Context-sensitive Personal Space for Dense Crowd Simulation

Real-time simulation of dense crowds is finding increased use in event planning, congestion prediction, and threat assessment. Existing particle-based methods assume and aim for collision-free trajectories. That is an ideal -yet not overly realisticexpectation, as near-collisions increase in dense and rushed settings compared to typically sparse pedestrian scenarios. This paper presents a method that evaluates the immediate personal space area surrounding each entity to inform its pathing decisions. While personal spaces have traditionally been modeled as having fixed radii, they actually often change in response to the surrounding context. For instance, in cases of congestion, entities tend to share more of their personal space than they normally would, simply out of necessity (e.g. leaving a concert or boarding a train). Likewise, entities travelling at higher speeds (e.g. strolling, running) tend to expect a larger area ahead of them to be their personal space. We illustrate how our agent-based method for local dynamics can reproduce several key emergent dense crowd phenomena; and how it can be efficiently computed on consumer-grade graphics (GPU) hardware, achieving interactive frame rates for simulating thousands of crowd entities in the scene.