In the recent past, several different methods for coordinating behavior in multi-robot teams have been proposed. Common to most of them is the use of communication to coordinate behavior. For many practical applications, however, communication might not be an option (e.g., because of energy constraints of embedded platforms, limited communication range of wireless transmitters, security risks of potential interception of messages in hostile territory, etc.). In this paper we examine alternative, low-complexity, lowcost strategies without communication for coordinated agent behavior. Specifically, we investigate the utility of a "social preference mechanism" and a "pairing mechanism" in territory exploration tasks (with many practical instantiations, e.g., two robots with different sensory capabilities investigating rock formations on a remote planet), where agents have to explore their environment to find and visit k checkpoints, which only count as "visited" when n agents are present at them at the same time. Both mechanisms are intended to increase the likelihood of two agents being at a same checkpoint at the same time. Experimental results indicate that pairing is the better strategy, thus raising interesting questions about tradeoffs between agent complexity and group size (e.g., whether fewer but more expensive agents that have sufficient resources to visit checkpoints individually are a better choice than more less expensive agents).
@inproceedings{schermerhornscheutz06aamas,
title={Social Coordination without Communication in Multi-Agent Territory Exploration Tasks},
author={Paul Schermerhorn and Matthias Scheutz},
year={2006},
month={May},
booktitle={Proceedings of the Fifth International Joint Conference on Autonomous Agents and Multiagent Systems (AAMAS-06)},
pages={654--661}
url={https://hrilab.tufts.edu/publications/schermerhornscheutz06aamas.pdf}
}