CARSystem: a distributed algorithm for efficient real-time vehicular traffic control

Abstract

Traffic operations using centralized adaptive systems require heavy infrastructure and processing power. Distributed strategies can provide scaleable and more fault-tolerant systems. Distributed traffic control systems, that lack effective information transfer between control points (traffic controllers at intersections), do not have the capability of operating in a coordinated fashion. This research attempts to develop a fully-distributed traffic control algorithm that is based on message transfer between individual traffic control processes. Real-time information -obtained by tracking the density profiles that constitute demand at each intersection in real-time-is used by an on-line optimization algorithm to provide real-time adaptive control. This thesis documents the development of a real-time distributed vehicular traffic control algorithm called CARSystem. The proposed algorithm is described and the performance is compared to conventional systems using TexSIM Network Traffic Simulation Model. Possible extensions of the algorithm to handle congestion, preemption, and other advanced strategies are also discussed.