Flocking: Don't need no stinkin' robot recognition

Abstract

Flocking is a common and widely studied spatial behavior exhibited by groups. This work highlights inconsistencies in the presentation of motion rules used for flocking, by implementing the well known rule by Hamilton on a robot system. We address a common assumption regarding the form of input to the motion rule: detection of whole agents and suggest that such detection is not necessarily justified. Our multi-robot system successfully exhibits flocking behaviors using an alternative detection method based entirely on low level sensor data. Furthermore, we show (under certain parameter settings), the behaviors exhibited by the agent-based and sensor-based detection are equivalent. We also discuss the various dynamics and implications the chosen detection process has on the behaviors of a motion rule.