Simulation of freeway incident detection using Bayesian networks

This paper reports the development of a new freeway incident detection algorithm using a microscopic traffic simulation tool. The proposed TSC algorithm has a modular architecture. A data processing module is designed to handle specialties of different sites. A Bayesian network based incident detection module is developed to manage general expert traffic knowledge and perform coherent reasoning. The algorithm is evaluated using both simulated and field incident data. A promising performance is obtained. In addition, the algorithm performs consistently when it is transferred to new freeway environments from its original study site. (a) For the covering entry of this conference, please see ITRD abstract no. E212640.