Enhanced lattice-based adaptive random testing
Tsong Yueh Chen, De Huang, Fei‐Ching Kuo, Robert Merkel, Johannes Mayer · 2009
Adaptive Random Testing (ART) has been proposed to improve the fault-detection capability of Random Testing (RT). Lattice-based ART (L-ART) is a distinctive ART method which generates test cases by systematically placing and then randomly shifting lattice nodes in the input domain. Previous studies showed that L-ART has a better fault-detection capability than RT, at the same generation cost. Test cases of L-ART however may be highly concentrated on certain parts of the input domain - a "skewed distribution of test cases". Because of this skewed distribution, when failure regions coincidentally reside in the area where L-ART selects a high density of test cases, L-ART can have a better fault-detection capability than when failure regions are in the low density area. Since failure regions can be in any part of the input domain, this dependency of fault-detection capability on the failure region location is undesirable. We have investigated the cause of such skewed test case distributions using L-ART. Based on our observations, we propose an enhancement to L-ART, which not only has a less-skewed test case distribution, but also demonstrates better and more consistent fault-detection capability than the original L-ART.