Much of the history of AI, especially in the 1970s, has involved well-publicized programs that reportedly achieved breakthroughs in their areas. Examples are SHRDLU, AM, and NOAH in natural language understanding, learning, and planning, respectively. Unfortunately, the publicity for such programs overshadowed their actual capabilities, which were in fact very difficult to determine. Consequently, the ideas from these “landmark programs” are difficult, if not impossible, to apply. The research reported here arose from the author’s attempt to use the standard description of NOAH [1] to build a hierarchical planner in his research. Several attempts to rewrite NOAH failed, and instead an algorithm for achieving conjunctive goals in a domain-independent planner emerged. This paper presents and analyzes the algorithm. Other previous work on planning is compared in a uniform framework.

Papers of this type are important if AI is to advance as a discipline. While the language of modal logic used is not to my taste, I recommend this paper as a useful reference for students or researchers building a planner.