Destination choice modeling of discretionary activities in transport microsimulations

Transport of persons and goods brings benefits and costs for individual actors and for the community. A main goal of transport planning is the maximization of net benefit or social welfare through influencing the transport system with due consideration of its environment. Naturally, the definition of net benefit is highly complex and subject to societal discussion. A first step for efficient control and understanding of transport system is transport modeling. While aggregate transport models, such as the 4-step procedure, still dominate the practice and, according to lex parsimoniae, probably will continue to play a significant role, disaggregate models (also called second generation models), with the individual as basic modeling unit, are becoming more and more important. They are able to address infrastructure management issues rather than being focused on infrastructure extension as are the first generation models. A prominent instance of disaggregate modeling are microsimulations that explicitly model the interactions of micro-units, here individuals or vehicles. Due to their conceptual appeal, the large research body, the continuously increasing computational power, and a large availability of microsimulation software packages, microsimulations have the potential to become state-of-practice in efficiently complementing the aggregate approach. A prerequisite for the exploitation of this potential is progress in terms of several crucial issues. On a general level, this concerns the methodically correct and computationally efficient handling of random variability in large-scale scenarios. On a level more specific for spatial choices, such as destination choices, the consistent choice set specification is a crucial problem so far missing a consistent solution. This thesis’ goal is contributing to such progress by providing an operational shopping and leisure destination choice module, implemented for the multi-agent transport simulation MATSim as an example, efficiently applicable for large-scale scenarios, and easily adoptable for other similar simulation models. The main contribution is the formulation of an efficient procedure to generate quenched randomness, i.e., to enable consistent handling of randomness in iterative large-scale frameworks.