A Multistage Stochastic Optimization Approach for Identifying Stable and Influential Clusters in Randomly Changing Networks
Abstract
The primary task of the proposed research is to develop novel decision models and solution strategies for topological characterization of inter-connected dynamical systems or sensor networks that can maintain functionality, performance and communication in highly uncertain operational settings. We first introduce a model for finding clusters (i.e., groups of nodes) that can control a network. Then, we develop a decision framework for identifying optimal sensor (e.g., satellite) placement locations such that network-wide communications can be ensured and restored/repaired within available resources before and after observing random structural changes. Given a network topology, where the nodes represent potential placement locations, emphasis is put on finding clusters that guarantee inter-sensor communication before and after the random changes occur. Our approach leverages on advances from multistage stochastic and combinatorial optimization to identify clusters that invoke such optimal performance. Results demonstrate that effective applications will facilitate reliable communication network designs evidenced by strong failure tolerances in precarious operational environments. In a supplemental task, we develop a deep neural network learning architecture that uses synthetic aperture radar data for navigation in GPS-denied environments. The described concepts have direct interpretations in, for example, developing localization strategies for deployment of multi-agent UAV systems, satellite networks and its vulnerability analysis, and so on.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 18, 2022
- Accession Number
- AD1176897
Entities
People
- Maciej Rysz
Organizations
- Miami University