Evolutionary Optimization in Code-Based Test Compression

Abstract

We provide a general formulation for the code-based test compression problem with fixed-length input blocks and propose a solution approach based on Evolutionary Algorithms. In contrast to existing code-based methods, we allow unspecified values in matching vectors, which allows encoding of arbitrary test sets using a relatively small number of code-words. Experimental results for both stuck-at and path delay fault test sets for ISCAS circuits demonstrate an improvement compared to existing techniques.