Abstract

A major concern in Advanced Transportation Management Systems (ATMS), one of the principal thrusts of the national program on Intelligent Transportation Systems (ITS), is providing decision support to effectively detect, verify and develop response strategies for incidents that disrupt the flow of traffic. A key element of providing such support is automating the process of detecting operational problems on large area networks. Successful detection of operational problems in their early stages is vital for formulating response strategies such as modifying surface street signal timing plans and activating or updating traveler information systems, including changeable message signs, in-vehicle navigation systems and highway advisory radio, altering emergency services, amongst others. Reliable surface street incident detection is also necessary for the development of integrated freeway-arterial control systems. Incident detection has been the subject of research for the past two decades. But the focus has been on detecting capacity reducing non-recurring congestion on freeways. Only recently has attention begun to focus on developing a methodology for surface street networks. The main focus of this research was to develop a methodology to detect different types of operational problems relevant to the operations of surface street networks. In this research, a modular architecture of neural network has been proposed to develop a comprehensive system to detect different types of operational problems, based on detector data from an urban traffic control system. The modularity of the classifier proposed decomposed the task of detecting different types of problems and produced an overall system of models that individually outperformed a single multi-layer feed-forward neural network model for lane-blocking incidents, special event conditions and detector malfunction, and also a statistically-based discriminant function model. The neural network-based models and the statistical models were developed and tested with simulated and field data from two test study areas in Anaheim and Los Angeles, California, USA. The higher detection rates and lower false alarm rates of the modular neural network model compared to other techniques demonstrated its potential of detecting different types of traffic operational problems on urban arterials.