Abstract
System partitioning is essential to the design of com- plex systems such as automobiles.
Complexity can be because of compounded phenomena or because of intricate behavior. This
paper focuses on the com- pounding effect of integrating the different subsys- tems that result
from the partitioning. In particu- lar, it concentrates on subsystems, or features, that have an
embedded computation part to them. Three types of feature integration are classified: (i) shared
resources, (ii) communicating features, and (iii) inter- acting control. The integration is addressed
from a Model-Based Design perspective. It is discussed how this allows managing the complexity of
integrating computations because it is operative at a higher level of abstraction than even highlevel
programming lan- guages.
Conclusions
The application of embedded computing power in the automotive industry is increasing at a
steady pace. The main benefit of embedded computations are the relatively low cost for the
design flexibility that it provides.
With the flexibility comes a complexity, though,
that can be difficult to negotiate. This complexity can be because of size, but also because of
compli- catedness. In this paper, it is concentrated on the former, and it is discussed how
partitioning aids in managing the complexity because of the compounded phenomena.
With partitioning comes the inevitable integra- tion, and three classes of integration are identified:
(i) shared resources, (ii) communicating features, and (iii) interacting control. It is argued that highlevel
programming languages such as C and C++ do not suffice in addressing those three
integration issues.
Instead, Model-Based Design is introduced as a means to support the subsystem integration at a
much higher semantic level, that is closer to the domain specific notions. Some of the benefits of
Model-Based Design for subsystem integration are presented. The models that are used are better
suited to address intricate matters such as interface compat- ibiliy, not just from a syntactic, but
also from a static and dynamic semantics perspective.