Modelling driving decisions: a latent plan approach

Abstract

Microscopic traffic simulation tools are becoming increasingly popular in evaluating transport options. Driving behaviour models (e.g. route choice models, lane-changing models, etc.) are essential components of these tools. The state-of-the-art driving behaviour models assume that drivers make instantaneous decisions. However, in reality, many of the driving decisions are based on a specific plan. The plan is however unobserved or latent and only the manifestations of the plan through actions are observed. Examples include selection of a target lane before execution of the lane change, choice of a merging tactic before execution of the merge. Ignoring the effect of plans in the decision framework can lead to incorrect representation of congestion in traffic simulation tools. In this article, we present a modelling methodology to address the effects of unobserved plans in the decisions of the drivers. The actions of the driver are conditional on the current plan and can be influenced by anticipation of downstream traffic conditions. The heterogeneity in decision making and planning capabilities of drivers are explicitly addressed. The methodology has been applied in developing lane-changing behaviour models with disaggregate trajectory data extracted from video recordings of an urban road using the maximum likelihood technique. Estimation results show that the latent plan models have a significantly better goodness-of-fit compared to the ‘reduced form’ models where the latent plans are ignored. The latent plan models were also found to outperform the reduced form models in validation case studies within the microscopic traffic simulator MITSIMLab.