critical cases that may result in potential collision.

What carries the argument

Social Force Model applied to e-scooter motion inside a data-driven reconstruction pipeline that produces configurable synthetic interactions from real trajectories.

If this is right

  • Vehicle collision avoidance systems can be exposed to more critical interaction scenarios than those present in raw traffic recordings.
  • The simulator supports configurable risk levels so that motion planning algorithms can be tested systematically across a range of danger.
  • A case study on one real-world interaction verifies that the pipeline produces usable synthetic outputs.
  • The generated scenarios remain anchored in naturalistic data while reaching conditions that may lead to collision.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same reconstruction approach could be applied to other micromobility users such as cyclists or pedestrians to fill gaps in existing safety datasets.
  • If the generated critical cases prove representative, they could be used to define quantitative safety thresholds for approval of autonomous systems in mixed traffic.
  • Integration of the pipeline with standard driving simulators would allow broader coverage of edge-case interactions during algorithm development.

Load-bearing premise

The Social Force Model, when used on e-scooter motion, produces behaviors representative enough of real risky interactions to serve as valid test cases for vehicle collision avoidance.

What would settle it

A controlled experiment in which an autonomous vehicle planner is run on both the original recorded interaction and the SFM-extended version and shows no measurable increase in required avoidance effort or failure rate.

Figures

Figures reproduced from arXiv: 2604.02573 by Abin Mathew, Lingxi Li, Yaobin Chen, Zhitong He.

Figure 1
Figure 1. Figure 1: General pipeline for VEI scenario generation from naturalistic driving [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Vehicle and E-Scooter Interaction Scenario Reconstruction System Architecture [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Vehicle Longitudinal Control Flow Chart The speed-dependent safe distance is defined as dsafe = max v 2 veh 2|amin| , Tsafe vveh + dbuf, (4) where vveh is the ego speed, amin < 0 is the maximum braking deceleration, Tsafe is a time–headway parameter (set to 5 s), and dbuf is a fixed standstill buffer in front of the e-scooter. If the e-scooter is in front of the ego vehicle and the Euclidean gap d is sma… view at source ↗
Figure 4
Figure 4. Figure 4: Reconstructed Scenario configured to induce collision (Top: Bird [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: GPS information of the example VEI scenario [PITH_FULL_IMAGE:figures/full_fig_p005_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Speed profile for ego vehicle across five synthetic VEI scenarios. [PITH_FULL_IMAGE:figures/full_fig_p005_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Standard deviation of ego vehicle speed across five scenarios. [PITH_FULL_IMAGE:figures/full_fig_p006_7.png] view at source ↗
read the original abstract

The increasing, high-risk interactions between vehicles and vulnerable micromobility users, such as e-scooter riders, challenge vehicular safety functions and Automated Driving (AD) techniques, often resulting in severe consequences due to the dynamic uncertainty of e-scooter motion. Despite advances in data-driven AD methods, traffic data addressing the e-scooter interaction problem, particularly for safety-critical moments, remains underdeveloped. This paper proposes a pipeline that utilizes collected on-road traffic data and creates configurable synthetic interactions for validating vehicle motion planning algorithms. A Social Force Model (SFM) is applied to offer more dynamic and potentially risky movements for the e-scooter, thereby testing the functionality and reliability of the vehicle collision avoidance systems. A case study based on a real-world interaction scenario was conducted to verify the practicality and effectiveness of the established simulator. Simulation experiments successfully demonstrate the capability of extending the target scenario to more critical interactions that may result in a potential collision.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper proposes a pipeline that reconstructs naturalistic vehicle-e-scooter interactions from collected on-road traffic data and augments the e-scooter trajectories with a Social Force Model (SFM) to generate more dynamic and potentially risky movements. A case study on a real-world scenario is used to demonstrate that the simulator can extend the interaction to more critical cases that may result in potential collisions, with the goal of providing configurable synthetic data for validating vehicle motion planning and collision avoidance algorithms in autonomous driving.

