Hybrid Systems: Motivation
and Challenges
Technology advances allow
designing systems whose complexity was simply unthinkable a few years ago.
Design time has become the bottleneck for bringing new products to market.
Traditional design paradigms are no longer effective. The most challenging
designs are in the area of safety critical systems such as the ones used to
control the behavior of transportation systems (e.g., airplanes, cars, and
trains) or industrial plants. The difficulties reside in accommodating
constraints on functionality and implementation.
Functionality has to guarantee
correct behavior under diverse states of the environment and potential failures;
implementation has to meet cost, size, and power consumption constraints. When
designing embedded systems of this kind, it is essential to take into
consideration all effects including the interaction between environment (plant
to be controlled) and design (digital controller). This calls for methods that
can deal with heterogeneous components that exhibit a variety of different
behaviors. For example, digital controllers can be represented mathematically as
discrete event systems, while plants are mostly represented by continuous time
systems whose behavior is captured by partial or ordinary differential
equations. In addition, the complexity of the plants is such that representing
them at the detailed level is often impractical or even impossible. To cope with
this complexity, abstraction is a very powerful method. Abstraction consists in
eliminating details that do not affect the behavior of the system that we may be
interested with. In both cases, different mathematical representations have to
be mixed to analyze the overall behavior of the controlled system.
Many are the difficulties in
mixing different mathematical domains.
In primis,
the very meaning of interaction may be challenged. In fact, when
heterogeneous systems are interfaced, interface
variables are defined in different
mathematical domains that may be incompatible. This
aspect makes
verification and synthesis impossible, unless a careful
analysis of the interaction
semantics is carried out.
In general, pragmatic solutions
precede rigorous approaches to the solution of engineering problems. This case
is no exception. Academic institutions and private software companies (e.g.
Mathworks) started developing computational tools for the simulation, analysis,
and implementation of control systems deploying first common sense reasoning and
then trying a formalization of the basic principles. These approaches focused on
a particular class of heterogeneous systems: systems featuring the combination
of discrete-event and continuous-time subsystem. Recently, these systems have
been the subject of intense research by the academic community because of the
interesting theoretical problems arising from analysis and design of these
systems as well as of the relevance in practical applications. These systems are
called hybrid systems.
Simulink, Stateflow and Matlab together provide excellent modeling and
simulation capability for the design capture and the functional verification via
simulation of embedded systems; however, often there is a need to subject the
models (developed in Simulink) to a more rigorous and domain-specific analysis
as well as to refine this high-level description into an implementation. In
addition, we expect that no single design framework will be capable of
encompassing all the needs of system designers. Hence, exporting and importing
design representations will be a necessity even for future powerful tools.
Remodeling the system in another tool’s modeling language while possible
requires substantial manual effort. Additionally, maintaining consistency
between models is error-prone and difficult in the absence of tool support. The
popularity of Matlab, Simulink, and Stateflow implies that significant efforts
have already been invested in creating a large model-base in Simulink/Stateflow.
It is desirable that application developers take advantage of this effort
without foregoing the capabilities of their own analysis and synthesis tools.
Owing to these factors a strong need has been expressed for automated semantic
translators that can interface with and translate the Simulink/Stateflow models
into the models of different analysis and synthesis tools.
On a more fundamental level, a unified approach to hybrid systems modeling is
needed to enable the use of joint techniques and a formal comparison between
different approaches and solutions.
Suggesting the guidelines to use for the development of a common interchange
language for hybrid systems modeling is our main objective.
|
Aims |
