Quantitative measures of crowd patterns in agent-based models of street protests

Abstract

In this work we describe the introduction of quantitative measures of emergent crowd patterns in an existing Agent-Based model (ABM) of street protests with multiple actors (police, protester and ‘media’ agents). The model was applied to a scenario of a police force defending a government building which protesters seek to invade. The improved model provided a coherent ‘narrative’ of the simulations and highlighted the realistic and unrealistic aspects of the agents’ interactions. Two new types of police agents – ‘defensive’ and ‘offensive’ – were introduced, leading to a realistic model representation of police cordons defending a site or charging to disperse clusters of violent protesters. The new quantitative measures provided information on cluster size and orientation of clusters of violent protesters, as well as police coverage and protester breaching of the defensive perimeter, together with the time history of the bursts of localized fights and arrests. It was shown how the quantitative measures of the emergent properties can be used for both parameterization and validation of the model.