Significance. If the SFM-augmented scenarios are shown to produce kinematically realistic and representative risk metrics, the work could provide a practical method for addressing the scarcity of safety-critical micromobility data, enabling more rigorous testing of AD systems in high-uncertainty interactions. The combination of data-driven reconstruction with physics-based augmentation is a reasonable direction for generating edge-case test scenarios.

major comments (2)
  1. [Case Study] Case Study section: The claim that simulation experiments 'successfully demonstrate the capability of extending the target scenario to more critical interactions' is unsupported because no quantitative metrics (e.g., distributions of time-to-collision, minimum distances, or trajectory deviation from real data), error analysis, or direct comparison to observed critical maneuvers are reported. Without these, the verification of practicality and effectiveness cannot be assessed.
  2. [Methodology] Social Force Model application (Methodology): The SFM is applied to e-scooter motion without any reported calibration or validation against real e-scooter trajectory data. Since the original SFM was developed for pedestrian crowds and e-scooters exhibit different speeds, acceleration profiles, and control dynamics, the generated 'risky' behaviors lack grounding; the central claim that these constitute valid test cases for collision avoidance therefore does not follow from the presented evidence.
minor comments (2)
  1. [Abstract] Abstract: The statement that experiments 'successfully demonstrate' the extension to critical interactions should be accompanied by at least one concrete quantitative outcome to avoid overstatement.
  2. [Throughout] Notation and terminology: Ensure consistent definition of risk metrics (e.g., TTC, minimum distance) when describing SFM outputs; currently these appear only qualitatively.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed comments, which help clarify the strengths and limitations of our work. We address each major comment below and outline the revisions we will make to strengthen the manuscript.

read point-by-point responses

  1. Referee: [Case Study] Case Study section: The claim that simulation experiments 'successfully demonstrate the capability of extending the target scenario to more critical interactions' is unsupported because no quantitative metrics (e.g., distributions of time-to-collision, minimum distances, or trajectory deviation from real data), error analysis, or direct comparison to observed critical maneuvers are reported. Without these, the verification of practicality and effectiveness cannot be assessed.

    Authors: We agree that the case study as presented relies primarily on qualitative illustration of the pipeline extending a real-world scenario to higher-risk outcomes. To provide stronger evidence, we will revise the Case Study section to include quantitative metrics, specifically distributions of time-to-collision and minimum distances across the original and SFM-augmented trajectories, along with trajectory deviation measures relative to the collected data. This addition will allow direct assessment of how the generated scenarios compare to observed interactions and better substantiate the claims of practicality. revision: yes

  2. Referee: [Methodology] Social Force Model application (Methodology): The SFM is applied to e-scooter motion without any reported calibration or validation against real e-scooter trajectory data. Since the original SFM was developed for pedestrian crowds and e-scooters exhibit different speeds, acceleration profiles, and control dynamics, the generated 'risky' behaviors lack grounding; the central claim that these constitute valid test cases for collision avoidance therefore does not follow from the presented evidence.

    Authors: The referee is correct that the original SFM was developed for pedestrians and that e-scooters have distinct dynamics. In the manuscript, we selected and tuned SFM parameters using observed speed and acceleration ranges from our collected on-road e-scooter data to produce more dynamic motions. However, we did not include a formal calibration procedure or validation against held-out e-scooter trajectories. In the revised version we will expand the Methodology section with an explicit description of the parameter adaptation process, add a limitations discussion on the lack of dedicated e-scooter calibration, and include a sensitivity analysis on key SFM parameters to demonstrate robustness of the generated risk scenarios. These changes will better ground the approach while acknowledging the current limitations. revision: partial

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper's pipeline starts from externally collected real-world on-road traffic data, applies the standard Social Force Model (an established external model not originated or fitted inside this work) to generate synthetic e-scooter interactions, and evaluates the output via case-study simulation. No equations, parameters, or claims reduce the generated critical interactions or risk metrics back to quantities defined by construction within the paper itself. No self-citations, uniqueness theorems, or ansatzes from the authors are invoked as load-bearing steps. The central demonstration therefore rests on the external applicability of SFM rather than any tautological re-derivation of its own inputs.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the applicability of the Social Force Model to e-scooter dynamics and the assumption that synthetic extensions preserve naturalistic properties.

free parameters (1)
  • SFM force parameters
    Parameters controlling repulsive and attractive forces in the Social Force Model are likely tuned or selected to match observed behaviors.
axioms (1)
  • domain assumption Social Force Model accurately represents dynamic e-scooter rider behavior in interactions
    Invoked when applying SFM to generate risky movements from real data.

pith-pipeline@v0.9.0 · 5471 in / 1103 out tokens · 64695 ms · 2026-05-13T20:28:10.829451+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

24 extracted references · 24 canonical work pages

  1. [1]

    United States Consumer Product Safety Commission, “E-scooter and e-bike injuries soar: 2022 injuries increased nearly 21

    work page 2022

  2. [2]

