Effects of Vehicle Interactions on Drivers’ Lateral Decisions during Lane Changes Based on a Visual Search

The vehicle lateral movement is usually modeled as a continuous process in traffic systems research; however, the phased modeling of lateral offsets is ignored. In this paper, the dynamic lane change process was discretized based on the time interval of continuous side mirrors to establish a multistep decision model between vehicle lateral offset and vehicle interaction scenarios. This paper proposed a method for modeling the road ahead and measuring the lateral offset by using perspective projection. The lane-changing decision time window was calibrated based on the lateral offsets and the side mirror glances. Furthermore, mutual information was introduced as a reward evaluation in Markov Decision Process (MDP) modeling to analyze the heterogeneity in lateral decisions under the impact of vehicle interactions. The results indicate that the vehicle space ratio and glance duration of the side mirror significantly affect the lane-changing decision time window. The driver’s lateral decision preferences for vehicle interaction scenarios regarding relative distance and relative displacement are revealed. This paper effectively realized the integration of information theory and decision model, providing a reference for the dynamic driving space modeling and driver decision behavior.