A logic-based perspective on agent reconfiguration: Preliminary report

Abstract

We investigate the problem of maintaining and reasoning with different configurations of a logic-based agent. Given specific contexts, there may be several possible usual configurations that the agent's knowledge base can be in, and that one may want to access at different times. This can happen due to foreseeable exceptional situations one has to cater for, or different environments in which the agent may have to operate, or simply due to upgrades of the agent's initial configuration. In all these cases, there is a need for a system capable of managing possibly conflicting versions of the knowledge base and allowing the agent to switch between any two given configurations at run time. Building on Franconi et al.'s framework for propositional knowledge base versioning, here we establish the logical foundations for a general semantic-based architecture of such a system. Central to our approach is the notion of logical difference, which allows us to determine the essential pieces of information on which two given configurations of an agent differ.