Robot Problem Solving* Without State Variables

Abstract

In the original robot problem-solving system based upon an idea presented by Green, situations were described by axioms of first-order logic that explicitly included terms representing states. Theorem proving by resolution was then used to prove, constructively, the existence of a state having the properties of a desired goal. The major disadvantages of that approach were (1) All unchanged and irrelevant properties of the world must be re-established after every state transition; (2) The efficiency of the system is very sensitive to the choice of domains and predicates used in the logical encoding, so that sometimes the most “natural” representations do not lead to practical task descriptions. Waldinger has proposed a scheme for getting around these problems by extending the logic to handle sets and tuples (still keeping an explicit state variable around).