Crowd Simulation: Improving Pedestrians' Dynamics by the Application of Lattice-Gas Concepts to the Social Force Model

Abstract

The social force (SF) model has been successfully applied to the simulation of flows of pedestrians. Nevertheless, in some scenarios with low density, experiments show that the simulated individuals do not behave as expected, working as irrational particles rather than smart people. For example, by the means of the cited model, it is common to see many simulated individuals going several times straightly against columns, before finding a way to deviate and safely exit the room. Aiming to deal with such problems, this article proposes a way to provide the simulated pedestrians the ability of changing the direction of their displacement at reasonable times, in order to bypass eventually blocked or crowded near areas. To do so, it applies concepts from the lattice-gas model to the SF model. Experiments were driven in order to evaluate the proposed model. As results, it maintained the ability of the SF model to reproduce phenomena like the formation of arcs in evacuated one-door rooms. Focusing on the scenario with a column-blocked one-door room, the simulated pedestrians presented softer and more coherent trajectories, when compared to the pure SF model solution.