Current embedded systems are increasingly more complex and heterogeneous, but they are expected to be more safe, reliable and adaptive. In consideration of all these aspects, their design is always a great challenge. Developing these systems with conventional design approaches and programming methods turns out to be di cult. In this paper, we mainly present the informative background and the general idea of an ongoing yet young research project, including the modelbased design and an architecture-centric approach, to address previous challenges. Our idea adopts a formal-methods-based model integration approach, dedicated to architecture-centric virtual integration for embedded software systems, in an early design phase. We thus expect to improve and enhance Correct By Construction in the design. The considered formal methods consist of timing speci cation, design by contracts, and semantics interoperability for models to be integrated in the system. The application domains of our approach include automotive and avionic systems.
Current embedded systems are increasingly more complex and heterogeneous,
but they are expected to be more safe, reliable and adaptive [
In this paper, we mainly present the informative background and the general idea of an ongoing yet young research project, including the model-based design and an architecture-centric approach, to address previous challenges. Our idea adopts an formal-methods-based model integration approach, dedicated to architecture-centric virtual integration for embedded software systems, in an early design phase. By applying formal methods in an early design phase, we expect to improve and enhance correct by construction. The formal methods to be considered consist of timing speci cation, design by contracts, and semantics interoperability for models to be integrated in the system. The application domain of our approach include avionic and automotive systems. 2
High-level modeling has been widely adopted as a promising solution to
address the system complexity issue [
All the languages mentioned previously are considered as candidate languages
in high-level modeling for embedded systems. Multi-languages can be used in
the same design because of system modeling from di erent views, for example,
software, architecture, etc.; and di erent purposes, such as analysis, veri cation,
and evaluation. Furthermore, di erent languages may adopt di erent formalism,
e.g., state machines, data ow, communicating sequential processes, di erential
equations, as backstage support. So the rst challenge at the modeling language
level is how to harmonize multiple paradigm modeling [
An avionic co-modeling example. Co-modeling for the system-level
design has been explored in [
In Polychrony, the integration is performed at the polychronous MoC level[
Based on the previous exploration of design issues and the state of the art of
solutions in research, we nd an architecture-centric model-based integration
framework is necessary for the next-generation design of automotive software
systems. The framework is expected to provide the following advantages: reliable
model integration, fast and early-phase design validation, architecture
optimization enabling, easy access to current matured software development tools and
environment, etc. With this objective in mind, we rst propose a model-based
architecture-centric virtual integration approach, in the framework of
modelbased systems engineering [
With all the concerns in the embedded system design, timing is one of the most signi cant ones. In general, the timing issue becomes more explicit when architecture is considered and the system is integrated, due to the gap between software and architecture design. In our project, we consider high-level, formalized timing constraints to be de ned, observed and analyzed based on software architecture, speci ed in AADL. From this point of view, an architecture centric approach is adopted for the model integration in our project. Considering abstraction in the system design, we advocate the modeling of synchrony and time as software and hardware events, which are related to synchronization in an architecture speci cation. Compared to real time, synchronous logical time, applied on both software and architecture, provides an algebraic framework in which both event-driven and time-triggered execution policies can be speci ed.
In the framework of our project, we de ne the semantics and algebra with
regard to logical timing constraints and speci cation, and support the
submission of a timing-related annex to the SAE standard AADL[
Design by contract [
One of the main issues in the composition of models is semantics di erence
between heterogeneous models and di erent formalism. One of the feasible
solutions to this issue is to have a common model as the intermediate formal model,
and all other models are translated into the common model. An example can
be found in [
In this position paper, we have presented several important issues in current system design related to embedded systems, such as multi-paradigm modeling, integration framework, and formal semantics issues. A brief survey of corresponding research topics was also presented. We, hence, propose a model-based architecture-centric integration approach, considering timing speci cation, design by contract and semantics interoperability as main topics of research. Based on these research, a model-based integration framework is expected to be built, which is dedicated to model-based systems engineering for next-generation automotive systems.
The authors appreciate the valuable advices from Ryo Ito and Kazuhiro Kajio (Toyota Motor Corporation).