Design space exploration for the development of embedded systems

The evolution of electronic devices has made a tremendous progress within the last 50 years, thus today’s world they can be found nearly everywhere, such as cell phones, camcorders, antiblock-brakes. The design of such complex system, that consists of hardware and software has to cope with several obstacles like for example high system complexity and increasing economical demands like shortened time-to-market. Those barriers get especially visible in the wireless domain. Here, design productivity lacks behind the possible computational complexity famously described with Moore’s law. The importance to cope efficently with these problems of system designing has been highlighted by the International Technology Roadmap for Semiconductors. This thesis examines one of the design tasks namely design space exploration. Since the description of systems raises constantly its level of abstraction which causes a higher ability for exploring design variants the automatic derivation of alternatives becomes a high reputation. Current approaches for design space exploration are based on manual exploration and hence suffer from a time consuming exploration and lead to sub optimal solutions. Even automated approaches are restricted due to the high system complexity and need to be enhanced. In this thesis a fast and efficient design space exploration approach is proposed. This is based on the characterisation of a system description, the estimation of design properties, and the automatic evaluation of design variants.