Research
EMBEDDED SYSTEMS
Embedded systems are programmable, electronic (often in combination with mechanical) systems that control and determine the functioning of devices (machines, appliances, instruments, constructions). Embedded systems are multidisciplinary by nature. Their design involves the disciplines software engineering, electronic engineering, and control theory, and -- depending on the application domain -- other disciplines as well.
The functionality of embedded systems is usually expressed by means of complex software structures that run on distributed hardware platforms. Embedded systems have to operate under hard constraints that pertain to real-time behavior, low-cost, low-energy, short time-to-market, high reliability, ease-of-use, etc. Designing systems that meet such constraints requires a global system approach that involves application domain expertise.
The ESI focuses on industrial embedded systems, systems that occupy volumes of cubic decimeters to cubic meters. This is the application/market domain of wafer steppers, printing systems, medical imaging equipment, televisions, automobiles, electronic microscopes, manufacturing machines, etc.
THE MULTIDISCIPLINARY SYSTEMS DOMAIN
This drawing illustrates the multidisciplinarity of embedded systems. At the bottom are the well-understood monodisciplinary technologies: mechanics, electronics, and software. From these building blocks systems are created that satisfy system-wide constraints pertaining to system qualities such as cost, performance, and reliability. Many, but not all, of these qualities can be expressed in well-understood units (in euros, in m/s, in bits/s, in mean-time-between-failure, for example). The part that is not well-understood is the question how to connect bottom and top, i.e. how to design systems expressed in the bottom technologies that meet the system constraints.
Currently at many institutes in the world research is carried out how to combine bottom disciplines, for example how to combine discrete and continuous methods into hybrid methods. The next step upward (in the picture the ovals) addresses the question how to come to design methods that are multi-disciplinary and single-quality, for example trade-off analysis between software and mechanics aimed at weight reduction. Another step upward in the systems hierarchy (in the picture the lower clouds) is about design methods that are both multi-disciplinary and multi-quality, for example multidisciplinary trade-off analysis that addresses both speed and energy.
The ESI focuses its research on multidisciplinary systems design. The research yields design methods of the ‘oval’ and ‘lower clouds’ types. The ESI relies on its network of ‘partner groups’ for the mono-disciplinary contributions in the ESI research projects.
RESEARCH FOCUS
The ESI works on design methods for embedded systems that
- exploit the potential of harmonious compositions of software, electronics, and other technologies,
- address multiple system qualities, and
- generalize over (at least) the domain of industrial embedded systems.
Many system qualities are important for embedded systems, such as reliability, performance, availability, cost, robustness, safety, maintainability, security, responsiveness, adaptivity, evolvability, survivability, nomadicity, manageability, and scalability. Although none of these qualities can be excluded, the ESI chooses a few of them on which it focuses and in which it acquires expert staff.
The system qualities the ESI focuses on are
- Performance (the system exhibits correct timing behavior and uses its resources efficiently)
- Reliability (the system keeps operating correctly while being used)
- Evolvability (the system is optimized for different applications over long operational lives without compromising safety and reliability)
These qualities are always studied in relation to other qualities.
There are developments that make it increasingly difficult to guarantee that systems meet these qualities at the required levels. An important development is that embedded systems become more open. They open up in two ways: they become part of a priori unknown ‘networked, distributed’ communities, and they increasingly involve parts that are (dynamically) supplied by third parties. Other developments jeopardizing these system qualities are the seemingly untameable growth of (embedded) software code, a phenomenon known as software bloating, the shortening design and life cycles of embedded systems, and the customization of embedded systems.
The drawing below shows how the current ESI projects address these system qualities. The horizontal lines are system qualities and other research targets; the vertical lines are the projects, where PMS stands for the research project with Philips Medical Systems as Carrying Industrial Partner, which started in the fall of 2005 as Darwin.
