Crowd modelling validation for modified social force model

Abstract

Crowd modeling is mainly used to observe and analyze the movement pattern of the crowd, including their behavior with the influence of building geometry. It also has been used widely in many application areas such as for transportation services, urban planning and event planning. Representation of crowd dynamics using a simulation tool is useful in various crowd studies, where experiments with humans are too dangerous and not practical to be implemented. As to ensure the validity and accuracy of the developed simulation model, it has to be validated with the real data, in which most of recent crowd modeling works are lacking. Therefore, in this paper, we propose three types of approaches to validate our proposed crowd simulation model, the Magnetic Social Force Model, which are the component testing, qualitative validation and quantitative validation. Real data of crowd movement at concourse area of a train station in Kuala Lumpur has been used for the validation purpose in this work. By comparing the simulation analysis with the real data, results for component testing shows that our proposed crowd model has successfully produced crowd trajectories that are similar to the real crowd data with an accuracy of 90%. Meanwhile, for the qualitative validation, the proposed model is able to produce collective types of self-organized crowd behaviours such as lane formation, counter flow formation and corner hugging formation. Furthermore, the model has also been validated using the fundamental diagram.