Dissertation
Controlling Complex Network Dynamics using Sparse Measurements and Actuation: Graph-Theoretic Perspective
Doctor of Philosophy (PhD), Washington State University
01/2014
Handle:
https://hdl.handle.net/2376/112215
Abstract
This thesis studies the control and estimation of hardwired network processes with sparse actuation and sensing from a topological perspective. This topological characterization is achieved in the context of several evocative linear models for network dynamics. For these models, formal relationships between the network's graph topology and properties of the network's actuated/measured dynamics have been developed.
One thrust of the research has been on estimation and identification of linear network synchronization processes. We provide graph-theoretic characterizations of 1) the mode estimation performance, as indicated by the Cramer-Rao Bound; and 2) structural-change detection, which is phrased as a Maximum A Posteriori probability detection problem. Our analysis indicates linkages between the network topology, the sensor locations, and the estimation/detection performance, which provide insight into optimal sensor placement. Further, we present a case study on fast fault location in power transmission networks using synchrophasor measurements.
A second thrust has been on feedback control of linear network processes with essentially-nonnegative state matrices, which encompass spread and synchronization processes. First, the input-output behavior of a network dynamics in which only one network component/channel can be actuated and another (in general remote) network component can be measured is studied. In this single-input single-output case, we characterize the infinite- and finite- zero structure in terms of the network's graph, and the actuation/measurement location. Using these results, we provide graph-theoretic conditions for determining whether the system is non-minimum phase, i.e. there are input-output zeros in the right-half plane, which provide insight into feedback-control performance and robustness. We also consider decentralized control of the network dynamics. In this case, we provide a constructive design of optimal static decentralized feedback, when control capabilities are available at a subset of the network's channels, and give graph-theoretic insights into the design.
Metrics
12 File views/ downloads
27 Record Views
Details
- Title
- Controlling Complex Network Dynamics using Sparse Measurements and Actuation: Graph-Theoretic Perspective
- Creators
- Jackeline Abad Torres
- Contributors
- Sandip Roy (Advisor)Ali Saberi (Committee Member)Mani V. Venkatasubramanian (Committee Member)
- Awarding Institution
- Washington State University
- Academic Unit
- Electrical Engineering and Computer Science, School of
- Theses and Dissertations
- Doctor of Philosophy (PhD), Washington State University
- Number of pages
- 287
- Identifiers
- 99900581733601842
- Language
- English
- Resource Type
- Dissertation