Hybrid long-range collision avoidance for crowd simulation

Abstract

Abstract-Local collision avoidance algorithms in crowd simulation often ignore agents beyond a neighborhood of a certain size. This cutoff can result in sharp changes in trajectory when large groups of agents enter or exit these neighborhoods. In this work, we exploit the insight that exact collision avoidance is not necessary between agents at such large distances, and propose a novel algorithm for extending existing collision avoidance algorithms to perform approximate, long-range collision avoidance. Our formulation performs long-range collision avoidance for distant agent groups to efficiently compute trajectories that are smoother than those obtained with state-of-the-art techniques and at faster rates. Comparison to real-world data demonstrates that crowds simulated with our algorithm exhibit an improved speed sensitivity to density similar to human crowds. Another issue often sidestepped in existing work is that discrete and continuum collision avoidance algorithms have different regions of applicability. For example, low-density crowds cannot be modeled as a continuum, while high-density crowds can be expensive to model using discrete methods. We formulate a hybrid technique for crowd simulation which can accurately and efficiently simulate crowds at any density with seamless transitions between continuum and discrete representations. Our approach blends results from continuum and discrete algorithms, based on local density and velocity variance. In addition to being robust across a variety of group scenarios, it is also highly efficient, running at interactive rates for thousands of agents on portable systems.

Modeling, Evaluation, and Scale on Artificial Pedestrians

“…The work by Narain et al [2009] presents a mixed representation (discrete and continuous) for the crowd with a collision avoidance model adequate for high density crowds. In the same line, the work by Golas et al [2014b] develops a collision avoidance model capable of working in the full range of densities. Combined with a pressure field to include inter-personal stress and individual discomfort, the model can simulate crowd turbulence [Golas et al 2014a].…”

“…From the data shown in Table II we can extract two main conclusions: the prevalence of the microscopic models respect to the macroscopic ones and the concentration of works in two bibliographic areas: on the one hand research publications related [Reynolds 1987;1999;Ondrej et al 2010;Martinez-Gil et al 2014 et al 2008a;Pettré et al 2009;Karamouzas et al 2009;Guy et al 2009;van den Berg et al 2011;Olivier et al 2012;He et al 2016] Path finding and route choice Mechanics Social force [Freimuth and Lam 1992;Helbing et al 1997c;Helbing et al 1997a;Gilman et al 2005 [Chenney 2004;Treuille et al 2006;Narain et al 2009;Golas et al 2014b] and virtual Agency Agent-based [Kuffner 1999] environments Datadriven Data-driven [Lerner et al 2007;Lee et al 2007;Ju et al 2010;] CA CA [Kneidl et al 2013 Yilmaz et al 2009;Pluchino et al 2014] to Computer Graphics and, more specifically, Computer Animation (Group A); on the other hand, publications related to Transportation Research. This indicates that fundamental studies have made way to research in fields where the models are applied.…”

Behavioral Spherical Harmonics for Long-Range Agents’ Interaction

“…Early, longdistance obstacle detection avoids the congestion so that each group rotates around the center instead of having a congestion there. We obtained interactive performance and reproduced the behavior modeled in [7] with a much simpler method than the original study's mixed approach, which used both continuum and discrete methods.…”

“…Behavioral comparison with local models. We tried to replicate the same test benchmark of Golas et al [7], where diametrically opposed groups try to move in opposite directions in space. Early detection of distant groups, combined with knowledge of agent direction, enable them to avoid the congestion in the center of the space.…”

“…Such long-range interaction requires more computation than local collision avoidance, even for distributed and parallel simulation. HybridCrowd [7] performs approximate, long-range collision avoidance via two approaches: low-density crowds are modeled with discrete methods, while highdensity crowds exploit continuum methods.…”