From Scattered to Focused: Task-Dependent Connectivity in Honey Bees, with Midge Swarms and Bird Flocks

Abstract

Collective motions in biological organisms such as honey bees, midges, and birds exhibit remarkable coordination and adaptability, emerging from local interactions among individuals, including those with tight constraints on neural material. Understanding the underlying mechanisms behind the collective behavior of biological swarms informs the dynamics of decentralized systems, including engineered aerial swarms. This study provides a systematic graph network extraction formulation applicable to physical domains trajectories including the practical effects of data dropouts, multiple agents, and multiple degrees of freedom. The analysis uses graph network modeling (in which nodes represent agents and edges denote interactions) to uncover the connectivity patterns governing swarm dynamics and quantify their differences. Three network identification methods, including sparse regression, Granger-causality, and crosscorrelation, are applied to determine the underlying graph from measured animal trajectories. An additional geometric anisotropy analysis used in bird flocks is also applied. The methods are applied to experimental recordings of honey bees involving scattered and focused "return-to-hive" behaviors, to midges in reproductive-related aggregations, and to flocking passerine birds. Across these studies, the number of agents a focal agent receives information from was quantified by analyzing the connectivity of identified interaction graphs. The results reveal honey bees show significant differences in connectivity and slight differences in anisotropy between experiments involving scattered and focused "return-to-hive" behaviors. In particular, scattered honey bees showed causal relationships up to 10 neighbors (consistent with flocking jackdaw data), while focused honey bees showed connections with up to 2-3 agents, and midges showed slightly lower connectivity. Finally, cross-correlation analysis indicated honey bees exhibit distance-sensitivity in close proximity, while flocking birds exhibit a distance-dependent correlation decline. Together, these results provide evidence for task-dependent in-degree modulation in honey bees, and provide a systematic analysis formulation that highlights the variety of approaches to coordinated motion, including those with and without geometric anisotropy. These results inform the design of bioinspired networked systems.