Pedestrian Traffic - Simulation and Experiments

In recent years and decades the development of ever more powerful computer hardware has been accompanied by the evolution of simulational or computer physics as a third element of physics next to theory and experiment. This thesis deals with the simulation of pedestrian traffic with a focus on evacuation processes. While theory and experiment, respectively empiricism, relied on each other since the dawn of modern physics, they do not necessarily rely on simulations, although they have begun to make heavy use of it. Simulations on the contrary can never be carried out meaningfully without theories and experiments backing them and making use of them in the interpretation process of the results. This dependence is reflected in this thesis, which includes elements of all three operation methods. It begins with an overview of elements that are necessary to build reliable simulation models. The interrelation between simulation, theory and experiment is set out in more detail there. Then a survey of existing models of pedestrian evacuation dynamics is given. The second chapter deals with the semantic - and therefore rather theoretical - problem of how a cellular automaton model can evolve toward a model which is better referred to as “discrete” model when the model is extended. This question is irrelevant for the issue of reliability, yet it is often asked. In the third chapter a discrete model of pedestrian evacuation dynamics is constructed and tested. The tests of the various elements of the model focus on the elements’ influence on the fundamental diagram, yet there are also some other tests which include some background from theory. The main results of this chapter are the construction of the model itself, the proof that it is very well able to reproduce a widely accepted empirical fundamental diagram up to a density of roughly four persons per square meter, and that - concerning computing times - the model is applicable to scenarios with a few million persons. The fourth chapter deals with the analysis of two observations and two experiments. The first observation was done during an evacuation exercise in a primary school. The empirical data was partly used to calibrate the parameters of the simulation and partly to compare them with the results of simulations which were done using these parameters. The second observation is a study of upstairs walking speed distributions on a long stair. In the counterflow experiment a rich variety of self-organisational structures showed up, which will be a challenge to model in the future. The finding that the sum of flux and counterflux is always larger than the flux in no counterflow situations may be seen as the most interesting result of this experiment. The main results of the “bottleneck experiment” is that the flux is neither a linear nor especially a step function of the width of a bottleneck and that therefore some legal regulations are based upon wrong assumptions. Chapter five consists of five examples with diverse focuses for the application of crowd simulations. The appendix includes a record of crowd disasters as well as - following from that - some considerations on human behavior in dangerous situations.