Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Research on a Driver's Perceived Risk Prediction Model Considering Traffic Scene Interaction

Chenhao Yang, Siwei Huang, Chuan Hu

TL;DR

This work addresses driver perceived risk in conditional autonomous driving by integrating subjective driver traits with objective ego-vehicle motion and environmental factors. It introduces a risk-field-informed deep learning framework that uses LSTM and a cross-attention mechanism to model interactions among traffic participants, paired with a clustering-based personalized modeling strategy. A driving-video scoring experiment validates the approach, achieving up to a $10$–$11 ext{ } ext{percent}$ improvement in AUC over baselines and demonstrating strong predictive accuracy across five risk levels. The framework offers practical implications for human-machine interaction and adaptive safety controls in intelligent driving systems, with data-driven guidance for warning systems and safety algorithms.

Abstract

In the field of conditional autonomous driving technology, driver perceived risk prediction plays a crucial role in reducing traffic risks and ensuring passenger safety. This study introduces an innovative perceived risk prediction model for human-machine interaction in intelligent driving systems. The model aims to enhance prediction accuracy and, thereby, ensure passenger safety. Through a comprehensive analysis of risk impact mechanisms, we identify three key categories of factors, both subjective and objective, influencing perceived risk: driver's personal characteristics, ego-vehicle motion, and surrounding environment characteristics. We then propose a deep-learning-based risk prediction network that uses the first two categories of factors as inputs. The network captures the interactive relationships among traffic participants in dynamic driving scenarios. Additionally, we design a personalized modeling strategy that incorporates driver-specific traits to improve prediction accuracy. To ensure high-quality training data, we conducted a rigorous video rating experiment. Experimental results show that the proposed network achieves a 10.0% performance improvement over state-of-the-art methods. These findings suggest that the proposed network has significant potential to enhance the safety of conditional autonomous driving systems.

Research on a Driver's Perceived Risk Prediction Model Considering Traffic Scene Interaction

TL;DR

This work addresses driver perceived risk in conditional autonomous driving by integrating subjective driver traits with objective ego-vehicle motion and environmental factors. It introduces a risk-field-informed deep learning framework that uses LSTM and a cross-attention mechanism to model interactions among traffic participants, paired with a clustering-based personalized modeling strategy. A driving-video scoring experiment validates the approach, achieving up to a – improvement in AUC over baselines and demonstrating strong predictive accuracy across five risk levels. The framework offers practical implications for human-machine interaction and adaptive safety controls in intelligent driving systems, with data-driven guidance for warning systems and safety algorithms.

Abstract

In the field of conditional autonomous driving technology, driver perceived risk prediction plays a crucial role in reducing traffic risks and ensuring passenger safety. This study introduces an innovative perceived risk prediction model for human-machine interaction in intelligent driving systems. The model aims to enhance prediction accuracy and, thereby, ensure passenger safety. Through a comprehensive analysis of risk impact mechanisms, we identify three key categories of factors, both subjective and objective, influencing perceived risk: driver's personal characteristics, ego-vehicle motion, and surrounding environment characteristics. We then propose a deep-learning-based risk prediction network that uses the first two categories of factors as inputs. The network captures the interactive relationships among traffic participants in dynamic driving scenarios. Additionally, we design a personalized modeling strategy that incorporates driver-specific traits to improve prediction accuracy. To ensure high-quality training data, we conducted a rigorous video rating experiment. Experimental results show that the proposed network achieves a 10.0% performance improvement over state-of-the-art methods. These findings suggest that the proposed network has significant potential to enhance the safety of conditional autonomous driving systems.

Paper Structure

This paper contains 16 sections, 8 equations, 8 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Current Research on Data for Perceived Risk Modeling
  3. Current Research on Perceived Risk Modeling Methods
  4. Contributes of This Study
  5. The Structure of The Perceived Risk Prediction Model
  6. Feature Selection
  7. Personalized Modeling Strategy
  8. Network Architecture of the Perceived Risk Prediction Model
  9. Experiment Design
  10. Objective and Deployment
  11. Detailed Design
  12. Analysis and Discussion of Experimental Results
  13. Significance Analysis of The Impact of Modeling Features on Perceived Risk
  14. Validation of the Effectiveness of the Deep Learning Network Architecture
  15. Validation of the Effectiveness of the Personalized Modeling Strategy
  16. ...and 1 more sections

Figures (8)

  • Figure 1: Perceived risk prediction model framework.
  • Figure 2: Perceived risk prediction network structure.
  • Figure 3: Experiment process display.
  • Figure 4: Rating and video material interface.
  • Figure 5: Rating and video material interface.
  • ...and 3 more figures