Cluster analysis of collective behavior for a robotic swarm

Abstract

Swarm robotics is a new research field of multi-robot systems which generally consists of many homogeneous autonomous robots without a global controller. Since a robotic swarm is controlled by an emergent way of many interactions with the other robots or an environment, such as a result of self-organization, robot learning or artificial evolution, no method has been known to grasp the macroscopic collective behavior in a practical sense, according to the best of our knowledge. In this paper, based on this situation, a novel method for analyzing the macroscopic collective behavior inspired by a technique in the field of complex network is proposed. The effectiveness of the proposed method is demonstrated on a robotic swarm behavior for the cooperative transport problem by extracting the community structure based on the modularity optimization. © 2013 Springer-Verlag.