Leader Selection and Clusterization Algorithms in a Static Robot Swarm

Abstract

The paper presents the problem of clustering and leader selection in a group of robots, using a static swarm model - fixed network at some point in time, consisting of the robots connected to one another via the communication channels. Robots use only local interaction, the topology of the swarm and the number of the robots is not known beforehand. It is proposed to take into account the relative positions of the robots and their neighbors, i.e., their local topology, which is known to them, and allows them, in the long run, to choose their leader out of the robots located close enough to the topological center of the whole group. It is known that the group has peripheral robots - those which have not all the communication channels occupied. They initiate the leader selection procedure by transmitting its weight to the center of the group. This allows to create there a subgroup of robots, with the biggest weights, one of which becomes the leader. It is considered as an option, when the static topology of the swarm changes, i.e. some robots are eliminated in the process of voting. It is demonstrated that in all these cases the leader selection algorithm succeeds. In addition, a clustering algorithm is proposed to solve the problem of the functional differentiation of robots, which will quickly produce their integration into subgroups. The conducted computing experiments prove the efficiency of the algorithms.