Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Integrating Naturalistic Insights in Objective Multi-Vehicle Safety Framework

Enrico Del Re, Amirhesam Aghanouri, Cristina Olaverri-Monreal

TL;DR

This work tackles the challenge of assessing traffic safety in mixed-vehicle environments by bridging objective surrogate safety measures with subjective human perception. It introduces a multi-vehicle safety framework that combines SSMs through a grid-based aggregation or an autoencoder–based MAE, and further incorporates a 2D safety-field–like treatment with position weighting for surrounding vehicles. Evaluations on naturalistic HighD and IAMCV highway data show that integrating all surrounding vehicles with grid-based SSMs yields the most stable and frequent positive correlations between safety-risk gradients and human driving responses, while AE-based risk offers high average correlation with greater variability. The framework provides a foundation for SSM-driven safety-field modeling that aligns more closely with human perception and can extend to scenarios requiring robust safety reasoning, such as hydrogen-related vehicle incidents.

Abstract

As autonomous vehicle technology advances, the precise assessment of safety in complex traffic scenarios becomes crucial, especially in mixed-vehicle environments where human perception of safety must be taken into account. This paper presents a framework designed for assessing traffic safety in multi-vehicle situations, facilitating the simultaneous utilization of diverse objective safety metrics. Additionally, it allows the integration of subjective perception of safety by adjusting model parameters. The framework was applied to evaluate various model configurations in car-following scenarios on a highway, utilizing naturalistic driving datasets. The evaluation of the model showed an outstanding performance, particularly when integrating multiple objective safety measures. Furthermore, the performance was significantly enhanced when considering all surrounding vehicles.

Integrating Naturalistic Insights in Objective Multi-Vehicle Safety Framework

TL;DR

This work tackles the challenge of assessing traffic safety in mixed-vehicle environments by bridging objective surrogate safety measures with subjective human perception. It introduces a multi-vehicle safety framework that combines SSMs through a grid-based aggregation or an autoencoder–based MAE, and further incorporates a 2D safety-field–like treatment with position weighting for surrounding vehicles. Evaluations on naturalistic HighD and IAMCV highway data show that integrating all surrounding vehicles with grid-based SSMs yields the most stable and frequent positive correlations between safety-risk gradients and human driving responses, while AE-based risk offers high average correlation with greater variability. The framework provides a foundation for SSM-driven safety-field modeling that aligns more closely with human perception and can extend to scenarios requiring robust safety reasoning, such as hydrogen-related vehicle incidents.

Abstract

As autonomous vehicle technology advances, the precise assessment of safety in complex traffic scenarios becomes crucial, especially in mixed-vehicle environments where human perception of safety must be taken into account. This paper presents a framework designed for assessing traffic safety in multi-vehicle situations, facilitating the simultaneous utilization of diverse objective safety metrics. Additionally, it allows the integration of subjective perception of safety by adjusting model parameters. The framework was applied to evaluate various model configurations in car-following scenarios on a highway, utilizing naturalistic driving datasets. The evaluation of the model showed an outstanding performance, particularly when integrating multiple objective safety measures. Furthermore, the performance was significantly enhanced when considering all surrounding vehicles.
Paper Structure (11 sections, 4 equations, 6 figures, 6 tables)

This paper contains 11 sections, 4 equations, 6 figures, 6 tables.

Table of Contents

  1. INTRODUCTION
  2. Related Work
  3. Traffic Scenario and Dataset description
  4. Traffic scenario
  5. Datasets
  6. Safety Model
  7. Metrics
  8. Safety risk estimation
  9. Model evaluation
  10. Model Evaluation Results
  11. Conclusion and Future Work

Figures (6)

  • Figure 1: Illustration of the safety analysis of the examined traffic scenario. Each ego vehicle (EV) is evaluated along with its leading vehicle (LV), following vehicle (FV), and parallel-moving vehicles (P).
  • Figure 2: The computation of Post Encroachment Time (PET) for parallel-moving vehicles involves assessing the lateral velocity component of $v_{PL}$. If this component signals a potential encroachment, defined by more than half of the width of the parallel vehicle crossing into the lane of the ego vehicle (EV) at the encroachment point, the PET can be estimated.
  • Figure 3: Means and standard deviations of the statistically significant Spearman coefficients for various SSM weights and AEs from Table \ref{['tab:SSM_weights']}. Weight combinations without PET are unable to evaluate parallel vehicles.
  • Figure 4: Means and standard deviations of the Spearman coefficient for various model configurations $1a3g$ on the HighD dataset, indicating the correlation between jerk and the gradient of the safety risk.
  • Figure 5: Car following scenario on the highway from the IAMCV dataset.
  • ...and 1 more figures