一句话：提出 DynCA 算法，为高度可配置系统高效生成更小规模的 3-wise 覆盖数组。
高度可配置系统需要组合交互测试（CIT）来检测配置相关故障。其中约束覆盖数组生成（CCAG）是核心难题：在 t=3 时，现有算法虽在 2-wise 上表现良好，却难以兼顾可扩展性与数组规模。本文提出 DynCA，一种针对 3-wise CCAG 的新算法，旨在为大规模实例有效生成更小的覆盖数组。作者在大量真实高度可配置系统上开展实验，与当前最优算法相比，DynCA 将生成的 3-wise 覆盖数组规模平均减小 10.8%，同时仅需约一半运行时间即可生成更小的数组，表明其在测试效率与可扩展性上的优势。
With the increasing demand for software customization, highly configurable systems have become increasingly important. Combinatorial interaction testing (CIT) is widely recognized as an effective technique for testing such systems. The most critical problem in CIT is constrained covering array generation (CCAG), which aims to construct a t-wise covering array (CA) with as small a size as possible, where t denotes the testing strength. Compared with the widely adopted pairwise testing ( 2-wise CIT), 3-wise CIT is capable of detecting more faults and provides greater practical benefits. However, it also introduces significantly higher computational complexity. Although many existing algorithms perform well on 2-wise CCAG, current state-of-the-art CCAG algorithms still suffer from the severe scalability challenge in 3-wise CCAG and have difficulty generating small-sized CAs for large-scale instances. Since the size of a CA directly affects testing efficiency in practice, generating small-sized 3-wise CAs is of great importance. In this work, we propose , a novel 3-wise CCAG algorithm that effectively generates small-sized CAs for large-scale instances. Experimental results on extensive real-world highly configurable systems demonstrate that, compared with existing state-of-the-art algorithms, reduces the size of generated 3-wise CAs by 10.8%. Moreover, is efficient, requiring only half of the running time to generate smaller 3-wise CAs than existing algorithms.
@inproceedings{paperbot12197,
title = {DynCA: An Effective Algorithm for 3-Wise Combinatorial Interaction Testing in Highly Configurable Systems},
author = {Ruizhi Shi and Shuangyu Lyu and Renyu Yang and Wei Wu and Chanjuan Liu and Chunming Hu and Chuan Luo},
booktitle = {ASE},
year = {2026},
doi = {10.1145/3832783.3834349}
}