Efficient algorithms for the microsimulation of travel behavior in very large scenarios

Our life is strongly influenced by travel which enables the mode of living we are used to in the first place. This importance of travel and the resulting desire for reliable and effective transportation drive the need for predictive models concerning our use of transport infrastructure. It is remarkable that usually travel is not an end in itself; that is, the motion of traveling is most often not the reason for it to be performed. Nearly always it merely serves the purpose to either transporting a person or goods from one place to another, to enable new activities not available otherwise. This thesis follows the conviction that travel behavior can only be modeled and conceived on the level of the individual as it is essential to understand the reasoning behind travel. Therefore, the approach under investigation herein is agent based traffic modeling. Agents, as computational representations of real life individuals, act according to simplified rules, by adhering to which they find useful daily activity plans. Then, they execute these plans in a traffic flow microsimulation, producing emerging phenomena such as congestion. The resulting flow patterns and the corresponding agents’ daily activity plans can later be used for traffic and travel behavior analysis. Representing regional travel behavior on the level of individuals creates major challenges in terms of computational costs of such methods. The aim of this thesis is to present algorithms and implementations able to reduce this high demand of resources, rendering agent based models practicable for engineering consultants on standard computing hardware available in most offices today.