Multi-Stakeholder Dynamic Optimization Framework for System-of-Systems Development and Evolution

Zhemei Fang, Purdue University

Abstract

Architecture design for an “acknowledged” System-of-Systems (SoS), under performance uncertainty and constrained resources, remains a difficult problem. Composing an SoS via a proper mix of systems under the special control structure of an “acknowledged” SoS requires efficient distribution of the limited resources. However, due to the special traits of SoS, achieving an efficient distribution of the resources is not a trivial challenge. Currently, the major causes that lead to inefficient resource management for an “acknowledged” SoS include: 1) no central SoS managers with absolute authority to address conflict; 2) difficult balance between current and future decisions; 3) various uncertainties during development and operations (e.g., technology maturation, policy stability); 4) diverse sources of the resources; 5) high complexity in efficient formulation and computation due to the previous four factors. Although it is beyond the scope of this dissertation to simultaneously address all the five items, the thesis will focus on the first, second, and fifth points, and partially cover the third point. In a word, the dissertation aims to develop a generic framework for “acknowledged” SoS that leads to appropriate mathematical formulation and a solution approach that generates a near-optimal set of multi-stage architectural decisions with limited collaboration between conflicted and independent stakeholders. This dissertation proposes a multi-stakeholder dynamic optimization (MUSTDO) method, which integrates approximate dynamic programming and transfer contract coordination mechanism. The method solves a multi-stage architecture selection problem with an embedded formal, but simple, transfer contract coordination mechanism to address resource conflict. Once the values of transfer contract are calculated appropriately, even though the SoS participants make independent decisions, the aggregate solutions are close to the solutions from a hypothetical ideal centralized case where the top-level SoS managers have full authority. In addition, the thesis builds the bridge between a given SoS problem and the mathematical interpretations of the MUSTDO method using a three-phase approach for real world applications. The method is applied to two case studies: one in the defense realm and one in the commercial realm. The first application uses a naval warfare scenario to demonstrate that the aggregated capabilities in the decentralized case using MUSTDO method are close to the aggregated capabilities in a hypothetical centralized case. This evidence demonstrates that the MUSTDO method can help approach the SoS-level optimality with limited funding resource even if the participants make independent decisions. The solution also provides suggestions to the participants about the sequence of architecting decisions and the amount of transfer contract to be sent out to maximize individual capability over time. The suggested decisions incorporate the potential capability increase in the future, which differentiates itself from allocating all the resources to the current development. The quantified numbers of transfer contract in this case study are equivalent capabilities that are relevant to equipment loan or technology transfer. The second case study applies the MUSTDO-based framework to address a multi-airline fleet allocation problem with emissions allowances constraint provided by the regulators. Two representative airlines including the low-cost airline and the legacy airline aim to maximize individual profit by allocating six type of aircraft to a given ten-route network under the emissions constraint. Both the deterministic and stochastic experiments verify the effectiveness of the MUSTDO method by comparing the profit in the decentralized case and profit in a utopian centralized case. Meanwhile, sensitivity studies demonstrate that higher minimum demand requirement and lower discount factor can further improve the efficiency of emissions allowances utilization in MUSTDO method. Comparing to an alternate grandfathering approach, the MUSTDO method can guarantee a high-level efficiency of resource allocation by avoiding failed allocation decisions due to inaccurate information for the regulators. In summary, the framework aids the SoS managers and participants in the selection of the best architecture over a period of time with limited resources; the framework helps the decision makers to understand how they can affect each other and cooperate to achieve a more efficient solution without sharing full information. The major contribution of this dissertation includes: 1) provide a method to address multi-stage SoS composition decisions over time with resource constraint; 2) provide a method to manage resource conflict for stakeholders in an “acknowledged” system-of-systems; 2) provide a new perspective of long-term interactions between stakeholders in an SoS; 3) provide procedural framework to implement the MUSTDO method; 4) provide comparison of different applications of the MUSTDO framework in distinct fields.

Degree

Ph.D.

Advisors

DeLaurentis, Purdue University.

Subject Area

Aerospace engineering|Systems science

COinS