    Electric scooter and bike accidents are soaring across the u.s

    University of California San Francisco, “Electric scooter and bike accidents are soaring across the u.s.” 2024, [Online]. Available: https://www.ucsf.edu/news/2024/07/428096/ electric-scooter-and-bike-accidents-are-soaring-across-us

    work page 2024

  3. [3]

    A survey on multimodal large language models for autonomous driving,

    C. Cui, Y . Ma, X. Cao, W. Ye, Y . Zhou, K. Liang, J. Chen, J. Lu, Z. Yang, K.-D. Liao, T. Gao, E. Li, K. Tang, Z. Cao, T. Zhou, A. Liu, X. Yan, S. Mei, J. Cao, Z. Wang, and C. Zheng, “A survey on multimodal large language models for autonomous driving,” inProceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) Workshops, J...

    work page 2024

  4. [4]

    Chen and L

    Y . Chen and L. Li,Advances in Intelligent Vehicles. Academic Press, 2013

    work page 2013

  5. [5]

    Future directions of intelligent vehicles: Potentials, possibilities, and perspectives,

    D. Cao, X. Wang, L. Li, C. Lv, X. Na, Y . Xing, X. Li, Y . Li, Y . Chen, and F.-Y . Wang, “Future directions of intelligent vehicles: Potentials, possibilities, and perspectives,”IEEE Transactions on Intelligent Vehi- cles, vol. 7, no. 1, pp. 7–10, 2022

    work page 2022

  6. [6]

    Comparing the characteristics of electric scooter and bicycle injuries: a retrospective cohort study,

    L. Seitakari, O. Pakarinen, V .-P. Harjola, M. Castr ´en, and H. Vasara, “Comparing the characteristics of electric scooter and bicycle injuries: a retrospective cohort study,”Scientific Reports, vol. 15, no. 1, p. 27599, 2025

    work page 2025

  7. [7]

    Machine learning in micromobility: A systematic review of datasets, techniques, and applications,

    S. Yan, C. Kaundanya, N. E. O’Connor, S. Little, and M. Liu, “Machine learning in micromobility: A systematic review of datasets, techniques, and applications,”arXiv preprint arXiv:2508.16135, 2025

    work page arXiv 2025

  8. [8]

    Performance evaluation of real-time object detection for electric scooters,

    D. Chen, A. Hosseini, A. Smith, A. F. Nikkhah, A. Heydarian, O. Shoghli, and B. Campbell, “Performance evaluation of real-time object detection for electric scooters,”arXiv preprint arXiv:2405.03039, 2024

    work page arXiv 2024

  9. [9]

    Intent-guided trajectory prediction for e-scooter riders and bicyclists,

    Z. Zhang, Z. Ding, Y . Chen, S. Chien, L. Li, R. Sheroy, J. Domeyer, and R. Tian, “Intent-guided trajectory prediction for e-scooter riders and bicyclists,” in2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC). IEEE, 2024, pp. 272–277

    work page 2024

  10. [10]

    Multi-trajectory prediction for e- scooter riders with multi-modal inputs,

    Z. Zhang, J. Peng, Y . Chen, J. Zhou, S. Chien, Z. Ding, R. Sherony, Z. Sun, S. Wang, and R. Tian, “Multi-trajectory prediction for e- scooter riders with multi-modal inputs,” in2025 IEEE 28th International Conference on Intelligent Transportation Systems (ITSC). IEEE, 2025, pp. 3358–3364

    work page 2025

  11. [11]

    Carla: An open urban driving simulator,

    A. Dosovitskiy, G. Ros, F. Codevilla, A. Lopez, and V . Koltun, “Carla: An open urban driving simulator,” inConference on robot learning. PMLR, 2017, pp. 1–16

    work page 2017

  12. [12]

    Scendd: A scenario-based naturalistic driving dataset,

    A. Prabu, N. Ranjan, L. Li, R. Tian, S. Chien, Y . Chen, and R. Sherony, “Scendd: A scenario-based naturalistic driving dataset,” in2022 IEEE 25th International Conference on Intelligent Transportation Systems (ITSC). IEEE, 2022, pp. 4363–4368

    work page 2022

  13. [13]

    Scendd++: An enhanced scenario-based naturalistic driving dataset,

    N. Ranjan, A. Prabu, Z. He, and L. Li, “Scendd++: An enhanced scenario-based naturalistic driving dataset,” in2024 International Con- ference on Cyber-Physical Social Intelligence (ICCSI). IEEE, 2024, pp. 1–6

    work page 2024

  14. [14]

