A First Order Theory of Planning, Knowledge, and Action

William J. Rapaport, Leora Morgenstern
1988 Journal of Symbolic Logic (JSL)  
Most AI planners work on the assumption that they have complete knowledge of their problem domain and situation so that planning an action consists of searching for an action sequence that achieves some desired goal. In actual planning situations, we rarely know enough to map out a detailed plan of action when we start out. Instead, we initially draw up a sketchy plan and fill in details as we proceed. This paper presents a formalism based upon a syntactic logic of knowledge which is expressive
more » ... enough to describe this flexible planning process. 99 100 SESSION 3 I. In~u~on Most AI planners work on the assumption that they have complete knowledge of theh¢ problem domain and situation, so that planning an action consists of searching through a pre-packaged list of action operators for an action sequence that achieves some desired goal Real life planning rarely works this way, because we usually don't have enough information to map out a detailed plan of action when we start out. Instead we initially draw up a sketchy plan and fill in details as we galn more exact infonuation about the world. A robust planning system must have sLrailar capabilities if it is to work ha a complex domain. In particular, it needs a solid theory of knowledge, action, and communication, so that it can produce a coherent plan even when its knowledge of the problem domain and problem situation is incomplete. Some work has already been done in this field, most notably by Robert Moore [Moore 1980], who has developed a theory of knowledge and action in which it is possible, for example, for an agent who knows the combination of a safe to reason that he knows how to open the safe. Due to the inherent limitations of Moore's formalism, however, Moore cannot attack the problems of knowledge, action, and communication in their full generality. This paper presents a logic that is expressive enough to deal with the problems that faced Moore;s system. In the next section of this paper, we review Moore's work on knowledge and action and demonstrate its logical limitations. Afterwakds, we present an alternate approach, discuss its power, and show how it can be integrated with Moore's work on knowledge and action. Finally, we show how this logic of knowledge and action can be extended to a more robust theory of planning, and demonstra~ solutions to a number of A FIRST ORDER THEORY OF PLANNING, KNOWLEDGE, AND ACTION probieras that cannot be handled by current theories of knowledge and action.
doi:10.2307/2274547 fatcat:456zi3mzyzhrzbculiv27f3hlm