Tabu Search in Covering Array Generation

Combinatorial testing can effectively detect faults caused by the interaction among the parameters of the system under test. In its 30 years of the development, covering array generation has been one of the key research areas, and relevant research articles have reached more than 200. As effective algorithms to generate covering arrays, existing tabu search algorithms have some advantages on the size of covering array, but there is still much room for improving the solution quality and calculation speed. Furthermore, the practical application of the existing algorithms is poor, because they can neither take account of constrains nor generate variable strength covering arrays. To solve the above problems, this paper proposes a new tabu search algorithm. Three improved aspects are presented. 1) The process of parameter tuning is divided into two stages: pair-wise and climbing to ensure that the optimal configuration is hit with a minimum number of configurations so as to further improve the size of covering arrays. 2) In order to improve the speed, the algorithm is parallelized. 3) Constrains and variable strength handling are added to make the algorithm adapt to various test scenarios. Experimental results show that the proposed algorithm has the advantage on the size of various covering arrays, such as fixed strength covering arrays, variable strength covering arrays and covering arrays with constraints. At the same time, the parallelization results in the increase of average speed of the algorithm by about 2.6 times. © Copyright 2018, Institute of Software, the Chinese Academy of Sciences. All rights reserved.