Configuration-Based Exploratory Design

In this paper, we present a method for adapting the use of configuration systems to design exploration. Typically, the design of complex systems is realized as the integration of several constituent components. To guide the development of each of these components, designers would like to select the most promising family of solutions early in the design process. However, since the best family of solutions is the one containing the best instance, selecting the best family would require analyzing every individual instance in each family. This is a costly exercise that we are trying to avoid in the first place by focusing only on the most promising family of solutions. To overcome this catch-22, we introduce the use of automated configuration systems to conduct design exploration at a reduced cost. Such a configuration system can explore families of solutions automatically, not necessarily in full detail, but in sufficient detail to exclude the worst solution alternatives. We have adapted the traditional Propose and Revise approach to configuration into a hierarchical, modular framework that avoids the large initial investment involved in creating a traditional configuration system. We conclude this paper by illustrating our approach with an initial implementation of our framework called LEGION.