Combining evolutionary algorithms with constraint solving for configuration optimization

In Search based Software Engineering, well-known evolutionary algorithms are utilized to find the optimal solutions and address the configuration optimization problem for software product lines and trade off multiple often competing objectives. Previous work by Henard et al. showed the weakness of the constraint expressiveness and the optimality and speed. In this work, we propose a multi-objective evolutionary algorithm, which significantly improves the expressiveness from Boolean constraints to quantifier-free first-order constraints, particularly without sacrificing much performance. Furthermore, we propose a parallel portfolio approach. Empirical results demonstrate that this approach presents the performance superiority compared with the state-off-the-art and improves optimality as far as possible within a limited time budget. Finally, we present an overview of challenges in future.