Modeling and Probing Strategy for Intelligent Transportation System Utilizing Lagrangian Traffic Data.

Abstract

Traffic flow model that provides accurate traffic prediction can be beneficial for traffic congestion management. Macroscopic traffic flow models were used in the past to incorporate probe vehicle data and to provide real-time traffic information, but the data collection has not been done efficiently. Also, prediction of traffic state, especially for unexpected traffic jam, is needed to compensate latency in data processing and to provide advance warning to the driver. The objective of this dissertation is to develop an analytical tool which predicts congested highway traffic by utilizing macroscopic traffic flow model and strategically collecting data from probing vehicles with real-time update.

First, Newtonian relaxation method is used to incorporate probing data into the LWR model in Eulerian coordinates for traffic status estimation. An investigation of probe vehicle deployment optimization is used to reveal the trade-off between the quality of traffic estimation and the probing cost. Synthetic data is used for numerical experiment, and Genetic algorithm is used to solve the optimization problem. The result indicates that optimal deployment of probe vehicle can reduce probing cost and estimation error by efficient usage of probe vehicles. It is possible to decrease probing data for congested traffic with negligible degradation on the quality of traffic estimation.

Second, a stochastic Lagrangian macroscopic traffic flow model is formulated which update the prediction of model parameters and traffic state with unscented Kalman filter in real-time. The proposed probing method tracks vehicles in pairs and utilizes loop detector data for additional information as needed. The model is validated with two sets of empirical data to demonstrate its capability of providing short-term prediction and using model parameter to detect traffic jam i advance. An adaptive probing scheme is presented to show that adjusting probing cell size based on the variance from stochastic model can improve the prediction accuracy.

This dissertation proposed a stochastic Lagrangian traffic flow model with the capability of traffic prediction and traffic jam detection, and also demonstrated the benefit of using adaptive probing. Future research interests include performance bounds investigation, optimization of adaptive probing, traffic information distribution, and probing commercial vehicle with optimal operation.