DATA-DRIVEN CONTROL OF DYNAMICAL NETWORKS: FUNDAMENTAL LIMITATIONS, ALGORITHMS, AND ROBUSTNESS GUARANTEES

Abstract

Networks are critically important in a variety of applications of interest to the DoD. Model-based techniques are traditionally used to operate these systems, requiring a detailed model of each component at all times. Importantly, accuracy and tractability of the models greatly affect the effectiveness of these methods, with the unfortunate trade-off that simple models are typically too inaccurate to be useful. In today’s data-rich world, the idea of leveraging collected data instead of, or in addition to, a cumbersome mathematical model is not only tantalizing, but also a necessity given the increasing complexity of modern distributed systems and networks. The potentials and pitfalls of model-free approaches have recently been demonstrated by the increasingly numerous applications of machine learning and artificial intelligence, for tasks ranging from autonomous driving, speech recognition, and navigation. In this project, the PI develop new theories and tools for data-driven control, secure operation, and adaptation of complex networks in contested, information-rich, and dynamic environments. While existing data-driven approches apply to non-structured systems, provide no optimality and robustness guarantees when data is maliciously corrupted or incomplete, and are task-specific, this project will undertake a rigorous approach based on control and network theories, optimization, and statistics to (i) formalize the applicability of data-driven methods for a variety of network control problems, (ii) reveal the fundamental robustness limitations of datadriven network control, and (iii) design data-driven algorithms to operate and optimize networks despite limited, corrupted, and outdated measurements.

Document Details

Document Type
DoD Grant Award
Publication Date
Aug 12, 2021
Source ID
FA95502010140

Entities

People

  • Fabio Pasqualetti

Organizations

  • Air Force Office of Scientific Research
  • United States Air Force
  • University of California Regents

Tags

Fields of Study

  • Computer science

Readers

  • Distributed Systems and Data Platform Development
  • Educational Psychology
  • Neural Network Machine Learning.

Technology Areas

  • AI & ML
  • AI & ML - Machine Learning Algorithms
  • AI & ML - Neural Networks