Representing actions: Laws, observations and hypotheses

We propose a modification ℒ₁ of the action description language script A sign. The language ℒ₁ allows representation of hypothetical situations and hypothetical occurrence of actions (as in script A sign) as well as representation of actual occurrences of actions and observations of the truth values of fluents in actual situations. The corresponding entailment relation formalizes various types of common-sense reasoning about actions and their effects not modeled by previous approaches. As an application of ℒ₁ we also present an architecture for intelligent agents capable of observing, planning and acting in a changing environment based on the entailment relation of ℒ₁ and use logic programming approximation of this entailment to implement a planning module for this architecture. We prove the soundness of our implementation and give a sufficient condition for its completeness.