Sufficient Statistics for System of Systems Control (S4C)

There is an urgent need to develop and demonstrate innovative control, design and analysis techniques to characterize the stability and performance of a system of systems (SoS) as a function of sub-system dynamics, network structure and control/decision processes. This is particularly true in the missile defense scenario, which involves a coordinated defense against threats. However, there is no existing theory that can guarantee stability and performance of such networked control systems. To address this need, Intelligent Automation, Inc. and Stanford University, propose to develop a control theoretic framework for analysis and design of system of interacting control systems. This framework leverages on our teams prior work on developing sufficient statistics for optimal control of multi-agent systems. These statistics were developed for certain problems of linear systems and extended to particular classes of nonlinear systems as well. In this effort, we will further generalize the results to more general classes of systems and extend them to complex nonlinear and Multi-Input Multi-Output systems. We will study the computational aspects and trade-offs between performance and approximate solutions. This effort will lay the foundation for systematic control, design and analysis of complex real-world SoS with nonlinearity, uncertainties and communication delays. Approved for Public Release 16-MDA-8620 (1 April 16)