Improving the efficiency and applicability of machine planning.

摘要

Although machine planning is an active research field, machine planners have seldom been used outside of AI labs, due partly to their computational complexity and partly to the difficulty of using a planner as one component of a larger system. In contrast, existing manufacturing systems have a variety of efficient algorithms for inventory management, but rely extensively on human intervention to design assembly lines and switch factory production from one product to another.; Our research seeks to bridge the gap between AI Planning systems and manufacturing control systems. We show that manufacturing representations, including assembly trees and Bills of Material can be converted into domain operators for a machine planning system. The output of the planner can be converted into four matrices, which a discrete event controller uses in a prototype manufacturing cell. In addition, two of these matrices allow for an efficient and improved version of the Manufacturing Resource Planning Problem.