Validation of lane-changing behaviour in a weaving section using a driving simulator experiment

Abstract

Driving simulators have been extensively utilized to investigate driving behaviour. Nevertheless, concerns persist regarding their validity and reliability in accurately replicating real-world driving. No study has directly compared lane-changing behaviour in actual expressway weaving sections with that observed in a driving simulator. To address the research gap, a test track is designed to replicate the geometric characteristics of a real-world weaving section. The experiment involves 42 participants. Data on driving behaviour and performance are collected to evaluate key traffic efficiency and lane change-related variables. Kolmogorov-Smirnov test assesses the discrepancies between field observations and simulation outcomes. The results identify the statistical similarities in speed and headway across most manoeuvres, lane change duration and accepted lead gap for mandatory lane changes, as well as generalized time-to-collision ($\mathrm{GTTC}$) in discretionary lane changes. However, simulation drivers tend to drive faster particularly in ramp scenarios due to reduced speed perception. Some novice drivers maintain larger headways and spend less time on mandatory lane changes. Furthermore, the location of merging and diverging positions significantly influences the lane change duration. During simulated driving, poor rear visibility leads to more forced lane changes and larger accepted gaps. The $\mathrm{GTTC}$ values from the simulation are higher than those from field, except for discretionary lane changes, with younger drivers exhibiting smaller $\mathrm{GTTC}$ values. Lastly, the simulation data suggest challenges in replicating the dangerous scenarios resulting from the smaller accepted gaps observed in field. The findings offer support for the broader implementation of driving simulation experiments for weaving sections.