Implementing Search-Based Software Engineering Approaches with Nautilus
Resumo
Search-Based Software Engineering (SBSE) approaches adopt search-based techniques to solve Software Engineering (SE) optimization problems. Among these techniques, evolutionary algorithms are the most popular and successfully used, such as multi-objective evolutionary algorithms. However, some challenges still need to be addressed. Firstly, SE problems are complex and commonly impacted by many conflicting factors. In this context, the use of many-objective algorithms is necessary. Secondly, the users very often do not recognise the found solutions as feasible because these solutions are usually not generated considering the users’ needs and preferences. Thus, to deal properly with this situation, preference-based algorithms should be applied. Moreover, there are some practical issues regarding the choice of operators, evaluation of algorithms and visualization of solutions. Existing frameworks do not provide support to address these challenges. To overcome these limitations, we present Nautilus, an open-source Java web-platform tool that works with plugins to ease the addition of new problem instances, implementation of search operators and different multi and many-objective optimization algorithms, guided (or not) by human participation. This paper describes Nautilus-NRP, an extension implemented to address the Next Release Problem (NRP). NRP refers to the selection of requirements to be implemented in the next release of a software and is used to illustrate Nautilus’ main functionalities and how it can be extended to solve a SE problem. Link for the video: https://youtu.be/2dbwslTrvhg.
