Risk-Adaptive Visual Attention Mechanisms Under Dynamic Lane Change Risk: Evidence from a Naturalistic Eye-Tracking Field Study

Understanding drivers’ visual attention adaptation under dynamic lane change risk supports human-centric interaction design. We synchronized eye-tracking metrics, vehicle motion logs, and front-view videos from 36 professional drivers in a naturalistic field experiment. A computer vision-based framework was proposed to extract surrounding vehicle kinematics from monocular footage, and a Modified Lane Change Risk Index was developed to quantify dynamic risk. Results reveal directional heterogeneity, with right lane changes involving conservative scanning and significantly prolonged fixations on the initial lane compared with leftward maneuvers. As risk escalated, drivers prioritized safety-critical areas, with increased saccade frequency and shortened fixation durations. In high-risk right lane changes, the Near Lane Line served as a visual anchor for trajectory correction. High-risk states persisted from execution into stabilization, while drivers tended to lock their gaze on the target lane and neglect side mirror checks. These findings inform risk-adaptive human-machine interfaces and human-centric co-driving systems.