Automatic test pattern generation with BOA

Abstract

We introduce a Bayesian Optimization algorithm (BOA) for the automatic generation of test sequences (ATPG) for digital circuit. We compare our approach, named BOATPG, to the two most known evolutionary approaches to ATPG (GATTO and STRATEGATE) and the currently most promising non-evolutionary approach to ATPG (namely, SPECTRAL ATPG). We show that our simple approach can easily outperform GATTO and performs as good as a more complex evolutionary approach like STRATEGATE. We also show that when BOATPG is coupled with spectral approach for seeding the population of initial test sequences, the resulting hybrid system, SBOATPG, performs better than the plain BOATPG although the improvements over SPECTRAL ATPG are limited. © Springer-Verlag Berlin Heidelberg 2006.