How to Progress a Database II: The STRIPS Connection

Ever since STRIPS was first introduced (Fikes and Nilsson [3]), i ts logical semantics has been problemat ic. There have been many proposals in the l i terature (e.g. Lifschitz [4], Ero l , Nau and Subrahmanian [2], Bacchus and Yang [ l ] ) . These all have in common a reliance on metatheoretic operat ions on logical theories to capture the add and delete lists of STRIPS operators, b u t i t has never been clear exactly what these operations correspond to declaratively, especially when they are applied to logically in complete theories. In this paper we provide a semantics for STRIPSl ike systems in terms of a purely declarative s i tuat ion calculus axiomat izat ion for actions and their effects. On our view, STRIPS is a mechanism for comput ing the progression (L in and Reiter [6], Pednault [8]) of an in i t ia l s i tuat ion calculus database under the effects of an act ion. We i l lustrate this idea by describing two different STRIPS mechanisms, and prov ing their correctness w i t h respect to their s i tuat ion calculus specifications.