ACCIDENT PREDICTION MODELS FOR UNSIGNALIZED INTERSECTIONS

Abstract

Intersections are the most hazardous locations among all the road networks. Crashes are the major drawback of these networks as well as major cause of death in both developed and developing countries. Intersections are mainly of two types; signalized and unsignalized. To evaluate safety at such intersections, many studies have mainly focused on signalized intersections, but did not put more emphasis on unsignalized intersections. For safety evaluation, many researchers have made their attempts in developing accident prediction models. In this dissertation, accident prediction model for unsignalized intersections were developed. This was done by analyzing total crashes and incorporating various geometrical parameters, traffic control etc. Accident prediction models (APMs) are invaluable tools that have many applications in road safety analysis. This report discusses the importance of accidents prediction models and its utilization at different unsignalized intersections. From very past, researchers had carried out many studies to predict crashes with respect to various parameters corresponding to traffic volume, geometric characteristics, area type etc. Several statistical techniques have been carried out to develop accident prediction models. Few studies were carried out for unsignalized intersections and most of them targets rural unsignalized intersections. Therefore, there is a need for developing accident prediction models for urban and semiurban unsignalized intersections. As the different traffic movements of unsignalized intersections are controlled by any traffic signal, the frequency as well as severity of crashes are more at unsignalised intersections compared to other roadway system. Therefore, to reduce the crash rate and to improve safety to save human lives, there is a rapid need of safety assessment at unsignalized intersections. As in developing countries, the accident database is not maintained in a systematic manner, APMs can be served as an effective tool to predict collisions from various traffic and geometric related exposures.