Effective Measurement and Actuation Paradigms for Diffusive Networks

Amirkhosro   Vosughi

The purpose of this thesis is to investigate the effectiveness of emerging control and monitoring paradigms for dynamical networks, from a graph-theoretic perspective. Two different directions of research are pursued: 1) analysis of target controllability and source estimation metrics for sparsely-actuated and measured network processes, 2) control design for diffusive network processes using stochastically-moving sensors and actuators. First, we study metrics for target controllability and source observability in dynamical networks, which are applicable for assessing security and enabling control in cyber-enabled networks like critical infrastructures. Specifically, the energy required to control a target node in a network from a remote input is characterized, and dually the fidelity with which a source state can be estimated from a remote measurement is studied. Several spectral and graphical results are presented, which allow discernment of the level of difficulty of target control and source estimation, and also permit comparison of the two metrics. Subsequently, scalable distributed algorithms are developed for computing the metrics. The second focus of this study is on analyzing the control of diffusive network processes using stochastically-moving sensors and/or actuators. Breakthroughs in Internet-of-Things technologies are making it possible to have platforms equipped with on-board sensor/actuator that can sense and manipulate network processes locally, even while they are involved in other missions. Here, we consider closed-loop control of diffusive or synchronization processes in networks using such platforms; both continuous-time and sampled-data formulations are considered. The stochastic movement of the platform is modeled by a Markov process, and Markovian Jump Linear System (MJLS) machinery is used to analyze moment stability. Small gain stability in a two-moment sense is verified. The settling time of the closed-loop system is also investigated in the continuous-time context. This framework is subsequently applied to design thermal regularization schemes for residential buildings which use occupant-location data. In this context, the framework is used to statistically characterize a cost metric that incorporates the deviation of temperature from the desired value and control effort. This framework is used to optimize the control parameters.

Effective Measurement and Actuation Paradigms for Diffusive Networks

Files and links (1)

Abstract

Metrics

Details