Functional Strips: A More Flexible Language for Planning and Problem Solving

Abstract

Effective planning requires good modeling languages and good algorithms. The Strips language has shaped most of the work in planning since the early 70's due to its effective solution of the frame problem and its support for divide-and-conquer strategies. In recent years, however, planning strategies not based on divide-and-conquer and work on theories of actions suggest that alternative languages can make modeling and planning easier. With this goal in mind, we have developed Functional Strips, a language that adds first-class function symbols to Strips providing additional flexibility in the codification of planning problems. This extension is orthogonal and complementary to extensions accomodated in other languyages such as conditional effects, quantification, negation, etc. Function symbols, unlike relational symbols, can be nested so objects need not be referred to by their explicit names and as a result more efficient encodings can be provided. For example, a problem like the 8-puzzle can be codified in terms of four actions with no arguments; Hanoi, can be codified with a number of ground actions independent of the number of disks; resources and constraints can be easily represented, etc. Functional Strips is both an action and a planning language in the sense that actions are understood declaratively in terms of a state-based semantics and operationally in terms of efficient updates on state representations. In this paper, we present the language, the semantics and a number of examples, and discuss possible uses in planning and problem solving.