Defeasible Argumentation Support for an Extended BDI Architecture

In this work, an agent architecture that combines defeasible argumentation and the BDI model is described. Argumentation will be used as a mechanism for reasoning about beliefs, for filtering desires considering the agent’s current environment, and for selecting proper intentions. The approach allows to define different types of agents and this will affect the way in which desires are filtered and hence, which intention is selected. For performing defeasible reasoning, the approach uses a concrete framework based on a working defeasible argumentation system: Defeasible Logic Programming (DeLP). A set of filtering rules, represented as a defeasible logic program, will be used to represent reasons for and against adopting desires. Thus, based on its perceived or derived beliefs, the agent will argue about which of its desires are achievable in the current situation. To clarify the ideas two applications will be introduced to show two significantly different types of agent that can be implemented using this approach.