Repository mining for changes in Simulink and Stateflow models

Model-Based Development (MBD) is widely used for embedded controls development, with MATLAB/Simulink/Stateflow being one of the most used development environments in the automotive industry. Simulink/Stateflow models are the primary design artifacts in automotive controls MBD development, and these models must be maintained over their lifetime. We achieve this in traditional software designs through the use of information hiding. It is thus beneficial to develop these models so that they facilitate likely changes that do not adversely impact the quality of the design. In order to do so, the types of frequently performed changes must be understood and appropriate language mechanisms must be available to support these changes. While some work has been done to analyze changes in Simulink/Stateflow models, a much deeper understanding is needed. We leveraged an extraordinary opportunity of having access to a comprehensive industrial software repository and its associated version control system to gain insight into likely changes for Simulink/Stateflow in automotive controls development. This analysis provides accurate feedback on actual changes made over many years to Simulink/Stateflow models, and classifies changes to suggest how particular model changes can impact system evolution.