Towards a Taxonomy for Simulink Model Mutations

A relatively new and important branch of Mutation Analysis involves model mutations. In our attempts to realize model-clone detector testing, we found that there was little mutation research on Simulink, which is a fairly prevalent modeling language, especially in embedded domains. Because Simulink model mutations are the crux of our model-clone detector testing framework, we want to ensure that we are selecting the appropriate mutations. In this paper, we propose a taxonomy of Simulink model mutations, which is based on our experiences thus far with Simulink, that aims to inject model clones of various types and is fairly representative of realistic Simulink edit operations. We organize the mutations by categories based on the types of model clones they will inject, and further break them down into mutation classes. For each class, we define the characteristics of mutation operators belonging to that class and demonstrate an example operator. Lastly, in an attempt to validate our taxonomy, we perform a case study on multiple versions of three Simulink projects, including an industrial project, to ascertain if the actual subsystem edit operations observed across versions can be classified using our taxonomy and present any interesting cases. While we developed the taxonomy with the specific goal of facilitating and guiding the injection of mutants for model clones, we believe it is fairly general and a solid foundation for future Simulink model mutation work.