| Chair | Pieter J.
Mosterman Real-time and Simulation Technologies The MathWorks, Inc. USA |
Model-Based Design involves engineers from many different disciplines applying models for purposes ranging from documentation, numerical simulation, code generation to hardware programming. Separate tools are typically employed for these respective purposes because dedicated to solving the task at hand. Furthermore, different tools may have different analysis facilities and support modeling formalisms with widely different semantics. To minimize the cost because of model exchange and integration, automated model exchange and execution of models designed with different tools is desired.
Paper 1 [Klaus Mueller-Glaser] Up to 70 electronic control units (ECUs) serve for safety and comfort functions in a car. Communicating over different bus systems most ECUs perform close loop control functions and reactive functions fulfilling hard real time constraints. Some ECUs controlling on board entertainment/ office systems are software intensive, incorporating millions of lines of code. The design of these distributed and networked control units is very complex, the development process is a concurrent engineering process and is distributed between the automotive manufacturer and several suppliers, this requires a strictly controlled design methodology and the intensive use of computer aided engineering tools. The CASE-tool integration platform ``GeneralStore'' and the ``E/E-Concept Tool'' for design space exploration supports the design of automotive ECUs, however, GeneralStore is also used for the design of industrial automation systems and biomedical systems.
Paper 2 [Jan Broenink] The work presented here is on setting up methodological support, including (prototype) tools, for the design of distributed hard real-time embedded control software for mechatronic products. The use of parallel hardware(CPUs,FPGAs) and parallel software is investigated, to exploit the inherent parallel nature of embedded systems and their control. Two core models of computation are used to describe the behavior of the total mechatronic system (plant, control, software and I/O): discrete event system (DES) and continuous time system (CTS). These models of computation are coupled via co-simulation, to be able to do consistency checking at the boundaries. This allows for integration of discipline-specific parts on the model level (during design phases) instead of on the code level (during realization and test phases). Crossview design-change influences get specific attention, to allow for relaxation of the tension between several dependability issues (like reliability and robustness), while keeping design time (and thus design costs) under control. Furthermore, the design work can be done as a stepwise refinement process. This yields a shorter design time, and a better quality product. The method is illustrated with a case using the tools being prototyped.
Paper 3 [Horst Salzwedel] Transition to model based design of systems at electronic system level (ESL) has greatly reduced the complexity by raising the level of abstraction. It resulted in improvements in quality of design, reduction of development time, and reduction of number of iterations in design. A new emerging problem is the widening of the gap between design at ESL and implementation. In this paper a method is presented to overcome this gap. For the design of an inexpensive high precision positioning system based on GPS, additional sensors, and differential signals, functional requirements are mapped into a MLDesigner model containing functional model, environmental model, and use cases. Annotations on the model are automatically translated in an architectural model with channels, resources, events, and memory from a developed standardized library. After design iterations on architecture from a selected sub-model, an implementation framework for hardware and software code is generated and I/O specific parameters are added. The code can be used on FPGAs and embedded processors.
Paper
4 [Gabriela Nicolescu] Continuous and discrete components may be
integrated in diverse embedded systems ranging across defense,
medical, communication, and automotive applications. The global
validation of these systems requires new validation techniques, the
main challenge being the definition of global simulation models able
to accommodate the different concepts specific to continuous and
discrete models. This paper presents a generic methodology for the
efficient design of continuous/discrete-events co-simulation tools.
Before the implementation stage, the methodology proposes several
steps enabling the gradual formal definition of the simulation
interfaces functionality and their internal architecture.