Dynamic assignment, surveillance and control for traffic network with uncertainties

This dissertation involves the development of three key components of the ATMIS: dynamic traffic assignment (DTA), surveillance, and management.;Traffic volume (or queue) control scheme is widely used in traffic control practice and has been proven to be effective in managing congestion or gridlock. However, DTA considering the effects of traffic volume control schemes has been missing from literature. To fill this gap, this dissertation considers the analytical traffic volume (queue) control for traffic networks under two route choice behavior assumptions, i.e. dynamic user equilibrium (DUE) and dynamic system optimum (DSO). Existence of equilibrium to the DUE with traffic volume control is proven. The DSO analysis highlights the differences between the dynamic externalities of the two vertical queue models. The results are applied to investigate the traffic induced air pollution pricing.;For the surveillance part, this thesis concentrates on the development of a macroscopic traffic flow model to capture traffic dynamics on networks influenced by demand and supply uncertainties. Based on the modified CTM and the switching mode model (SMM) a stochastic cell transmission model (SCTM) is proposed. The uncertain wavefronts are captured by probabilities of occurrence of operational modes which describe different congestion levels. The SCTM is calibrated and validated by several empirical studies. We compare the performance of the SCTM with Monte Carlo Simulation of the MCTM (MCS-MCTM). The results confirm that the SCTM outperforms the MCS-MCTM. We apply the SCTM to estimate the queues and delays at signalized intersections and compare the results with some well-known delay and queue estimation formulas. The comparison results show a good consistency between the SCTM and these formulas.;In the traffic management part, optimal and robust decision making problems for managing uncertain network traffic are investigated. The traffic management problems are formulated as stochastic dynamic programming problems. A closed form of optimal control law is derived. The robust decision making problem, which aims to act robustly with respect to the supply uncertainty and to attenuate the effect of demand uncertainty, can be recognized as an equivalent optimal decision making problem. The applications of the proposed methods to incident management are also highlighted.