Self-Organization and Collective Dynamics of Cognitive Self-Steering Active Agents in Semi-Dense Systems

The efficient movement of cognitive active agents, such as pedestrians, relies on adapting their motion in response to neighboring particles. We present a generic model for such systems, which consists of intelligent active Brownian particles (iABPs) in two spatial dimensions, moving at a constant speed along their heading direction. Essential features of this model are (i) steering torques that change the direction of motion related to visual perception for distancing and goal fixation, and (ii) a distinction between on-coming and co-moving agents. We examine semi-dense systems where excluded-volume interactions are negligible, and the constant-speed assumption applies. We use this model to describe two scenarios: motile agents at finite density, aiming to maintain a large neighbor distance, and agents at a three-way crossing, where there is no simple way to self-organize. In the first case, we find that the agent dynamics for medium-to-large vision angles only depend on the ratio Pe3/2/Ω of their Péclet number Pe and maneuverability Ω, while for narrow vision angles, an avoidance-induced flocking state is observed. In the latter scenario, where agents aim to reach a goal, their behavior is governed by the competition of maneuverability, goal-fixation, the vision angle and the inflow rate.