Beyond the Ordinary: Design of Embedded Real-time Control
2002 - 2007
The Bode-RC project objective is to define the approach on multi-disciplinary embedded systems design based on a backbone concept of integral embedded system architecture reasoning and a common multi-disciplinary hierarchical model.
Modern embedded systems, such as those for example encountered in high-speed digital printing equipment, rely on the input from different disciplines (like electronics, software engineering, IC design, sensor engineering, control theory) during their design. These disciplines use different formalisms, such as discrete events, classes, state-machines, differential equations and electromagnetic fields and the coupling between subsystems defined by two disciplines might be essential for the overall system performance.
The development practices commonly applied for complex embedded systems let the three major disciplines involved (mechanics, electronics and informatics) work on their part of the system in isolation, and even considering their part in turns: first the mechanics discipline defines its part, and then the electronics discipline takes over. Finally the software engineers develop the software for the system defined by mechanics and electronics. Obviously, a lot of hardware and software components are introduced and have to work together, but it never happens: unforeseen problems always appear. For such systems it is very difficult to analyze and predict the impact of different decisions and solutions that were made in isolation over the whole system and the correctness and performance can only be checked in the final stages of the design by extensive testing.
In order to be aware of and to prevent the problems that appear, the development methodology should include two phases: modeling and synthesizable descriptions. The primary purpose of modeling is analysis. Models should concentrate on one or more aspects of the system, while abstracting from the others. On the other hand, description means synthesis. All parts of the system that should be synthesized are present in the system description.
The advantage of a top-down or model-driven approach compared with a bottom-up or code-driven approach is the consistency of the implementation with the model, the execution according with the semantics of the model, prediction about timing behavior which is critical for real-time embedded systems.
The focus of this PhD research is to get from a model of a real-time embedded software system (which takes into account also the environment it interacts with) to its implementation on a heterogeneous hardware platform, in a predictable way. In an iterative fashion the model is refined until it reaches the stage of a synthesizable description which can automatically be transformed into an implementation. The modeling method allows different hardware architectures and mappings to be explored in the early phases of the design trajectory. At each stage of the model refinement, evaluations has to be made in order to find possible bottlenecks that otherwise would not allow a correct implementation of the system.
People involved from TU/e – ICS-ES