Investigating the dimensionality problem of Adaptive Random Testing incorporating a local search technique

Adaptive random testing (ART) has been proposed to enhance the effectiveness of random testing. By spreading test cases evenly within the input domain, ART techniques may reduce the number of test cases necessary to detect the first failure by up to 50%. However, the most effective ART strategies are little effective in higher dimen- sions. This fact distinctly affects their applicability since in a real testing area input domains usually are far from being one- or two-dimensional. The present work addresses this problem. It discusses the shortcomings of existing solu- tions and describes how prior knowledge can help solving the problem. Since in general no prior knowledge is avail- able, this work proposes a solution which--though not fully solving the dimensionality problem--seems to be very close to the theoretical optimum. The proposed approach is based on the ideas of the local search technique ´Hill Climbing´.