Potential of Heterogeneity in Collective Behaviors: A Case Study on Heterogeneous Swarms

Abstract

Research in swarm robotics and collective behaviors is often focused on homogeneous swarms. However, heterogeneity in behaviors can be advantageous as we know, for example, from studies on social insects. Our objective is to study the hypothesis that there are potential advantages of heterogeneous swarms over homogeneous swarms in an aggregation scenario inspired by behaviors of juvenile honeybees. Even without task switching – that is, with predefined, static roles for certain swarm fractions – we find in our case study that heterogeneous swarms can outperform homogeneous swarms for a predetermined set of basic behaviors. We use methods of evolutionary computation to define behaviors imitating those found in honeybees (random walkers, wall followers, goal finders, immobile agents) and also to find well-adapted swarm fractions of different predetermined behaviors. Our results show that non-trivial distributions of behaviors give better aggregation performance.