Travel behaviour modelling for scenarios with exceptional events

Over the past years, micro-simulations have become increasingly important as supporting tool for the decision making process in the fields of traffic analysis and traffic management. In contrast to traditonally used aggregated models, they are able to produce a larger range of outputs including detailed information on each synthetic traveller. Moreover, the decision making process of each person can be modelled individually. Common tools that use micro-simulations anticipate that the simulated scenario represents a typical day where no unexpected events occur. Furthermore, they assume that each traveller act as a homo oeconomicus. Related to this is the assumption that all travellers have complete information, for example about the traffic flows or the available capacity in parking spaces. If a scenario contains events that cannot or only partially be forseen, this assumption is not valid anymore. Simulating such scenarios with a common simulation based tool could lead to illogical behaviour of the simulated population or even false results. The interest in such scenarios has grown in recent years. Therefore, this dissertation describes in detail how the simulation framework MATSim was extended to support scenarios with exceptional events. The extent of such events can range from small scale to large scale. While the first ones have limited spacial impacts, the latter ones have major impacts on the transport systems and the behaviour of the affected population. A car accident would be a small scale incident while earthquakes or terrorist attacks would be examples for large scale incidents. To support such scenarios, several extensions have been integrated into the framework. In a first step, the ability to simulate pedestrians, cyclists and car passengers is added to the framework. So far, those transport modes were not simulated in detail and the travel times were based on assumptions. As a result, the positions of agents travelling using one of those modes were not known. The implemented multi-modal simulation extension produces more accurate travel times and allows tracking agents while they are travelling.