Any operational activity-based travel demand model can be divided into two major components:​ activity generation and activity scheduling. Although considerable progress in behavioural theory and modelling techniques has been achieved for the activity scheduling component, activity generation in many ways is an overlooked or under researched area. This thesis concentrates on this critical issue from the theoretical perspective as well as empirical model development.

Based on a comprehensive literature review this thesis develops a conceptual framework for modelling activity generation, considering it as a process. The consideration of the activity generation as a process is to ensure the dynamic interrelationship with the activity scheduling process as well as meaningful integration with other medium- to long-term household decisions (employment, housing, automobile ownership etc.) processes. For modelling the non-skeletal activity-agenda, this thesis adopts the concept of activity utility and proposes a large scale demand system modelling technique. It is demonstrated that a multi-day modelling framework is necessary to ensure both within-day as well as day-to-day dynamics in activity-agenda formation. At the same time it is also necessary to ensure that the activity classification is sufficiently disaggregate so that the trade-offs involved in time allocation to different activities during activity generation process is properly addressed. Long-term development will require operationalization of the proposed activity generation models to improve the policy sensitivity of activity-based travel demand model as well as allow better integration of the travel demand model within an Integrated Land Use Transportation and Environment (ILUTE) framework.

The conceptual framework leads to a step by step procedure for investigating behavioural hypotheses and developing empirical models. Use of week-long activity diary survey (CHASE:​ Computerized Household Activity Scheduling Elicitor) data allows the investigation of behavioural elements within an econometric modelling frameworks of activity generation. The classification of activities into skeletal versus non-skeletal types allows adoption of different modelling techniques consistent with the dynamics and behavioural linkages between short-term travel demands with the medium-to long-term household decisions. Investigation of non-skeletal activities reveals that activity sequencing is more the job of activity scheduling than activity generation.

For modelling the non-skeletal activity-agenda, this thesis adopts the concept of activity utility and proposes a large scale demand system modelling technique. It is demonstrated that a multi-day modelling framework is necessary to ensure both withinday as well as day-to-day dynamics in activity-agenda formation. At the same time it is also necessary to ensure that the activity classification is sufficiently disaggregate so that the trade-offs involved in time allocation to different activities during activity generation process is properly addressed. Long-term development will require operationalization of the proposed activity generation models to improve the policy sensitivity of activitybased travel demand model as well as allow better integration of the travel demand model within an Integrated Land Use Transportation and Environment (ILUTE) framework.