The Analysis of Results of Stochastic Evacuation Models

Sammanfattning: All humans are unique, we have different characteristics and we make different decision. This is a challenge when it comes to modelling human behaviour. The models we design are based on mathematics where there is no room for inconsistencies such that are present in human behaviour. So how do we account for this when we try to model human behaviour in evacuation modelling? Most models provide the possibility to include mathematical distributions and algorithms that are supposed to represent this uncertainty that we refer to as behavioural uncertainty. This thesis aims at providing a method based on functional analysis and statistics that can be used to study the effects of behavioural uncertainty on evacuation simulation results and in particular to study convergence in results for important output parameters. Previous work has been done for the most important evacuation model output, namely the evacuation time. The work in this thesis contributes by including other output parameters which are of importance both for the evacuation time and other aspects of evacuation safety. The method is accompanied by a tool that helps the user in this kind of quantitative assessment. A demonstration of the method and the tool has been provided through the use of a case study. The results showed that the analysis is efficient in analysing convergence in results for a variety of output parameters, even for output parameters with varying number of data points. This enables a more comprehensive and detailed analysis than what has previously been shown, ensuring that the behaviours that govern evacuation time also have converged in results.