Evolving Herding Behaviour Diversity in Robot Swarms

Abstract

Behavioural diversity has been demonstrated as beneficial in biological social systems, such as insect colonies and human societies, as well as artificial systems such as large-scale software and swarm-robotics systems. Evolutionary swarm robotics is a popular experimental platform for demonstrating the emergence of various social phenomena and collective behaviour, including behavioural diversity and specialization. However, from an automated design perspective, the evolutionary conditions necessary to synthesize optimal collective behaviours (swarm-robotic controllers) that function across increasingly complex environments (difficult tasks), remains unclear. Thus, we introduce a comparative study of behavioural-diversity maintenance methods (swarm-controller extension of the MAP-Elites algorithm) versus those without behavioural diversity mechanisms (Steady-State Genetic Algorithm), as a means to evolve suitable degrees of behavioural diversity over increasingly difficult collective behaviour (sheep-dog herding) tasks. In support of previous work, experiment results demonstrate that behavioural diversity can be generated without specific speciation mechanisms or geographical isolation in the task environment, although the direct evolution of a functionally (behaviorally) diverse swarm does not yield high task performance.