From Centralized to Distributed Selective Overhearing

Proceedings of the AAAI Conference on Artificial Intelligence, 21

Volume

Technical Papers

Multiagent Systems

Overhearing is an approach for monitoring open, distributed, multi-agent systems by listening to the routine communications taking place within them. Previous investigations of overhearing assumed that all inter-agent communications are accessible to a single overhearing agent. However, as multiagent systems grow both in size and distribution two problems arise. First, in large-scale settings, an overhearing agent cannot monitor all agents and their conversations, and must therefore be selective in carefully choosing its targets. Second, a single overhearer would encounter difficulties overhearing agents acting in a geographically-distributed environment. This paper tackles these challenges by addressing distributed teams of overhearing agents involved in selective overhearing. Building on prior work on centralized selective overhearing, we consider the consequences of transitioning from overhearing teams working in a centrally-coordinated manner to distributed overhearing teams. In doing so, we distinguish the various factors influencing the level of distribution within these teams and determine their importance in terms of effective selective overhearing.