Distributed Control of Large-Scale Networks: From Theory to Applications

Abstract

Networked systems analysis has been on the forefront of interdisciplinary research over the past few years with applications spanning from a wide variety of research literature, ranging across control systems, learning, power grids, social network, and economics. Control of large-scale networked systems often necessitates the availability of complex models for the interactions amongst the agents. This dissertation tackles some of the aspects of control and learning in such complex interconnected systems from both theoretic and applied viewpoints. We consider networks as ``systems'' and study the advantage they bring into control, estimation, and optimization of various applications such as antagonistic networks, network games, and layered networks. Next, we generalize our viewpoint towards networks as a ``system-of-systems'' where each system contains a subsystem and consider the distributed control problem on such hierarchy. This part considers methods for synthesis of linear quadratic controllers for large-scale systems that consist of interactive agents. Lastly, we put forward ideas on online iterative synthesis when the control horizon is limited or data collection is expensive. This part suggests potential remedy to some shortcomings that are incurred due to potential faulty assumptions in the preceding chapters of the thesis. An underlying theme to most parts of the manuscript is how well a network's structure facilitates the machinery of control and optimization of systems that are considered in this category. The functionality of methods in each part is demonstrated via illustrative simulations.