Social Crowd Simulation: The Challenge of Fragmentation

Michelangelo Diamanti, Hannes Högni Vilhjálmsson

Abstract: Crowd simulation is the process of simulating the movement and behavior of a large number of people. This field is continuously being improved by incorporating different theories of how humans move and interact with their surroundings, steadily increasing the realism of the simulation. Furthermore, new techniques for calibrating simulation parameters, and evaluating the accuracy of simulation output, keep being proposed. This paper presents a brief overview of these foundations and argues that a fragmentation of the field into multiple incompatible solutions may impede progress towards comprehensive social behavior models. It finally argues that abstractions of human intent and behavior, proposed within the Embodied Conversational Agents community, may suggest a useful path towards bringing social crowds to new levels of realism.