Directed motion of cognitive active agents in a crowded three-way intersection

Understanding the navigation through semi-dense crowds at intersections poses a significant challenge in pedestrian dynamics, with implications for facility design and insights into emergent collective behavior. To tackle this problem, a system of cognitive active agents at a crowded three-way intersection is studied using Langevin simulations of intelligent active Brownian particles (iABPs) with directed visual perception (resulting in non-reciprocal interactions) and self-steering avoidance—without volume exclusion. We find that the emergent self-organization depends on agent maneuverability, goal fixation, and vision angle, and identify several forms of collective behavior, including localized flocking, jamming and percolation, and self-organized rotational flows. Additionally, we demonstrate that the motion of individual agents can be characterized by fractional Brownian motion and Lévy walk models across different parameter regimes. Moreover, despite the rich variety of collective behavior, the fundamental flow diagram shows a universal curve for different vision angles. Our research highlights the impact of collision avoidance, goal following, and vision angle on the individual and collective dynamics of interacting pedestrians. The study of self-organisation of pedestrian movement at crossing is important for the design of strategies facilitating pedestrian flow in crowded areas and the mitigation of crowd-related accidents. The authors study the motion of pedestrians using a model inspired from active matter systems finding interesting phases of three interacting streams of agents, including jamming, and the emergence of a vortex state.