This paper will present a lane choice model for urban arterial intersections. This is the first model of this kind and replaces the rule-based assignment technique of vehicles in their subsequent lanes used in state-of-the-art microscopic traffic simulators. The lane choice at intersections is modeled as a two step process: target lane choice and immediate lane selection based on the selected target lane. The target lane is the lane the driver perceives as the best to be in taking a wide range of factors and goals into account. These include path-plan considerations and lane-specific attributes and can vary with the planning capability and aggressiveness of the driver. However, maneuver to the target lane may not be possible immediately. The observed trajectories only consist of the immediate lane choices of the drivers. The choice of immediate lane is conditional on the target lane selection and affected by maneuverability considerations and aggressiveness of the driver. The parameters of the target lane and immediate lane choice models are jointly estimated with detailed vehicle trajectories. The heterogeneities of the driver population, both in terms of planning capability and aggressiveness, are explicitly taken into account in the model formulation. The estimated model is compared with a single level intersection lane choice model to demonstrate the improvements in the goodness-of-fit. The improvements are further strengthened by validation studies within the microscopic traffic simulator MITSIMLab where the simulation results using the proposed model yield better matches with observed data compared to the rule-based models.
Abstract