Investigating morning commute route choice behavior using global positioning systems and multi-day travel data

One of the major impediments to developing a larger body of knowledge in travel behavior is the lack of detailed travel data. The lack of sufficient data is the result of the inherent complexity of gathering and subsequently analyzing observations of the phenomena of interest. This is particularly true for route choice, a topic on which scant link-by-link data appear to be available, especially for multiple days. In fact, very little empirical work is based on real world observation. This dissertation studies the factors that influence morning commuters' route choice and route switching based on objective real-world observations of travel behavior during a multi-day period.; This dissertation analyzes the spatial deviation patterns for morning commutes made by the same drivers captured by GPS units. Such patterns were impossible to discern with earlier conventional survey methods. The results indicate commute distance has an impact on different types of spatial route deviations. This dissertation tests the current route choice model assumption that assumes travel time or travel distance is the only factor influencing drivers' route choice decisions. Investigation of the objective route choice factors confirms that minimizing travel time, although very important, is not the only factor impacting route choice. Several other factors have been identified to impact commuters' route choice. This dissertation examines the choice between using single or multiple morning commute routes. The results indicate the strong explanatory power of work schedule flexibility and trip-chaining on the dependent variable comparing to the commuters' socio-demographic characteristics and commute route related attributes. This dissertation also presents an extensive effort in analyzing GPS-based travel behavior data and develops a methodology to subtract route choice information and trip-level travel information from the GPS-based vehicle activity data.