Modeling crash risk via pre-crash trajectory analysis: Evidence from a matched case-control study

Area of Science:

• Road safety research
• Transportation engineering
• Public health

Background:

• Traffic crashes are a major global public health issue, causing significant injuries and fatalities worldwide.
• Improving road safety necessitates a comprehensive understanding of driving behaviors immediately preceding a crash.

Purpose of the Study:

• To investigate crash-related driving behaviors using a large-scale naturalistic driving dataset.
• To identify kinematic indicators associated with increased crash risk.
• To quantify the relationship between specific driving behaviors and crash likelihood.

Main Methods:

• Utilized a large-scale nationwide naturalistic driving dataset from China (545,052 crash-involved trajectories).
• Employed a matched case-control design, pairing crash trajectories with four non-crash counterparts.
• Extracted kinematic indicators (speed, acceleration, jerk, yaw rate) and applied conditional logistic regression with nonlinear and interaction terms.

Main Results:

• Root mean square of deceleration (dec_rms) was the strongest predictor, increasing crash odds over fourfold.
• Root mean square of jerk (j_rms) and mean yaw rate (yaw_mean) were positively associated with crash risk.
• Higher mean speed (v_mean) and moderate speed variability (v_cv) were linked to lower crash odds, with a U-shaped relationship for v_cv.

Conclusions:

• Trajectory-based kinematic indicators are valuable for assessing crash risk.
• Specific behaviors like high deceleration and jerk are critical risk factors.
• Findings can inform real-time safety monitoring and advanced driver-assistance systems.