Practical reasoning with procedural knowledge: (A logic of BDI agents with know-how)

Abstract

In this paper, we present a new logic for specifying the behaviour of multi-agent systems. In this logic, agents are viewed as BDI systems, in that their state is characterised in terms of beliefs, desires, and intentions: the semantics of the BDI component of the logic are based on the well-known system of Ran and Georgeff. In addition, agents have available to them a library, of plans, representing their ‘know-how’: procedural -knowledge about how to achieve their intentions. These plans are, in effect, programs, that specify how a group of agents can work in parallel to achieve certain ends. The logic provides a rich set of constructs for describing the structure and execution of plans. Some properties of the logic are investigated, (in particular, those relating to plans), and some comments on future work are presented.