Microscopic Bus Performance Analysis Using Real-time Data in Greater Sydney

Abstract

Traditional methods of measuring bus performance rely on manual surveys of arrival and departure times, which are costly and inefficient. With GPS-equipped buses, real-time updates on vehicle locations and stop arrivals can now be collected more efficiently. The General Transit Feed Specification (GTFS) provides a structured format for managing this data. Using GTFS-Realtime feeds from public transport authorities, bus performance can be evaluated at high resolutions.

Despite its potential, the adoption of GTFS-Realtime has been limited by challenges such as non-human-readable formats and complex data cleaning requirements. This thesis develops a data pipeline to overcome these barriers, transforming 25 months of GTFS-Realtime Trip Updates data in Sydney, Australia.

The first study applies panel regression models to assess how traffic signals, priority measures, and cross-traffic turns (left turns in right-hand drive countries) affect marginal delays across stop-to-stop segments. Variables include traffic volumes, precipitation, and COVID-19 restrictions. Findings show that bus-taxi and bus-HOV lanes reduce delays and improve reliability, while traffic signals and cross-traffic turns significantly increase delays.

The second study focuses on bus cross-traffic turns using both Trip Update and Vehicle Position data. Statistical and trajectory analyses show that cross-traffic turns increase delay and variability while reducing operational speed. Cross-validation supports the accuracy of this microscopic approach.

To address these issues, the final study introduces the bus cross-traffic turn priority box—a bus queue jump lane that enhances bus speed and reliability without reducing road capacity or green time for other vehicles. This thesis demonstrates the value of GTFS-Realtime data for microscopic bus performance analysis and offers actionable strategies to improve urban bus operations.