A modeling process to understand complex system architectures
Balestrini Robinson, Santiago
MetadataShow full item record
Military analysis is becoming more reliant on constructive simulations for campaign modeling. Requirements for force-level capabilities, distributed command and control architectures, network centric operations, and increased levels of systems and operational integration are straining the analysis tools of choice. The models constructed are becoming more complex, both in terms of their composition and their behavior. They are complex in their composition because they are constituted from a large number of entities that interact nonlinearly through non-trivial networks and in their behavior because they display emergent characteristics. The modeling and simulation paradigm of choice for analyzing these systems of systems has been agent-based modeling and simulation. This construct is the most capable in terms of the characteristics of complex systems that it can capture, but it is the most demanding to construct, execute, verify and validate. This thesis is focused around two objectives. The first is to study the possibility of being able to compare two or more large-scale system architectures' capabilities without resorting to full-scale agent-based modeling and simulation. The second objective is to support the quantitative identification of the critical systems that compose the large-scale system architecture. The second objective will be crucial in the cases where a constructive simulation is the only option to capture the required behaviors of the complex system being studied. The enablers for this thesis are network modeling, graph theory, and in particular, spectral graph theory. The first hypothesis, stemmed from the first objective, states that if the capability of an architecture can be described as a series of functional cycles through the systems that compose them, then a simple network modeling construct can be employed to compare the different architectures' capabilities. The objective led to the second hypothesis, which states that a ranking based on the spectral characteristics of the network of functional interactions indicates the most critical systems. If modeling effort is focused on these systems, then the modeler can obtain the maximum fidelity model for the minimum effort.