Turning Movement Estimation in Real Time

Abstract

Fast processors offer exciting opportunities for real-time traffic monitoring. Conventional transportation planning models that assume stable and predictable travel patterns do not lend themselves to on-line traffic forecasting. This paper describes how a new traffic flow inference model has the potential to determine comprehensive flow information in real time. Its philosophical basis is borrowed from the field of operational research, where it has been used for optimizing water and electricity flows. This paper shows how road traffic turning movement flows can be estimated from link detected flows at small recurrent intervals, in real time. The paper details the formulation of the problem, outlines the structure of the data set that provides the detector data for the model input and observed turning flows for the model evaluation. The theoretical principles that define the model are described briefly. Turning movement flow estimates, at 5-min intervals, from two independent surveys are presented and analyzed. The results show an overall mean coefficient of determination (