Improving matrix estimation pertaining to detailed traffic information and sophisticated traffic state

Abstract

Technical innovation and extensive application of adaptive signal control at intersections have made turning flow information that provide more precise constraints for Origin-Destination matrix (O-D matrix) estimation easily available in great quantity and more accurate than ever. However, the influence of turning flow and duplication of information on the existing matrix estimation models and on the accuracy of O-D matrix estimation has not been broadly investigated. Also, traffic phenomenon in networks becomes complicated and difficult to explain with the increase in number of vehicles, variety of daily activities and sophisticated travel behaviors. As such, general congested traffic state as well as diverse travelers’ perception about travel time should be taken into consideration in O-D matrix estimation models. In this paper, the influence of applying finer and duplicated flow information as well as route choice proportion estimates on the performance of the Information minimization (IM) and the modified IM models were examined. It has shown that duplicate information has adverse effect on the accuracy of matrix estimation, whereas additional turning flow information can improve estimation accuracy. Based on the examination results a methodology using the IM model, the stochastic user equilibrium (SUE) assignment and the information screening process, was proposed to optimize the goodness of estimation and enhance the IM model to deal with the traffic situation more realistically. The respective convergence and required computation time were also examined. Furthermore, an empirical route choice study was conducted in order to help determining the size of a route set used in the SUE assignment model.