An incident detection algorithm using artificial neural networks and traffic information

Abstract

Incident detection methods for the automatic recognition of incidents and other freeway events requiring emergency responses have existed for over forty years. Most of the developed and implemented algorithms rely on inductive loop data. Inductive loops are the most commonly used traffic sensor and collect data such as volume and velocity at a point. However, the implemented algorithms using inductive loop data work with mixed success. Recently, there has been renewed interest in incident detection algorithms partly because of new sensors for obtaining traffic information. One of these new sensors is a Two-way Probe Car System (TPCS), which was developed as a mobile detector for measuring link travel speeds in South Korea. TPCS is mainly a means of collecting enhanced roadway condition information and then broadcasting related traveler information and various alerts back to vehicles. In this paper, we suggests a new model for incident detection using TPCS data and neural networks.