IMPROVING URBAN BUS SERVICE REGULARITY MODELS USING REAL-TIME DATA AND A ROUTE-LENGTH COEFFICIENT

Abstract

This paper substantiates the limitations of the commonly used trip-count-based regularity indicator in assessing urban bus service regularity, since it does not capture real-time operational factors such as trip travel time, headway variability, and passenger waiting time. The study aims to improve a service regularity model integrated with a platform that enables estimation of trip travel time and passenger waiting time under real operating conditions based on AVL/GPS observations. In the proposed approach, the actual trip time for each run is decomposed into motion time over route segments, dwell time at stops, delay time at signalized intersections and congestion points, and terminal buffer time; a travel-time coefficient is then defined using deviations from planned values. In addition, a headway-based coefficient reflecting passenger waiting time is introduced, and an overall trip quality score is formed as a weighted combination of time-related factors. The model also incorporates a separate coefficient that represents the natural degradation of regularity as route length increases, and the final regularity index is computed with the route-length effect explicitly accounted for. The improved model is evaluated through a computational experiment on two routes: Route 1 (17.7 km; 10 stops; 44 min one-way; 8 buses) and Route 13 (18.9 km; 13 stops; 36 min one-way; 13 buses). The results indicate that incorporating real-time platform data and the route-length factor increases the sensitivity and interpretability of the regularity assessment and expands diagnostic capabilities for dispatching actions such as buffer setting and holding control.