Development of link tension index for flocking agents

Techniques are needed to control mobile agents as a flock. These agents observe each other within a neighboring distance. The agent and its observing relation can be considered as nodes and links of a graph. To control the flock, the graph needs to be connected. However, following the rules of flocking, a flock splits into smaller flocks; i.e., the graph becomes disconnected. We thus propose a link tension index that predicts which links in the graph will remain intact and which will break a short time later. We believe that the index can be applied to the control of a flock. The link tension index represents whether the maximum distance of the link between two agents is within a local observing area. In an actual flock, there is no guarantee that this local index always makes correct predictions because there are other agents besides the agents directly connected. Thus, this paper evaluates the validity of the index for a flock of many agents. We show that the index correctly predicts the links of the flock that will remain intact a short time later but poorly predicts the links that break because the index focuses on the energy of the relative motion between two agents.