ITSC 2025 Paper Abstract

Paper FR-LA-T36.1

Tse-Yi, Kuo (National Yang Ming Chiao Tung University), Yong-Ci, Chang (National Yang Ming Chiao Tung University), Fu-You, Huang (National Yang Ming Chiao Tung University), Hsueh-Yi, Lai (National Yang Ming Chiao Tung University)

Machine Learning Classification of Crash Risk During Takeover Based on Driver Readiness and Scenario Features

Scheduled for presentation during the Regular Session "S36c-Behavior Modeling and Decision-Making in Traffic Systems" (FR-LA-T36), Friday, November 21, 2025, 16:00−16:20, Surfers Paradise 3

2025 IEEE 28th International Conference on Intelligent Transportation Systems (ITSC), November 18-21, 2025, Gold Coast, Australia

This information is tentative and subject to change. Compiled on October 18, 2025

Keywords Driver Behavior Monitoring and Feedback Systems for Semi-autonomous Vehicles, Ethical Decision Making in Autonomous and Semi-autonomous Vehicles, Human-Machine Interaction Systems for Enhanced Driver Assistance and Safety

Abstract

Despite the promise of conditionally automated driving, drivers remain responsible for taking control when the system reaches its operational limits. Takeover is a complex task influenced by driver readiness—often affected by non-driving-related tasks (NDRTs)—and by the specific characteristics of the driving scenario. Furthermore, takeover involves not only reaction time but also post-takeover control, both essential for assessing takeover quality. Thus, this study focuses on crash occurrence as an integrated outcome of these factors. A driving simulator experiment was conducted with 64 participants, featuring varied time budgets, road curvatures, and obstacle positions. Driver readiness was manipulated through NDRTs to induce different levels of cognitive and physical readiness, measured primarily via eye-tracking. Using XGBoost, crash occurrence was successfully classified with an F1-score of 0.9389. Feature importance analysis revealed that scenarios involving curved roads and limited maneuvering space (e.g., fewer lanes) are more prone to crash occurrence. Physical readiness emerged as a particularly critical factor in mitigating these risks. Therefore, future automated vehicles should integrate real-time assessments of both driver state and driving environment to enable adaptive and context-aware driver assistance systems.