SUMMARY

When engineers design a system, they have a large number of system alternatives to consider. They systematically prune this space of alternatives into a system specification. This specification has two components: the architecture specification and the specification for the components in this architecture. We call the process designers follow to arrive at the system specification architecture exploration. Architecture exploration is important in many product domains, especially those with a large number of plausible alternatives none of which are clearly dominant. With current technology, broad exploration is prohibitively expensive or time consuming, and instead, designers focus their efforts on appropriately sizing a small number of candidate architectures. To alleviate this problem, others have investigated the use of computational synthesis methods, but these methods are inefficient because they fail to effectively capture and apply the designer’s knowledge about the solution domain. One major shortfall is that architecture synthesis is considered separately from component sizing, and the analysis knowledge needed to evaluate the architecture is not captured or used within the same framework as the synthesis knowledge. In this proposal, we explore how designer knowledge should captured and then formulated into an architecture exploration problem We then develop and evaluate two approaches that incorporate analysis knowledge into the architecture synthesis process. The first approach is an iterative approach where architectures are first synthesized and then sized in sequential steps. The second approach is an integrated approach where the system behavior is considered during the architecture synthesis step so that an alternative is synthesized and sized simultaneously. The two approaches will be compared based on expressivity, quality of solutions generated, and the scalability.