Self-organized behaviour in a modified multi-agent simulation model based on physical force approach

Abstract

The multi-agent simulation models are very useful in predicting crowd behaviours, where experiments with human are too dangerous. However, works that include intelligence in the physical force-based model are very limited. Moreover, most works based on the physical force-based model are only restricted to producing behaviours of crowds in emergency situations. Utilizing a simpler mathematical representation, this paper proposes a modification to an existing crowd simulation model based on the physical force approach. The proposed method incorporates the concept of magnetic force interaction into the existing social force model of an agent movement. It simulates the agent's interaction within their boundaries, preventing any collisions from occurring and to follow others when the agents move in a same direction. The proposed method incorporated agents' intelligence to choose the shortest path in their movements towards their destinations. When the number of agents increases in the simulation environment, the model is able to produce a self-organized behaviour, such as the lane formation behaviour pattern when the agents are in a bi-directional movement as well as in a counter flow movement at intersections.