Simulating police containment of a protest crowd

This paper presents an agent-based computational model of the crowd containment tactic known as kettling, which involves cordons of police officers surrounding a crowd of protesters in order to restrict its movement. Our model uses steering behavior techniques to simulate a series of simple scenarios in which an unorganized group of protester agents clash with an opposing group of police agents holding position in a stationary, multiple-rank line formation. We investigate the stability and penetrability of the police formation and how they are influenced by the formation's thickness (number of layers) as well as the specific behavioral strategy employed by individual police agentswhether using global or local references, or combinations of both. Our results show that a strategy using purely global references produces optimal performance, while increasing the number of formation layers enables strategies using combined references to approach this optimum to different degrees.