The Rutgers CAP Project Design Associate

The Design Associate is intended to be an interactive environment that supports decision-making, performance evaluation, design-record management and knowledge acquisition tasks. The system is speci cally intended to handle the design of complex, physical structures, such as ships and planes, among others. Two key di culties characterize this class of design problems: (1) Design goals depend on global properties of an artifact. (2) Evaluation of the performance of an artifact is computationally expensive. The Design Associate provides a set of tools for attacking each of these problems: Global constraints are attacked by methods of automatically abstracting and decomposing search spaces. Computational costs of evaluation are attacked by methods of intelligently selecting evaluation models at varying levels of approximation. Future work will extend the Design Associate by building a Model Associate for automatically generating new performance evaluation models. This extension is intended to support innovative design by diminishing the time and monetary costs of developing new models needed to evaluate radically new designs. Research on the Design Associate is expected to contribute to the eld of Computer-Aided Design by formalizing the generic task structure of complex, physical structure design. It is also expected to contribute to the eld of Arti cial Intelligence, by attacking problems such as search control, search space formulation, abstraction, decomposition, model selection and model formation. The research reported in this paper was supported by the Defense Advanced Research Projects Agency (DARPA) and the National Aeronautics and Space Administration (NASA), under DARPA-funded NASA grant No. NAG 2-645. This research has bene ted from numerous discussions with all members of the Rutgers CAP project, including Saul Amarel, Andrew Gelsey, Haym Hirsh, Chun Liew, Ringo Ling, Gerry Richter, Louis Steinberg, Chris Tong and Ke-Thia Yao.