    A coupled driving simulator to investigate the interaction between bicycles and automated vehicles,

    J. Lindner, A. Keler, G. Grigoropoulos, P. Malcolm, F. Denk, P. Brunner, and K. Bogenberger, “A coupled driving simulator to investigate the interaction between bicycles and automated vehicles,” in2022 IEEE 25th International Conference on Intelligent Transportation Systems (ITSC). IEEE, 2022, pp. 1335–1341

    work page 2022

  15. [15]

    Simulation framework for vehicle and electric scooter interaction,

    Z. He and L. Li, “Simulation framework for vehicle and electric scooter interaction,” in2023 IEEE 26th International Conference on Intelligent Transportation Systems (ITSC), 2023, pp. 4479–4484

    work page 2023

  16. [16]

    Scenario-based trajectory generation and density estimation towards risk analysis of autonomous vehicles,

    E. Johansson, M. S ¨onnergaard, S. Rahrovani, P. Basimfaret al., “Scenario-based trajectory generation and density estimation towards risk analysis of autonomous vehicles,” in2023 IEEE 26th International Conference on Intelligent Transportation Systems (ITSC). IEEE, 2023, pp. 1375–1380

    work page 2023

  17. [17]

    Automatically improving sce- nario descriptions derived from recorded traffic,

    N. Kolb, F. Huber, and A. Pretschner, “Automatically improving sce- nario descriptions derived from recorded traffic,”IEEE Transactions on Intelligent Vehicles, 2024

    work page 2024

  18. [18]

    Assessing the completeness of traffic scenario categories for automated highway driving functions via cluster-based analysis,

    N. Roßberg, M. Neumeier, S. Hasirlioglu, M. E. Bouzouraa, and M. Botsch, “Assessing the completeness of traffic scenario categories for automated highway driving functions via cluster-based analysis,” in 2025 IEEE Intelligent Vehicles Symposium (IV), 2025, pp. 1787–1794

    work page 2025

  19. [19]

    A diffusion-tgan framework for spatio-temporal speed imputation and tra- jectory reconstruction,

    Y . Qian, X. Li, J. Zhang, X. Meng, Y . Li, H. Ding, and M. Wang, “A diffusion-tgan framework for spatio-temporal speed imputation and tra- jectory reconstruction,”IEEE Transactions on Intelligent Transportation Systems, 2025

    work page 2025

  20. [20]

    Fine-grained trajectory reconstruction by microscopic traffic simulation with dynamic data- driven evolutionary optimization,

    H. Naing, W. Cai, J. Yu, J. Zhong, and L. Yu, “Fine-grained trajectory reconstruction by microscopic traffic simulation with dynamic data- driven evolutionary optimization,”IEEE Transactions on Intelligent Transportation Systems, 2024

    work page 2024

  21. [21]

    Sumo2unity: An open-source traffic co-simulation tool to improve road safety,

    A. Mohammadi, P. Y . Park, M. Nourinejad, M. S. B. Cherakkatil, and H. S. Park, “Sumo2unity: An open-source traffic co-simulation tool to improve road safety,” in2024 IEEE Intelligent Vehicles Symposium (IV). IEEE, 2024, pp. 2523–2528

    work page 2024

  22. [22]

    Adaptation and calibration of a social force based model to study interactions between electric scooters and pedestrians,

    Y . Valero, A. Antonelli, Z. Christoforou, N. Farhi, B. Kabalan, C. Gi- oldasis, and N. Foissaud, “Adaptation and calibration of a social force based model to study interactions between electric scooters and pedestrians,” in2020 IEEE 23rd International Conference on Intelligent Transportation Systems (ITSC). IEEE, 2020, pp. 1–7

    work page 2020

  23. [23]

    Risk analysis in vehicle and electric scooter interaction,

    Z. He, Y . Guo, Y . Chen, B. King, and L. Li, “Risk analysis in vehicle and electric scooter interaction,” in2024 IEEE Intelligent Vehicles Symposium (IV), 2024, pp. 1316–1322

    work page 2024

  24. [24]

    Driver’s manual - chapter 8: Defensive driving,

    New York State Department of Motor Vehicles, “Driver’s manual - chapter 8: Defensive driving,” 2011, [Online]. Available: https://dmv. ny.gov/about-dmv/chapter-8-defensive-driving

    work page 2011