Analysis of Trends in Bus Dwell Time Data for Real-Time Predictive Modeling

Abstract

Transit vehicles create special challenges for urban traffic signal control. Signal timing plans are typically designed for the flow of passenger vehicles, but transit vehicles, with frequent stops and uncertain dwell times, may have very different flow patterns that fail to match signal coordination plans. The presence of transit vehicles stopping on urban streets can also restrict or block other traffic on the road, resulting in further disruption to coordination. These factors can result in increased overall wait times and delays throughout the system for transit vehicles and other traffic. Transit signal priority (TSP) systems are often used to mitigate some of these issues, primarily addressing delay to the transit vehicles. However, predominant existing TSP strategies give unconditional priority to transit vehicles, thereby exacerbating quality of service for other modes.

In areas where transit vehicles have significant effects on traffic congestion, particularly in urban areas, using more realistic models of bus behavior in traffic signal control strategies could reduce delay for all travel modes, particularly in a connected vehicle context using adaptive control. However, estimating the arrival time of a transit vehicle at an intersection requires an accurate model of transit stop dwell times. As a first step toward developing a dwell time model for purposes of predicting bus arrival times, this paper analyzes trends in automatic vehicle location (AVL) data provided by the Port Authority of Allegheny County (PAAC) collected over the two year period from September 2012\,--\,August 2014 for two major bus routes. Our analysis enables several inferences to be drawn. First, the statistical properties of dwell times are similar (for most stops) across years for a given season and hence it is fine to join the data for the same season (or month) across years. Second, the probability of a non-zero dwell time varies from stop to stop in a given route suggesting that buses need not be given same priority at all signalized intersections. Third, cumulative density functions (CDFs) of dwell time distributions do provide insights into reliability of dwell times for a given stop; this information is especially useful in real-time control decisions; Fourth, fifteen minute interval dwell time CDFs of peak hour demonstrate the highly stochastic nature of dwell times. Based on this trend analysis, we argue that an effective predictive dwell time distribution model must treat independent variables as random or stochastic regressors.