Simulating service reliability of a high frequency bus route using automatically collected data

Abstract

High frequency bus routes are subject to a variety of influences that can affect the quality of service provided to passengers. Since they have short headways and high passenger demand interaction between buses can easily develop, causing degradation in service reliability. This, in turn, can prompt service interventions to correct service reliability. Transit agencies are implementing new technology that provide rich data sets for analysis and are also experimenting with a variety of operating policies to improve service reliability. This research develops a simulation model of high frequency bus service in order to study the causes of service unreliability and strategies to alleviate it. The model is designed to be used in conjunction with data recorded by the Automatic Voice Annunciation System (AVAS), Automatic Passenger Counting (APC), and Automatic Fare Collection (AFC) systems and is calibrated to represent route 63, a key bus route in the Chicago Transit Authority (CTA) network The simulation model is first used to conduct a sensitivity analysis of the factors influencing reliability, such as passenger demand, terminal departure behavior, and unfilled trips. Next, several operating strategies, including terminal departure and timepoint holding for schedule or headway, are modeled and evaluated for their potential to improve reliability. The sensitivity analysis and application testing support the use of passenger-centric metrics such as passenger-experienced waiting time and crowding over more aggregate headway measures such as large headways and bunching. Model results show that headway management strategies implemented at the terminal can significantly improve bus service reliability and ameliorate the impacts of unfilled trips on route 63, as measured by passenger waiting time, crowding, and big-gaps / bunches.

(cont.) The simulation model is a valuable research tool for applications beyond those tested in this thesis. The model developed can be applied with from data collected by automatic collection systems which is a particularly useful feature for transit agencies.