Addressing Data Latency in GTFS (General Transit Feed Specification) Realtime to Improve Transit Signal Priority

• Published in: Transportation research record 2024-07
• Main Authors: Zhou, Joseph Q., Jackson, Jay, Stravitz, Pam, Barlow, Gregory J., Koonce, Peter
• Format: Article
• Language: English
• ISSN: 0361-1981; 2169-4052
• DOI: 10.1177/03611981241255027

Transit signal priority (TSP) is a strategy that provides preferential treatment at signalized intersections. TSP reallocates green time to reduce the delay of transit vehicles at traffic signals. To be effective, a transit vehicle (bus) must communicate its location to the traffic signal to make the reallocation of time beneficial. In GTFS (General Transit Feed Specification) Realtime, latency poses a significant challenge for the implementation of GTFS-based TSP. Using data from four transit agencies, this research identifies issues with current GTFS Realtime feeds and proposes a solution using machine learning algorithms to address latency compensation. Experimental results demonstrate that the performance of two machine learning models surpasses the baseline approach, which relies on hourly means for bus speeds and dwell times. This paper tackles multiple issues related to existing GTFS data, enhancing the practicality and feasibility of GTFS-based adaptive TSP. In contrast to conventional approaches focusing on estimation of bus arrival time, this paper emphasizes estimation of bus location and presents an effective method to compensate for latency and improve estimation of bus location and dwell time.