SCHEMATIC GENERATION WITH AN EXPERT SYSTEM

This paper reports on a study of schematic generation by a knowledge-based system (KBS). Schematics must demonstrate features (continuity of signal flow and logical clustering) of a design w hile remaining simple (few turns and crossings). Algorithmic and manual techniques deal badly with the multiple conflicting objectives. A successful switchbox router in a KBS suggested the same approach to schematic generation. Schematic generation in a KBS still needs to be partitioned into sub-problems to make it tractable. Subproblems include ordering of symbols and the placement of nets (i.e., routing). The ordering of symbols in the direction of information flow is based on a precedence augmented with a rule-based component which recognizes cycles and breaks them intelligently. A heuristic length-based ordering in the other direction tries to minimize crossings. The abstract representation of nets has a major impact on the parsimony of the knowledge representation. Placements of progressively more detailed parts of nets try to achieve desirable configurations specified by rules; placements generate constraints on subsequent details which are propagated forward. The KBS paradigm is particularly applicable to the routing steps. © 1990 IEEE