Multiprocess Software Architecture for AI Problem Solving.

摘要

This dissertation describes the design and development of a knowledge-based artificial intelligence problem-solving organization that is suitable for efficient implementation on a closely-coupled multiprocessor computer system. The method is a result of formulating the problem-solving organization in terms of the hypothesize-and-test paradigm for heuristic search, with communication between the various hypothesizers and testers being effected by writing intermediate results in a shared blackboard-like data base. These hypothesizers and testers are expressed in terms of knowledge sources which represent bodies of suitably ortanized subject-matter knowledge pertinent to the task domain of the problem being solved. The various system organization problems connected with such a multiprocessing scheme are discussed, and solutions to these problems are presented. The major contributions of this work lie in the analysis and solution of the various multiprocessing problems that have arisen in the course of specifying this problem-solving organization.