Investigation of a Neural Network Model for Freeway Incident Detection

Abstract

A major source of traffic delay in many large urban areas in the United States is non-recurring congestion caused by incidents such as accidents, disabled vehicles, spilled loads, temporary maintenance and construction activities, signal and detector malfunctions, and other special or unusual events, that disrupt the normal flow of traffic. For example, estimates of the proportion of urban freeway delay in the U.S. attributable to non-recurring congestion range up to about 60%. The automated detection of freeway incidents is an important function of a freeway traffic management center, examples of which are increasingly to be found in large urban areas. However, conventional incident detection algorithms have generally met with mixed success in terms of performance criteria such as detection rate, false alarm rate, and the mean time to detect incidents. The need for improved techniques is pressing, particularly with the advent of intelligent vehicle-highway system concepts for integrated freeway and arterial networks that will rely heavily on the ability to automatically detect non-recurring traffic congestion. In this paper we present the initial results of an exploratory study investigating the application of neural network models from the field of artificial intelligence to the automated detection of non-recurring congestion on urban freeways. The results are most encouraging and, we believe, demonstrate the feasibility of the technology and the potential for further research in this area.