Needle’s eyes—A general decomposition principle for practical uses

Abstract

If efficient control knowledge is available, search in the state space is a practical way of solving non-toy problems. For many applications the structure of the state space itself offers a reliable means of control: Certain states called “needle’s eye” can be used as partial goals so that less complex problems remain to be solved. Originally, we formalized the “needle’s eye principle” for the acquisition and representation of control knowledge for a model-based planner. We show that it is equally useful for a highly complex configuration task arising in the material sciences. From the crystallographical problem setting a natural problem decomposition is derived, such that a hierarchical configuration procedure can be used as an implicit model of “disorder”. After a brief, informal review of the needle’s eye principle we show that the control knowledge inferred on the top level of our knowledge-based configuration program is, in essence, a chain of needle’s eyes.