Development of an expert system for identifying effective countermeasures at rural unsignalized intersections

Frey, Rebecca
Journal Title
Journal ISSN
Volume Title
University of Delaware
Intersections are among the most dangerous points in a transportation network. Both shortcomings in design and human error contribute to the frequency of crashes at intersections. In this thesis, a knowledge based expert system is created to assist engineers and transportation officials in the process of improving the safety at rural, unsignalized intersections. Currently, the process of identifying countermeasures for rural, unsignalized intersections with a chronic safety problem involves a great deal of experience and subjective assessments. The knowledge based expert system developed attempts to provide a more logical progression of the intersection safety process and incorporate quantitative assessments of potential options. The American Association of State Highway and Transportation Officials’ (AASHTO) 2010 publication, the Highway Safety Manual seeks to provide the first quantitative assessment of intersection safety countermeasures. These values are incorporated into the knowledge based expert system (KBES). The KBES for countermeasure identification is the result of a comprehensive literature search to identify and determine the effectiveness of countermeasures of rural, unsignalized intersections. The countermeasures considered in this project include intersection lane narrowing, advanced warning rumble strips, flashing beacons, advanced warning pavement markings, signage improvements, conversion to roundabout, signalization, conversion to all-way STOP control, left and right turn lanes, shoulder bypass lanes, sight distance improvements, intersection realignment, intersection illumination, and performing a skid test for pavement quality. The KBES was created using the software CORVID produced by Exsys® Incorporated. The system is constructed of variables joined together using IF…THEN statements in the logic block and is designed to represent the thinking process of a human expert. The KBES was tested using data from three high crash rural, unsignalized intersections in Delaware which were examined during a previous 2012 Delaware Department of Transportation (DelDOT) study. The results produced by the expert system were comparable to the results achieved by the DelDOT study. Testing the system in this manner begins the process defined by the Federal Highway Association (FHWA) for verification and validation of expert systems.