Robust estimation of the dynamic origin-destination matrix for a freeway network

Most existing dynamic O-D estimation methods were based on an assumption that flow rate information at all entrances and exits of the traffic system is available and accurate. However, the reality is that such information is neither complete nor accurate at the desirable level. In contending with such issues, this study presents a dynamic O-D model that formulates the travel time distribution over a range of intervals. The proposed model has the following key features: a) it is more realistic, especially when the travel time is longer than the control unit interval, and the speed distribution exhibits a wide range of variation; b) the number of unknown variables is reduced due to the embedded travel time distribution; and c) it is more robust in response to the potential measurement errors associated with the available input data.