Tour-based Mode Choice Model in Activity-based Modelling Framework

This thesis presents a tour-based mode choice modelling structure for activity-based travel demand models by exploiting the classical dynamic discrete choice modelling approach. Many activity-based modelling systems rely on either trip-based or ‘simplified’ tour-based mode choice models that in many cases completely overlook the dynamics of mode choice behaviour. To contribute to filling this gap, this thesis applies a systematic approach of investigation to better understand the nature of tour-based mode choices and to propose a parsimonious modelling structure for it. The first investigation looks into the trip-based mode choice behaviour of post-secondary students commuting to universities in the City of Toronto. The second investigation uses a heteroskedastic dynamic discrete choice model for tour-based mode choices modelling with an empirical investigation of university students’ daily mode choices in Toronto. The third investigation uses a computationally tractable dynamic discrete choice modelling framework for modelling tour-based mode choices. The fourth investigation uses a random utility maximization -based dynamic discrete-continuous modelling approach to capture individuals’ tour-based modes and continuous time-expenditure choice trade-offs in a 24-hour time frame. The model results reveal that individuals’ sensitivity to travel costs varies, while their sensitivity to travel time remains stable. The empirical model reveals that users of newly introduced mobility services (e.g., Uber, Lyft) tend to have different mode choice patterns and value of travel time savings than non-users of these services. The forward-looking component reveals that availability of the modes for subsequent trips in the tour represent a significant portion of the utility of the current mode choices. In terms of the time-expenditure choice model, it is found that full-time employees and younger individuals tend to leave home earlier than part-time employees and older individuals. It is found that individuals are likely to spend long hours at work or school if they leave home early. Furthermore, individuals are likely to schedule non-mandatory activities, such as shopping, later in the day. The validation and policy evaluation results are promising. While the models proposed here can be easily developed in different regions across North America, opportunities also exist for the application of this type of analysis globally.Ph.D