The Effect of Gossip in Heterogeneous Swarms

From large herds of migrating bison to human groups communicating to converge on a policy decision, there are various instances of decentralized swarms where agents with differing amounts of information must converge their behavior. Peer-to-peer decentralized communication can allow swarms to more reliably converge on a target, something that has been studied in existing gossip modelling and collective consensus work. In this paper, we incorporate a gossip model in a flocking algorithm, which allows the flock to be controlled by a leader in a decentralized manner, using gossip to disseminate the signal about a target direction. We show that, by varying our defined parameters such as agent reliability, shyness, we can modulate the swarm dynamics. We find that the swarm converges on the leader’s target direction to within 5 degrees.