Artificial Intelligence Techniques in Design and Test

A rtificial intelligence (Al) has long been the topic of university research. However, only recently has this work begun providing usable Al system development tools allowing applications of Al in other research areas. This growing availability has fostered a recent trend toward applying Al techniques to improve CAD tools used to design and test integrated circuits. Both industry and universities are investigating the possibilities. This issue of D&T provides a sampling of current work in applying Al to the design and test fields. Three of the articles represent university research, while the other three cover industrial efforts. Adding to the balance of this issue is the fact that three articles deal with design issues, two with test issues, and one with human factors, or user interface. The first article, "A Rule-Based Expert System for Optimizing Combinational Logic," by Aart de Geus and William Cohen, illustrates their current work in industry. The authors present a design aid that can substitute equivalent gate configurations to improve the area and speed of a design. The implementations generated by their system are comparable in area and speed to circuits designed by human experts. The second article describes how researchers at AT&T Bell Laboratories use expert systems to automate portions ofthe integrated circuit design process. "The VLSI Design Automation Assistant: From Algorithms to Silicon," by T.J. Kowalski and colleagues, presents a collection of design aids for implementing a wide range of system design decisions. Part of this system, the VLSI Design Automation Assistant, was transferred from university research. Robert Mueller and Joseph Varghese, in their article, "Knowledge-Based Code Selection Methods in Retargetable Microcode Synthesis," apply expert system technology to generate microcode. Their article, based on university research, illustrates the use of both knowledge-based techniques and procedural knowledge in this design step. They show how a knowledgebased synthesis system can act as the final stage of a microcode compiler. The fourth article, "A Knowledge-Based System for Designing Testable VLSI Chips," by Magdy Abadir and Melvin Breuer, describes university research leading toward a methodological framework for testable system design. The authors use a case study to illustrate their approach to a knowledge-based system for applying and evaluating different design-for-testability methods. An industrial application of an expert system, the Machine for Intelligent Diagnosis (MIND), is described in the fifth article. "MIND: An Inside Look at an Expert System for Electronic Diagnosis," by A. Jesse Wilkinson, explains how the MIND diagnostic tool can be applied to automatically troubleshoot a system. The MIND system will be at a beta test site this summer. Another area suitable for applying Al techniques is natural language understanding. The sixth paper, "Natural-Language Interaction for Computer Aided Design -A First Step," by Tariq Samad and Stephen Director, describes the development of a user-friendly front-end to an electronic circuit simulator. The system can respond