Prof. Dr. Holger Giese


Performance Assessment of UML-Models for Automotive Software


The development and maintenance of Automotive-Software-Systems is one of the most difficult challenges in the Automobile-Industry. By now 80% of today´s innovations are realised through software. This results in rapidly increasing complexity of such systems. The usage of UML is one approach to cope with this problem. The purpose of this approach is to encounter the higher design- and development-effort with early planning and documentation. Such an early definition of the architecture and the deployment is the basis for the reuse of components inside multiple vehicle types and the development and integration of complex software-systems where different stakeholders are involved. These stakeholders could be manufacturers, suppliers, sub-suppliers and so on.


The demand of resources and timing properties are crucial factors of software components, which are used inside an automobile. These factors are highly relevant for the reliability and efficiency of an automobile. The classical approach is to develop more isolated functionalities for which these properties could be estimated based on experiences. This approach has lead to very complex systems with over 10 million LOC, executed on more than 80 different distributed ECUs, which are connected over up to 4 different types of busses (CAN, LIN, MOST, FlexRay). In future projects, it is expected that more functionalities would be realized on more complex ECUs, so that the classical approach is not suitable without bigger problems. Much more methods and techniques are needed, which allow to do an early performance prediction and comparison of architectural alternatives.




Inside the project a methodical concept should be created, which allows to do a performance prediction during the development process using adequate analysis techniques and to evaluate architectural alternatives. Inside the developed methodology, it should be considered, which parts of the UML-Standard and what kind of UML-Profiles could be used to fulfill the above mentioned objectives. Beyond this, a prototypic solution for the analysis of architectural alternatives should be realized. For this purpose, a transformation from the UML-Models to a Task-Model has to be implemented. This Task-Model should be loaded inside the INCHRON toolset chronSim for an analysis of real-time-properties. The aim of such an analysis is to predict the performance of the late integrated system. This concerns especially properties like resource-demands and timing aspects.



The project is collaboration between INCHRON, a member of the Hasso-Plattner-Ventures in Potsdam, and the Hasso-Plattner-Institute. The students will cooperate with INCHRON to acquire the relevant requirements and to evaluate the results of the project.



Inchron Logo       HPI Logo