Performance-information analysis and distributed feedback stabilization in large-scale interconnected systems

Abstract

Large-scale interconnected systems are characterized as large and complex systems divided into several smaller autonomous systems that have certain autonomy in local optimization and decision-making. As an example, a class of interconnected linear stochastic systems, where no constituent systems need to have global information and distributed decision making enables autonomous systems to dynamically reconfigure risk-value aware performance indices for uncertain environmental conditions, is considered in the subject research. Among the many challenges in distributed and intelligent control of interconnected autonomous systems is performance uncertainty analysis and decentralized feedback stabilization. The theme of the proposed research is the interplay between performance-information dynamics and decentralized feedback stabilization, both providing the foundations for distributed and autonomous decision making. First, recent work by the author in which performance information availability was used to assess limits of achievable performance will be extended to give insight into how different aggregation structures and probabilistic knowledge of random decision processes between networks of autonomous systems are exploited to derive a distributed computation of complete distributions of performance for interconnected autonomous systems. Second, the resulting information statistics on performance of interconnected autonomous systems will be leveraged in the design of decentralized output-feedback stabilization, thus enabling distributed autonomous systems to operate resiliently in uncertain environments with performance guarantees that are now more robust than the traditional performance average.