Understanding Bus Travel Time Variations with AVL Data

Historically, studies of bus networks have focused on a single bus route using time point information at the stop level. In contrast, we approach bus performance with a computational architecture that enables us to compare routes and systems across multiple cities seamlessly, bringing a new perspective to bus system performance. We exploit transit agencies´ publicly available General Transit Feed Specification (GTFS) data and their live XML feeds of bus GPS coordinates. Here we specifically focus on the Massachusetts Bay Transit Authority (MBTA) bus network centered around Boston through May 2011. After performing map-matching of real time coordinates to the shapes of the road network, we can analyze travel time dynamics and, more importantly, detect the time when each bus arrives and departs to and from each stop. We characterize disruptions at the bus level by comparing the statistics of travel times between stops, allowing us to detect the origin of delays in space and time and the emergence of bus bunching. This architecture provides a spatiotemporal analysis of the bus system performance. Driving direction Driving