Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
6941314 | Pattern Recognition Letters | 2014 | 9 Pages |
Abstract
Crowd behaviour models are divided into agent-based, flow-based and particle-based in terms of whether the behaviour emerges from simulating all people (agents) individually (Koh and Zhou, 2011; Braun et al., 2005; Luo et al., 2008; Pan et al., 2007; Shendarkar et al., 2006; Narain et al., 2009), is programmatically defined a priori using fluid dynamics models (Hughes, 2002, 2003; He et al., 2011), or employ a particle system governed by physical laws (Helbing et al., 2000; Bouvier et al., 1997; Treuille et al., 2006; Cucker and Smale, 2007). In agent-based models, computationally intense problems, such as global navigation, hinder the efficient real-time modelling of thousands of agents. In this paper we present a novel approach to crowd behaviour modelling which couples the agent-based paradigm of allowing high level of individual parametrization (group behaviour between friends, leader/follower individuals) with an efficient approach to computationally intensive problems encountered in very large number of agents thus enabling the simulation of thousands of agents in real time using a simple desktop PC.
Related Topics
Physical Sciences and Engineering
Computer Science
Computer Vision and Pattern Recognition
Authors
Vassilios Kountouriotis, Stelios C.A. Thomopoulos, Yiannis Papelis,