A Framework for Decision Support in Systems Architecting

The objective of this thesis is to provide a method and tool to leverage computational resources to empower a systems architect to reason about architectural decisions more comprehensively and effectively compared to traditional approaches. This thesis provides a computational framework for decision support called the Architecture Decision Graph framework. It supports human decision-making by providing a methodology for generating and analyzing architectures as the result of a set of interrelated decisions. ADG’s explicit representation of an interconnected decision problem is a bipartite graph of decision variables, property variables, logical constraints, and property functions. The Architecture Decision Graph’s framework provide tools for reasoning about the structure of a decision problem, generating the set of feasible combinations of decisions, and simulating their outcome. The underlying computational engine used by ADG is the Object-Process Network (OPN) kernel. The contribution of this thesis to the field of systems architecting falls into three areas: First, the thesis contributes the ADG representation of an architectural candidate space as a set of interrelated decision variables. Second, the thesis contributes the ADG framework, which leverages the ADG representation of architecture to transform an architecting problem into a computational problem. Third, this thesis contributes decision space viewing tools, which present the potential impact of changes in the assignments of the decision variables to an architect. The ADG representation, analysis methodology, and tools are demonstrated with two applications. The first application is a retrospective study of the architectural decisions related to the development of the Apollo moon project of the 1960’s. The second application is a study of decisions in support of NASA’s lunar outpost architecting effort. The applications include discussions of the practical considerations related to the use of ADG as a decision representation method, the efficiency of the simulation algorithm, and a discussion of the architecting insights that can be drawn from the results. Thesis Supervisor: Edward F. Crawley Title: Ford Professor of Engineering