Dynamic traffic control: Decentralized and coordinated methods.

Abstract

An overall dynamic traffic management system that seeks to maximize network-wide performance is the primary focus of this research. Specifically, this dissertation deals with the development of efficient techniques for the dynamic control of signalization in traffic networks in the context of Intelligent Transportation Systems. It comprises three complementary components: decentralized control, coordinated control, and coordinated control in an Advanced Traveler Information System (ATIS) environment. For the first component, an algorithm to optimize, in real-time, traffic signals for individual intersections in traffic networks is presented; it uses an efficient decision-tree searching technique to minimize delay. This decentralized algorithm for traffic controllers is called Adaptive Limited Lookahead Optimization of Network Signals (ALLONS-D). Two perspectives are addressed as part of the second component: (i) a hierarchical control architecture for enabling local controllers to maximize system performance and (ii) an iterative process (ALLONS-I) to determine an equilibrium set of control policies for traffic-responsive signal controllers like ALLONS-D. The first perspective divides local signal choice and coordination of these local controllers into two layers of control. An optimization problem is formulated to determine the coordination requirements that are imparted to the local controllers from the higher layer. The ability of this scheme to improve performance on arterial and grid networks is tested via software simulation. In the same manner, this hierarchical scheme is shown to be useful in improving the flow of transit vehicles in a traffic signal network relying on ALLONS-D controllers. The second perspective for the second component of this dissertation deals with a form of coordination achieved by iteratively recalculating the signal control policies at the intersections; this iterative method is a dynamic adjustment process. This process is successful in converging to a set of coordinated traffic signals in some cases; its convergence properties are analyzed using a game-theoretic model. The result of this analysis is a proof of convergence for a specific class of traffic networks and traffic demand. Finally, the third component of this dissertation develops a traffic optimization process that incorporates drivers' route selections as well as the resulting adaptive traffic signal control policies. This iterative signal optimization - traffic assignment technique is developed and shown to converge to a dynamic user-equilibrium solution through simulation experiments and formal analysis. This iterative method allows an examination of the long term effect of the adaptive signal control scheme on drivers' route choices.