Enhancing Realism in Modeling Merge Junctions in Analytical Models for System-Optimal Dynamic Traffic Assignment

The existing analytical system-optimal dynamic traffic assignment (SO-DTA) model formulated with the linear programming (LP) approach usually assumes system control over vehicles in the entire network. This property would give rise to unreasonable priorities at merge junctions that are sometimes physically impossible to realize for the given roadway configuration. In this paper, we demonstrate that models with and without considering the merge-priority ratio would exhibit very different traffic patterns and route-choice behavior. To realistically model traffic flow on a transportation network, one should properly distinguish the level of control by drivers, roadway geometry, and system providers. This paper also attempts to develop an LP module that explicitly considers the merge-priority ratio of a merge junction and can potentially be incorporated into the existing LP formulation of the SO-DTA problem based on the cell-transmission model. By more realistically modelling the behavior of vehicles at merge junctions, the obtained solution can be used as a benchmark to compare control strategies developed without explicitly considering the merge-priority ratio at merge junctions or strategies developed with heuristic approaches.

I. Background

One of The objectives of Dynamic Traffic Assignment (DTA) is to predict traffic flow patterns over time and space on a transportation network for a given set of time-varying origin–destination (OD) demands based on some predefined conditions. The system-optimal (SO) state corresponds to the condition for distributing vehicles on a network in such a way that minimizes the total vehicle travel time or total vehicle delay in the system. SO control aimed to achieve system optimum can be employed to develop control strategies used for daily operations. Moreover, traffic patterns generated under the system-optimum condition would help system providers identify the location of the potential bottleneck in a network and, thus, effectively allocate limited resources to alleviate congestion.