Traffic Assignment Based on Parsimonious Data: The Ideal Flow Network

Abstract

Traffic assignment models are used to estimate and distribute flows in a road network so that congestion and travel time delay are minimized. However, most of these models require the costly origin-destination (OD) data. In this paper, through the ideal flow network (IFN) and the maximum entropy principle, traffic flows can be estimated even with limited road network data. The macro level OD matrix is mathematically transformed into micro level OD on each intersection, which yields richer transportation structure data such as capacity and lane width. When data available is limited to only a few links or intersections, the remaining missing data can be set by computing for the stochastic matrix from the capacity ratio. Even if the capacity ratio is not available, the connectivity of the road network structure can be used to derive the stochastic matrix. With this characteristic of the IFN, even with limited and parsimonious data that can be collected from any link or intersection using a video camera, GPS, or any ITS sensing device, link flows of an entire network can be updated dynamically. Link flow results using IFN are almost the same as actual results, which is illustrated using the Sioux Falls transportation network.