The advent of the Advanced Public Transportation Systems program by the Federal Transit Administration has encouraged bus transit operators to implement automatic vehicle location (AVL) systems for real-time monitoring. While the primary focus has been on the implementation of technologies, such as AVL systems, it is necessary and, perhaps, important to develop advanced performance analysis and evaluation procedures that can assist in the schedule planning and real-time service control tasks taking into advantage the real-time monitoring data. One potentially useful and effective approach for assisting in service control tasks, is the schedule behavior modeling concept. In this research effort, this concept is introduced to model the schedule behavior of buses on a route using schedule deviation information. The schedule behavior modeling approach presented in this study represents an innovative concept for modeling the performance of bus transit operations.; This research focussed on investigating the application of artificial neural networks (ANN) and the Box-Jenkins technique for developing and testing schedule behavior models using data obtained for a test route from Tidewater Regional Transit's AVL system. The three ANN architectures investigated were: Feedforward Network, Elman Network and Jordan Network. In addition, five different model structures were investigated. The time-series methodology was adopted for developing the schedule behavior models. The modeling experiments provided no conclusive results. However, Jordan network model provided encouraging results and performed well. Finally, the role of a schedule behavior model within the framework of an intelligent transit management system is defined and the potential utility of the schedule behavior model is discussed using an example application.
展开▼
