Joint modelling of longitudinal and lateral dynamics in lane-changing maneuvers
Ang Ji, Mohsen Ramezani, David Levinson
Abstract
This study models a lane-changing event as two behaviourally interconnected phases: ‘stay’ and ‘execution’. The model considers the ‘stay’ phase duration stochastically depending on external traffic conditions. The ‘execution’ phase is modelled as the longitudinal speed profile deployed to perform the lane change. The model comprises a Bayesian survival analysis to predict the probability of the stay duration before a new ‘execution’ phase while tackling the censoring issue of survival methods. Using naturalistic vehicular trajectory data, this paper quantifies what factors influence driver behaviour in lane-keeping and lane-changing execution. The parameter estimation results demonstrate that drivers' decisions on phase transitions are influenced by surrounding conditions, lane-changing purpose, directions, and departure lanes. The findings reveal that urgency (stemming from the purpose) and patience (satisfaction with the existing situation) are the main reasons for leaving the current lane. Adequate distances and relative speeds compared with surrounding vehicles induce or dissuade acceleration behaviour during the execution phase.