.png){kind=icon}
.png){kind=icon}
Return
Evolutionary many-objective satisfiability solver for configuring software product lines
.png){kind=icon}
PRE
AI
DOI：10.1007/s10489-021-02918-x.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Software product line configuration (SPLC) is a typical many-objective optimization problem. Its purpose is to optimize multiple objectives simultaneously in a highly constrained space for searching. Recently, the concept of combining evolutionary algorithms with satisfiability (SAT) solvers has drawn increasing attention to this problem. However, most computational approaches often suffer from model instability and poor performance. Therefore, to address these limitations with higher efficiency, we propose a new algorithm, MySPLC, for configuring software product lines, which involves three main strategies. SMS is a new operator with higher performance than other traditional genetic operators. MSPS selects multiple invalid solutions in a population iteration to process. This balances the time required for environment selection and invalid solution processing, making each environment selection more efficient and meaningful. MCVFS makes the population evolve in a more diversified direction. We performed experiments on 10 feature models, and the feature numbers of these feature models ranged from 544 to 31012. The results show that our algorithm can find more diverse solutions than state-of-the-art methods. In addition, we study the related parameters,genetic operators, convergence and scalability of the algorithm. These studies further expand the application scenarios of the algorithm.
Keywords:
Software product line configuration
Many-objective optimization
Evolutionary algorithm
Satisfiability solvers
Feature models