A Multi-tiered Approach to Context and Information Sharing in Intelligent Agent Communities

This paper presents an expanded role for context in agent communities. We propose a multi-tiered approach in which the contextual information is segregated and encapsulated in three different contextual levels, namely Local. Focal. and Zonal. Each level contains specific kinds of contextual information, thereby pruning the search space and making the process of reasoning about context more manageable. These levels can be thought of as super-contexts within which each of the subordinate contexts become potentially valid. Such a multi-tiered approach helps organize. manage. select, and maintain the large amount of contextual information required by future multi-agent systems.