Modeling multiple route choice paradigms under different types and levels of ATIS using correlated data

Advanced Traveler Information Systems (ATIS) provide qualitative and quantitative real-time traffic information before the trip and while driving offering potential solutions to enhance the transportation system performance and improve the quality of travel in urban areas. The success of these smart systems relies heavily on the understanding of drivers' behavior under ATIS which has not been fully investigated, and therefore confines the implementation of ATIS in real life. This dissertation presents the design and development of an interactive windows-based travel simulator to collect dynamic pre-trip and en-route route choice data under ATIS. It simulates commute home-to-work morning trips and uses a realistic 25-node and 40-link urban network from Orlando. The network has a fixed origin-destination pair and comprises different types of highways. The simulator provides five different levels of traffic information to the subjects in five different scenarios. It presents ten simulated days (two days for each scenario). The simulator accounts for delays caused by intersections, recurring congestion, non-recurring congestion (incident), queuing at toll plazas, and weather condition effects. In the simulator, ATIS provides information about travel time, weather conditions, incidents, and in some scenarios advice and information about the shortest path.; The dissertation presents models of the following; (1) drivers' diversion from normal routes under ATIS, (2) drivers' route choice under long-term pre-trip traffic information, (3) drivers' route choice under short-term en-route traffic information, (4) the effect of ATIS on travel time and travel time variability, and (5) repeated multinomial route choices under ATIS. All models were suffering from correlation between repeated choices made by the same subject. Model 5 was furthermore suffering from correlation between overlapping alternatives. This dissertation introduces Binary and Multinomial Generalized Extreme Equations (BGEE and MGEE) techniques as new methods to account for correlation in binary and multinomial route choice modeling, respectively. Also Mixed Linear Models (MLM) is introduced as a new method to account for correlation in linear models in transportation research.; The modeling results showed that gender is the only socioeconomic factor that does not affect any of the above route choice paradigms. Familiarity with the device that provides the information has a significant effect in the first four models. Expressway users are shown as the most travel-time savers who would divert if they are guided to a less-travel-time alternative. Number of traffic signals on the normal route and advised route affect diversion from the normal route and compliance with pre-trip advised routes. The results show also that drivers can identify and follow the shortest path when they are provided with advice-free traffic information. The odds of choosing a certain shortest-path route, advised or not, vary from route to another depending on its characteristics. As the level of information increases (adding en-route to the pre-trip and advice to the advice-free information) the average travel time decreases. In addition, it was shown that ATIS has significantly reduced the variability of the travel time as well. The effect of correlation was found statistically significant in most of the cases which shows the importance of accounting for correlation in route choice models that may lead to significantly different travel forecasts and policy decisions.