Exploiting domain physics: Using stigmergy to control cluster building with real robots

Abstract

This paper explores the application of stigmergy as a mechanism for the control and coordination of cluster building with 10 simple, autonomous, mobile real robots. Social insects collectively carry out extraordinary tasks using simple reactive rule sets with little or no memory and without recourse to internal world models over which they reason. Control and coordination of behaviour can be mediated externally through the environment. In particular, building tasks may employ the configuration of each construction phase to cue the next phase. This paper shows how, using physical robots, the deterministic clustering algorithm of Beckers et al [1994] can be modified to create a cluster of objects against the arena wall. The modification takes the form of either applying probabilistic rule selection or by altering the robot sensor morphology. These two approaches illustrate that equivalent qualitative emergent consequences can be generated algorithmically or by exploiting the domain physics.