Bus probe based urban travel time prediction

This dissertation seeks to utilize transit buses equipped with a real-time automatic vehicle location (AVL) system as probes to detect non-transit vehicle traffic conditions on signalized urban streets. The explored real-time AVL system sends bus point speed, location, time stamp and other information to the control center about every 30 seconds, which differentiates this dissertation from past bus probe studies that applied bus data from archived AVL systems that cannot transmit bus travel information to a central computer in a real-time fashion.;Based on field real-time AVL bus data and a VISSIM-simulated test bed, this study demonstrates the bus probe concept and further develops a general methodological framework for real-time urban arterial travel time estimation and prediction. The logical framework consists of three steps: quantifying the relationships between bus and car speeds, updating their speed profiles based on newly incoming stream of bus speed reports, and forecasting future car travel times. The first step is most critical as it is the basis of the entire study. Exploratory analyses of the real-time AVL data and the test vehicle data are presented first, followed by the statistical models including multiple linear regressions, seemly unrelated regressions and multivariate state space models. The results have confirmed that there exist significant relationships between bus travel and general vehicle travel and reasonable non-transit vehicle travel times can be estimated. The second and third steps involve a Bayesian updating and a multivariate time series forecasting algorithm respectively. The results have shown it is possible to use new incoming bus speeds and forecasted bus speeds to derive reasonable concurrent non-transit vehicle travel times via the relationships identified in the first step.;The current study focuses on the feasibility, framework and algorithms in utilizing buses as probes. Further studies that focus on on-line data processing and broadcasting are needed before the idea of using bus probes becomes fully operationable.