WEAVER Implementation

Abstract

There are several steps involved in the design and implementation of a knowledge-based expert system: the initial familiarization of the knowledge engineer with the application area, the initial gathering of knowledge, the specification of a suitable problem solving paradigm, the selection of a formalism to represent the domain specific knowledge, the control structure which specifies when and how a chunk of knowledge should be applied, and finally the modular and incremental modification of the system based on a set of test problems until satisfactory performance is achieved. In the case of WEAVER, the familiarization with the detailed routing of VLSI chips was done in the summer of 1982 when the author interviewed several of the best designers at INTEL Corporation [Kramer 82, Wilde 82]. Because of the author’s background and familiarity with the application domain he acted later on as an expert to modify and improve the system. WEAVER is organized around a multidirectional opportunistic search and constraint propagation paradigm. It is a multidirectional opportunistic search because it is a global router. Rather than concentrating on one net until it is completely routed WEAVER concentrates on all the nets and opportunistically routes segments of all the nets when circumstances present multiple options. It uses the constraint propagation paradigm when, as the result of global routing it finds itself in a situation where there is at most one choice of action left. The global routing feature of WEAVER becomes more clear in Chapter 6 when an example describes the step by step routing of a channel.