USING REAL-TIME TRACKING DATA TO BETTER UNDERSTAND PUBLIC TRANSIT

Public transportation is an essential part of urban life. In recent years, transit agencies have begun live GPS tracking of their vehicles and many agencies make this data available to the public in real time. This data can be used to study the motion and interactions of trains and buses in ways that were previously impossible. There is an extensive literature on the dynamics of buses and trains, but almost all of it is theoretical or based limited data (often only a few hours worth) due to the expense and difficulty of obtaining detailed data. We study the behavior of a light-rail rapid transit line in a major American city. This line runs at grade, making frequent stops to load passengers and wait for crossing road traffic. We used the API provided by the transit agency to request the train locations every 10 seconds. Using millions of train positions recorded over several months, we compare the motions of these vehicles to those of particles in an asymmetric exclusion process and a unidirectional random walk. Trains are released from one end of the line at reliable intervals. As they move along the route, their motion is affected by several highly-variable factors including wait times at red lights, and the time required to load/unload passengers. The number of passengers waiting at a station not only varies according to the time of day, but also depends strongly on the amount of time since the last train has passed. This causes an effective attraction between trains. Near the beginning of the line, the distribution of headways is sharply peaked around the average frequency. This distribution broadens along the route and we observe a distinct bunching of trains. We also examined travel times on this line and found that the average speed of the trains was significantly faster than walking, but with a large variance. Commuters are sensitive to the distribution of travel time, not just the average. The large variance in travel time causes inefficient commutes: people leave far earlier than their average commute time in order to guarantee that they arrive on time most of the time (not just on average). Thus, transit agencies could greatly improve service by reducing the variations in travel time, even if this results in no reduction in the average time required to complete a trip.