| It is known that higher efficiency, reliability and fault tolerance call be achieved in decentralized systems, where multiple, largely autonomous, components are integrated. While continuous advancements in computer, communication, and control technologies have expanded the possibilities for creating such systems, there are still some fundamental issues in designing distributed algorithms that achieve a prescribed level of performance. The objective of our research was to address some of these issues with more emphasis on the control and communication aspects.; In the first part, we address the notion of control-oriented value of information. Considering Linear Quadratic Gaussian (LQG) systems, we show how the measurements can be evaluated based on their effect on the performance and flow information measures may be incorporated in a given performance index. Furthermore, we investigate how communicating various pieces of information among the local stations in a decentralized system may affect the overall performance and specifically the global stability of the decentralized system.; In the second part, we consider the Witsenhausen counter-example, which is a simple two-stage stochastic decentralized system with a non-classical information pattern. We present various reformulations of the problem and elaborate on the difficulties involved in designing optimal strategies. Assuming that the two stations communicate through a low noise channel, we show, through an asymptotic analysis, that the linear strategies still satisfy the necessary condition for optimality.; In the final part, we focus on power control problem, which can be regarded as a stochastic decentralized regulation problem in cellular wireless systems. We unify the two main approaches for information-feedback distributed power control design and obtain an insightful sufficient condition for network feasibility. Moreover, we use a robust control framework for global stability analysis of a power-controlled network. We then design a Kalman predictive distributed power control algorithm. We show, through extensive system-level simulations, that under the dynamics of user arrivals and departures and user mobility, significant improvement in performance can be achieved when our predictive power control algorithm is integrated with a distributed minimum interference dynamic channel assignment scheme. |
