Reasoning about Plan Revision in Agent Programs

Abstract

This talk is on reasoning about agent programs written in Belief, Desire and Intention (BDI) agent programming languages. BDI programming languages (for example, [1], [2], [3]) have high-level programming primitives which correspond to the beliefs, goals and plans of an AI agent. A program contains a set of rules which allow the agent to adopt plans given its current beliefs and goals. Plans are essentially imperative programs. For example, an agent may have a rule which says that if it believes that it is currently located in room 1 and its goal is to be in room 2, then a suitable plan to adopt would be to exit room 1, turn right, move forward for 3 meters, turn right, and enter room 2.