Crowd modelling

Mikkel Hartmann, Wence Xiao, Troels Christensen, Dan Elmkvist Albrechtsen, Malik Thrane & Toke Høiland-Jørgensen

Studenteropgave: Semesterprojekt


We look at social force models as a way to model the behaviour of human crowds, in order to eval- uate how well these types of models simulate crowd behaviour, and what the models’ strengths and weaknesses are. In order to do this evaluation, we implement a computer simulation of an exemplary social force model. In order to create this simulation, we pick an exemplary model that is well described in the article that presents it, and analyse it in detail, filling in details from other articles when necessary. Based on this analysis of the model, we go from the abstract model formulation to a concrete numerical simulation by filling in required details, such as how to approximate the movement of pedestrians, how to set initial conditions and values, and how to implement the interaction between pedestrians and walls in practice. From our results, it is clear that our simulation (with the right parameters) exhibits reasonable pedestrian behaviour upon visual inspection. While we successfully replicate some results from the literature, other effects do not manifest themselves. We discuss several reasons for this dis- crepancy, including features that are missing from the model, parameter values, effects of using random numbers to generate the initial conditions and possible errors in our implementation of the model. Based on the results of our own simulations and our review of the social force modelling field, we assess social force models and their strengths and weaknesses. We conclude that social force models are not based on any theories for the behaviour of crowds, but are created to replicate a set of observations. As such, any confidence in their predictions must come from a record of pro- ducing results fitting observations; and since the field is relatively new, they have not yet reached this state. Social force models do, however, provide a practical way to simulate something that would otherwise be impossible to simulate. As such, they are the best available way to provide e.g. guidance when designing facilities that must accommodate many pedestrians, and given time the accuracy of their predictions will probably increase.

UddannelserMatematik, (Bachelor/kandidatuddannelse) Bachelor el. kandidat
Udgivelsesdato4 jan. 2011
VejledereViggo Andreasen


  • evacuation
  • Crowd Modelling
  • mathematical modelling
  • social force
  • computer simulation
  • pedestrian
  • Agent-based model
  • panic