On-Line Control Architecture for Enabling Real-Time Traffic System Operations

Abstract

Abstract: Advances in information technology and inexpensive high-end computational power are motivating a new generation of methodological paradigms for the efficient information-based real-time operation of large-scale traffic systems equipped with sensor technologies. Critical to their effectiveness are the control architectures that provide a blueprint for the efficient transmission and processing of large amounts of real-time data, and consistency-checking and fault tolerance mechanisms to ensure seamless automated functioning. However, the lack of low-cost, high-performance, and easy-to-build computing environments are key impediments to the widespread deployment of such architectures in the real-time traffic operations domain. This article proposes an Internet-based on-line control architecture that uses a Beowulf cluster as its computational backbone and provides an automated mechanism for real-time route guidance to drivers. To investigate this concept, the computationally intensive optimization modules are implemented on a low-cost 16-processor Beowulf cluster and a commercially available supercomputer, and the performance of these systems on representative computations is measured. The results highlight the effectiveness of the cluster in generating substantial computational performance scalability, and suggest that its performance is comparable to that of the more expensive supercomputer.