Supporting human interaction with robust robot swarms

Abstract

In this paper we propose a bio-inspired model for a decentralized swarm of robots, similar to the model proposed by Couzin [5], that allows for dynamic task assignment and is robust to limited communication from a human. We provide evidence that the model has two fundamental attractors: a torus attractor and a flock attractor. Through simulation and mathematical analysis we investigate the stability of these attractors and show that a control input can be used to force the system to change from one attractor to the other. Finally, we generalize another of Couzin's ideas [4] and present the idea of a stakeholder agent. We show how a human operator can use stakeholders to responsively influence group behavior while maintaining group structure.