The following technical report is available from
http://aib.informatik.rwth-aachen.de:
Ansatz zur variantenreichen und modellbasierten Entwicklung von
eingebetteten Systemen unter Berücksichtigung regelungs- und
softwaretechnischer Anforderungen
Andreas Polzer
AIB 2015-13
Nowadays, model-based development is a common method to deal with the
complexity and wide rages of functionality in embedded systems.
Especially in automotive domain MATLAB/Simulink is often used to develop
controller software for embedded systems. However the usually used tools
are originally not designed to model a product-line of embedded systems
applications. Therefore there is a high complexity of the designed
models. Additionally different tools are used for different design
artefacts like requirements and implementation model. These tools are
not connected to each other's which causes inconsistencies.
Different domains like electrical engineers, control engineers and
computer scientist are involved in designing embedded systems. The
different domains use different domain specific languages and have
different requirements.
This thesis introduces an approach considering requirements of different
domains designing embedded systems. The main focus is the further
model-based development of existing embedded applications, in particular
the cooperation of control engineering and computer science. Both
domains have partly different requirements. E. g. in computer science it
is important to reuse software. However in control engineering there is
a need for great flexibility in terms of external interface (actuators
and sensors) and the according software parts since the actuators and
sensor have critical influence on the performance of the controller.
Hence both domains have conflicting requirements.
In the second part of this thesis an approach on software product-line
is adopted to the model-based development process. Model-based
transformations and analysis are used to support the development. The
transformations use all design artefacts to achieve consistent models
with respect derived variants. The result of the analysis are views.
This views are implemented using standardized transformations provided
by a framework. Therefore new analysis views can be created combining
and adopting these transformations.
