17-03-2014, 10:37 PM
Abstract
As a software technique, expert system technology has several advantages over more traditional programming techniques. Chief
among these are the ease of modifiability of the code and ease of representation and use of ad-hoc'knowledge. One of the areas
where expert system technology has been successfully applied commercially has been in the area of Electronic Design
Automation (EDA). The demand for more sophisticated computer based tools in the EDA has been driven by the improvements
in semiconductor technology along with the increasing complexity of systems that are being designed. The improvements in the
semiconductor technology has made implementation of designs in application specific integrated circuits (ASIC) a cost effective
solution in many areas. The rapid improvements in semiconductor technology has been coupled with a shortening of product
life-cycles. Both of these trends has produced pressure to improve the efficiency of the design process. Because of the
complexity of most design problems, it has not been possible to apply traditional programming techniques in solving many of the
emerging design problems associated with using ASICs without making a great deal of simplifying assumptions or accepting
solutions that are far from optimal. Recently, expert system techniques that rely on task specific knowledge have been very
successfully applied in the design area However, because these expert systems rely on large amounts of domain knowledge for
their problem-solving power, a different set of issues as compared to writing traditional programs must be addressed in building
these systems. This paper examines the trends in the use of expert systems and the associated problem of knowledge acquisition
in several areas of ASIC design.
As a software technique, expert system technology has several advantages over more traditional programming techniques. Chief
among these are the ease of modifiability of the code and ease of representation and use of ad-hoc'knowledge. One of the areas
where expert system technology has been successfully applied commercially has been in the area of Electronic Design
Automation (EDA). The demand for more sophisticated computer based tools in the EDA has been driven by the improvements
in semiconductor technology along with the increasing complexity of systems that are being designed. The improvements in the
semiconductor technology has made implementation of designs in application specific integrated circuits (ASIC) a cost effective
solution in many areas. The rapid improvements in semiconductor technology has been coupled with a shortening of product
life-cycles. Both of these trends has produced pressure to improve the efficiency of the design process. Because of the
complexity of most design problems, it has not been possible to apply traditional programming techniques in solving many of the
emerging design problems associated with using ASICs without making a great deal of simplifying assumptions or accepting
solutions that are far from optimal. Recently, expert system techniques that rely on task specific knowledge have been very
successfully applied in the design area However, because these expert systems rely on large amounts of domain knowledge for
their problem-solving power, a different set of issues as compared to writing traditional programs must be addressed in building
these systems. This paper examines the trends in the use of expert systems and the associated problem of knowledge acquisition
in several areas of ASIC design.