Tailoring complexity metrics for simulink models

Abstract

The size and complexity of Simulink models is constantly increasing, just as the systems which they represent. Therefore, it is beneficial to control them already at the design phase. In this paper we establish a set of complexity metrics for Simulink models to capture diverse aspects of complexity by proposing new and redefining existing metrics. To evaluate the applicability of our metrics, we compare them with the closed-source metric proposed by Mathworks. Moreover, through a case study from the automotive domain, we relate such metrics to quality attributes as determined by domain experts, and correlate them to known faults. Preliminary assessment suggests that complexity is closely related to analysability, understandability, and testability.