A Robust Portfolio Optimization Approach to System of System Architectures

Abstract

The realization of military capability as a System of Systems (SoS), presents significant development challenges across technical, operational and programmatic dimensions. In particular, tools for deciding how to form and evolve SoS which consider performance and risk are lacking. This research leverages tools from financial engineering and operations research perspectives in portfolio optimization to assist decision making in this setting. Our approach facilitates evolutions of SoS architecture through a framework that supports architecture selection at a given decision-epoch of the evolutionary process. The approach models hierarchies of interdependent systems as generic nodes on a network that, subject to connectivity and compatibility constraints, work cohesively to fulfill overarching capability objectives. A robust portfolio algorithm is employed to address inherent real world issues of data uncertainty, inter-nodal performance and developmental risk. A naval warfare scenario illustrates application of the method to find "portfoliosˮ of systems from a candidate list of available systems. Results show how the framework effectively reduces the combinatorial complexity of tradespace exploration (e.g., connectivity rules, feasibility of solutions, optimality of solutions) by allowing the optimization problem to handle the mathematically intensive aspects of the decision-making process. As a result, human decision-makers can focus on choosing the appropriate weights for risk aversion in making final decisions. C⃝ 2015 Wiley Periodicals, Inc. Syst Eng 18: 269-283, 2015

Architecting a portfolio of systems

“…A “Collaborative System” or “System‐of‐Systems.” This is a collection of systems following the base definitions in Maier and Maier and elaborated in the literature since, see Sage and Cuppan and Haimes among many others. Portfolio concepts have been identified in this context previously, but this work applies them outside of the specific system‐of‐systems conditions. The three key prongs of this definition are that the systems in the collection: (a)Interact with each other to produce results none of them can achieve alone, (b)If a member is disconnected from the collective, it continues to fulfill useful purposes on its own (operational independence of the elements), (c)While interconnected, the constituent systems continue to be managed for their own purposes independently of the collective, at least in part (managerial independence of the elements).…”

Architecting federations of systems: A framework for capturing synergy

“…There is a rich body of literature about SoS, concerned with definitions, developing specific approaches, and addressing key issues such as dependency analysis of SoS, distributed resource management in SoS, decentralized optimization in SoS, and incentive‐based negotiation models . FoS are a subclass of SoS, and as per the current definitions of SoS, fall under the category of virtual SoS.…”

“…Hence, architects designing a system to be federated are trading design options that depend both on network externality effects 8 and specifications of systems outside their control. This means that There is a rich body of literature about SoS, concerned with definitions, 9 developing specific approaches, 10,11 and addressing key issues such as dependency analysis of SoS, 12 distributed resource management in SoS, 13 decentralized optimization in SoS, 14 and incentivebased negotiation models. 15 FoS are a subclass of SoS, and as per the Several authors have discussed the nature of an SoS, in an attempt of distinguishing them from conventional systems.…”

Model‐based architecture and programmatic optimization for satellite system‐of‐systems architectures

“…Individual systems were incorporated into an overall architecture model, and quantified relationships were used to exercise data analysis techniques to perform SoS architecture optimization for the desired parameters (such as portfolio capability, schedule, and costs). While examples of data analysis techniques integrated with architecture modeling have been demonstrated, emphasis has generally been on the design, integration, and test phases of a development program, and more work is needed to demonstrate the utility and successfully integrate optimization into a systems architecting framework for the predesign phase where the greatest impact on cost occurs.…”

“…Uncertainty analysis has also been explored within architecture optimization. Davendralingam and DeLaurentis showed how to apply mean‐variance optimization techniques originally developed for the financial sector by Markowitz to SoS architecture design to optimize performance against development time risk. If those initial risk values can be quantified with some known uncertainty, then methods exist to rigorously propagate that uncertainty through the architecture model to determine uncertainty bounds and assess the fidelity for the output of the optimization Xu et al demonstrated a method to utilize unscented transformations as a less computationally demanding alternative to Monte Carlo to capture the effects of uncertainty in an optimization.…”