Design Synthesis: A Model of Hierarchical Case-Based Reasoning

Abstract

A variety of artificial intelligence techniques have been used in attempts to automate design synthesis tasks. Two common approaches are case-based and decompositional design. While powerful techniques in their own right, their integration has lead to a new generation of design synthesis systems capable of tackling a larger range of problems with greater effectiveness. In this paper previous attempts at integrating these approaches are examined in a number of design systems. Although significant advances have been made, important shortcomings still exist. The main focus of this paper is to address the limitations of these design synthesis models. To this end Deja Vu, a new hybrid model of design synthesis, has been developed and is described. Deja Vu integrates decompositional and case-based approaches in a framework that exploits the power of experiential knowledge, and benefits from far greater domain applicability when compared to existing design techniques. Two implementations of this model, that are targeted at real-world software design tasks, have yielded encouraging preliminary results and are also described in this paper.