Emergent Crowd Grouping via Heuristic Self-Organization

TL;DR

This work tackles the problem of achieving efficient crowd motion without explicit grouping by enabling self-organized, implicit grouping among moving agents. It rotates each agent's original preferred velocity $\mathbf{P}^t$ by an angle $\Theta^t$ computed from local crowd information to produce $\mathbf{G}^t$, which is then used by a standard collision-avoidance framework (ORCA) to update velocities. Key contributions include a bottom-up mechanism for emergent grouping, quantitative evidence of reduced congestion and improved alignment between $\mathbf{P}^t$ and the actual velocity $\mathbf{V}^{t+1}$, and a Python package for easy reuse. The approach offers practical benefits for real-time crowd animation and simulation by fostering natural lane formation and lower internal conflicts in densely packed scenes.

Abstract

Modeling crowds has many important applications in games and computer animation. Inspired by the emergent following effect in real-life crowd scenarios, in this work, we develop a method for implicitly grouping moving agents. We achieve this by analyzing local information around each agent and rotating its preferred velocity accordingly. Each agent could automatically form an implicit group with its neighboring agents that have similar directions. In contrast to an explicit group, there are no strict boundaries for an implicit group. If an agent's direction deviates from its group as a result of positional changes, it will autonomously exit the group or join another implicitly formed neighboring group. This implicit grouping is autonomously emergent among agents rather than deliberately controlled by the algorithm. The proposed method is compared with many crowd simulation models, and the experimental results indicate that our approach achieves the lowest congestion levels in some classic scenarios. In addition, we demonstrate that adjusting the preferred velocity of agents can actually reduce the dissimilarity between their actual velocity and the original preferred velocity. Our work is available online.

Figures (6)

• Figure 1: The initial distribution of agents of two scenarios
• Figure 2: Curves of $M_1$ (the dissimilarity between preferred velocity and actual velocity of agents) over time steps of different crowd simulation models in four scenarios. In each scenario, although our approach exhibits a brief period of significant dissimilarity at the initial stage, it maintains relative stability throughout the subsequent simulation process.
• Figure 3: Curves of $M_2$ (the crowd congestion level) over time steps of different crowd simulation models in four scenarios. Our method consistently maintains a low level of congestion when agents encounter others (time steps around the middle of the x-axis)
• Figure 4: Agent states for three crowd simulation models in the AsyCircle scenario. Both ORCA and Implicit exhibit congestion at the area where the agents converges in the middle. Our method results in loosely formed groups of agents with similar directions before they encounter each other. Upon meeting, there is no significant congestion observed.
• Figure 5: Agent states for three crowd simulation models in classic three-agent scenario. Compared to the close encounters in ORCA, Implicit and our method demonstrate more realistic behaviors by incorporating necessary avoidance maneuvers and detours.