Humboldt-Universität zu Berlin - Faculty of Mathematics and Natural Sciences - Model Driven Software Development

SimuComp: Simulink Architecture Comprehension and Analysis

 

HEIBRIDS-LOGO

Funding: BMBF

Period: 2019 - 2021

Project partner: Prof. Dr. Andreas Vogelsang,
Model Engineering Solutions GmbH, Daimler AG

 

Classic embedded systems increasingly become independent and open systems which heavily interact with their cyber-physical environment and co-operate with other systems. Modeling is a promising approach to master the complexity of such multifunctional systems. For this purpose, large systems are usually decomposed into hierarchically organized subsystems which then implement a set of dedicated functions. On the functional level, Matlab-Simulink has been established as a de-facto standard for model-based development of embedded systems in many companies, for example, in the automotive industry. However, the different subsystems of a multifunctional system are developed largely independent of each other, e.g., in different departments of a company, and they are integrated into the overall system at a very late stage of the overall development process. A multifunctional system therefore is defined through a set of loosely coupled and independently developed Simulink models. The knowledge on their intended communication relations gets more and more lost during software evolution. Consequently, undesired yet unanticipated behavior often occurs when the different functions are integrated into an overall system. To date, an analysis of the respective interactions of single functions is only possible based on the generated source code. That is at odds with the paradigm of model-based software development and currently causes high integration costs. This project aims to consistently lift all integrative analytical tasks to the model level. Therefore we develop methods for automatic extraction of complex dependency relationships between Simulink models as well as analytical methods for early identification of architectural flaws and risks for a later integration of the functions. This leads to an increased architectural understanding of the set of sub-models and further on to a reduction of the integration costs. All developed concepts will be prototypically implemented and integrated into the renowned quality management tool M-XRAY of the company MES and evaluated in case studies, together with our associate partner Daimler AG.