Self-organized Flocking with Conflicting Goal Directions

In flocking, a large number of individuals move cohesively in a common direction. Many examples can be found in nature: from simple organisms such as crickets and locusts to more complex ones such as birds, fish and quadrupeds. Reynolds was the first to propose a computational model of flocking [1]. The behavior of each individual is made of three parts: separation, cohesion, alignment. Separation means that the individual moves away from its neighbors. Cohesion means that the individual stays close to its neighbors. Alignment means that the individual matches the velocity of its neighbors. This paper studies flocking in the robotics setting. One of the earliest attempt to realize flocking in robotics was done by Matariç [2]. She created a set of “basic behaviors”: safe-wandering, aggregation, dispersion and homing. Turgut et al. [3] implemented flocking on real robots using two behaviors: proximal control and alignment control. Proximal control combines the separation and cohesion components and was realized using the framework of artificial physics as done by Spears et al. [4]. Alignment control is realized through a novel sensing system which they called virtual heading sensor. More recent research in biology showed that only a small group of informed individuals who have information about a desired goal direction is sufficient lead the whole group in that direction [5]. These leaders are implicit, in the sense that the rest of the swarm is not aware of their presence. Inspired by [5], Çelikkanat [6] extended [3] by providing the goal direction to only a proportion of the robots, which they referred to as informed robots. They found that, similarly to [5], only a minority of informed robots is enough to guide the whole group. In this paper, we study flocking of a swarm of robots when information about two distinct goal directions is present in the swarm. This case can be instantiated