Uncertainty consideration in the knowledge based approach to road safety investigation
Abstract
Road safety has become the critical issue in transportation around the world. Safety specialists are continuously striving for improvement in this area. The most difficult part; however, is determining the road features, driver behaviors, and vehicular failures that unduly increase the risk of crash. One of the frequently used methods of safety improvement is investigating roads with an excessive number of crashes. However, the lack of standard techniques causes investigative teams to rely on their own experience and judgment. Moreover, the road safety investigation process includes various assumptions that are needed to deal with the lack of data and the inconsistency among expert opinions. The simple approach is to neglect uncertainty, which, unfortunately, can lead to more issues and to a distortion of the findings. This dissertation establishes and evaluates a technique to incorporate uncertainty considerations into the road safety investigation process. First, a knowledge base is developed and organized into a Bayesian Belief Network structure. The knowledge is acquired from multiple sources including the National Cooperative Highway Research Program (NCHRP) reports, safety guidelines, observation of experts during work, and road safety investigation reports. Next, the uncertainty is incorporated into the developed network, and Bayesian inference is used to update the network probabilities. The method is evaluated by comparing the results obtained by the participating experts and non-experts using two methods; namely, with and without uncertainty consideration. The evaluation results show that taking into account uncertainty can significantly improve the findings and prevent overlooking important factors. Finally, a discussion of the research contribution and possible future improvements are presented.
Degree
Ph.D.
Advisors
Tarko, Purdue University.
Subject Area
Civil engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.