Development of a cognitive and decision-based model for pedestrian dynamics

Research on pedestrian dynamics is always an interplay between empirical and experimental observations and theoretical modelling and simulations. Thereby, pedestrian models are not only used for theoretically reproducing empirical data, but also to better analyse and understand the mechanisms and behavioural aspects that underlie pedestrian dynamics. The model approach that is presented in this work assumes pedestrian motion to result from cognitive and decision-based processes. The model is set in continuous space, but discrete time and therefore belongs to a model class whose potential has been rarely investigated yet. However, compared to other model classes that are widely used in pedestrian dynamics, this approach is highly advantageous considering fidelity and simplicity in its structure. A pedestrian is considered as an autonomous entity that gains information on the surrounding by visual perception and anticipation. On this basis, the agent takes a decision on its movement for the next time step. The main focus during the development of the approach was on modelling the interaction and collision avoidance with other agents. Particularly for the collision avoidance, stochastic procedures are used in order to consider uncertainties of human decisions explicitly which makes the modelling approach more realistic. As simulation results show, the new approach is able to reproduce characteristic effects of pedestrian motion very well. For typical scenarios that have been used as test cases the simulated results fit well, at least qualitatively but often even quantitatively, to experimental data. Especially, important macroscopic effects, particularly collective phenomena, are observed in the results that are reproduced by modelling individual interaction of a single pedestrian with others. During the development of the model its parameters were specifically adjusted for the single scenarios, considering the empirical data basis. In addition, several cognitive mechanisms were supplemented. By this means, it is possible to identify and understand the important intrinsic properties and motivations of a pedestrian. Furthermore, this provides the opportunity to gain insight into how cognitive and decision-based approaches can model pedestrian behaviour as realistically as possible