Towards a unified agent architecture that combines rationality with reactivity

Abstract

In this paper we analyse the differences between rational and reactive agent architectures, and propose a uniform agent architecture that aims to capture both as special cases. For this purpose we employ a proof procedure, to control the agent's behaviour, which combines definitions with integrity constraints. The proof procedure is general, and has been shown elsewhere to unify abductive logic programming, constraint logic programming and semantic query optimisation. We also employ a resource-bounded formalisation of the proof procedure which allows the agent's reasoning to be interrupted and resumed, so that observations and actions can be performed.KeywordsLogic ProgramLogic ProgrammingGoal StatementIntegrity ConstraintHorn ClauseThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.