資訊物件萃取理論的計算複雜度分析

Abstract

在傳統的軟體設計導向中，資訊物件的定義都是從實體物件的外觀所決定，然後再分析並控制實體物件的行為。本論文則以一相反的角度出發，首先先分析系統中實體物件的動態行為，進而將資訊物件定義出來，其定義的方式是根據裴式圖中系統行為模式的穩態不變量。文中針對資訊物件萃取提出四種方法進行比較，分別是啟發法（heuristic approach）、演算法（algorithmic approach）、枝界法（branch and bound）及A*法（A*approach）。從最後的結果中顯示，利用這種新方法的優點是可以刪除不一致（inconsistent）及多餘（redundant）資訊物件。

Traditonally information objects are defined with respect to the physical objects in a software-based control system. Then the behavior of physical objects are analyzed and controlled. In this thesis, a completly opposite method is proposed. That is to analyze the behavior of physical objects at first. Then the definition of information objects are defined through the place-invariant analysis in Petri net based behavior model. Four different computation models in the place-invariant analysis are compared: heuristic approach, algorithmic apporach, branch and bound approach and A* approach. The research result shows the advantages of this new approach which removes the inconsistent and redundant information objects.