一种基于注意力机制的群集运动模型

Motivated by the capability to compress information efficiently of selective attention mechanism, we proposed a selective attention based flocking model. In this model, individuals sense other members by fixed number of neighbors guided by line of sight method and calculate saliency of these neighbors according to their contribution to local order. Then individuals decide whether to activate themselves to respond to stimuli determined by threshold model. We found a well-defined hierarchical network concerning stimulus transfer among the flock. Individuals switch between hierarchical and egalitarian strategies according to their status of activation. Results reveal that the selective attention can enhance the transfer of social cues and make flock members able to respond to social cues rapidly. The decision to turn caused by external stimuli can propagate across the group quickly using this model, which is consistent with natural flocks.