Towards an Improvement of Bisection-Based Adaptive Random Testing

Bisection-based adaptive random testing (B-ART) is a lightweight method for test case generation. However, its failure detection effectiveness is not so ideal. In this paper, two strategies are designed to overcome the disadvantage as above. The first one is the flexible partitioning strategy, in which the splitting line (or plane) is determined according to the relative position of the test case within the region to be bisected. Secondly, given an empty sub-region, candidate strategy is applied to select an appropriate candidate whose boundary distance is the largest in the set of random candidates as the next test case. Based on these two strategies, an improved algorithm named B-ART-FPCS is proposed. To verify the effectiveness of B-ART-FPCS algorithm, simulation analysis is performed for the comparison between the original B-ART and B-ART-FPCS. The experimental results show that B-ART-FPCS exhibits the stronger failure detection capability than B-ART for block failure pattern and most cases of point pattern. In addition, the linear-order time complexity of B-ART-FPCS is analyzed in theory and confirmed by experiments.