A domain independent method to assess system of system meta-architectures using domain specific fuzzy information

This research proposes a domain independent method to build and assess systems of systems (SoS) architecture models. A simplified binary, meta-architecture containing each component system's participation and a first order, system-to-system interface is proposed. The method describes how to elicit desired SoS attributes from stakeholders. Measures of the attributes depend on systems' participation, characteristics and interfaces, that is, on the SoS architecture. The goal is to model a realizable SoS configuration, optimized over multiple attributes. Key attribute measures are combined in a fuzzy inference system to assess an overall fitness measure for any SoS within the meta-architecture. A genetic algorithm is used to find 'good' SoS architectures with a fitness that depends on the participation framework. This research illustrates a method to define architecture sensitive attributes and build the fuzzy assessor. These are two segments of the Missouri S&T developed, nine part Flexible Intelligent Learning Architectures for SoS (FILA-SoS) research approach to architecting SoS. A desirable SoS architecture found this way may be handed off to an agent-based model to examine the impact of various negotiation behaviors or policies on realization of the SoS. The final configuration may evolve over several development epochs as desired in the wave model.;The method is demonstrated on SoS in several domains to illustrate its broad generality. Two intelligence, surveillance and reconnaissance (ISR) SoS, a search and rescue (SAR) SoS, two versions of the MITRE Toy problem, and a validation using an actual SoS for a large training program are analyzed. The method provides researchers and designers with a novel way to think about the effects of imprecise stakeholder desires, sensitivity to inputs, and acquisition policies on SoS architecting.