INTELLIGENT DISTRIBUTED SYSTEMS

Abstract

We propose to develop asynchronous algorithms for estimating the state of a dynamical system over a distributed time-vary agent network in the face of modeling errors. We will exploit recent advances in distributed estimation which enable one to “split” certain distributed estimators into two parts - one for which conventional tools can be used to adjust convergence rates and the other for which convergence rate can be controlled by switching and averaging. Using ideas from parameteradaptive control, we will attempt to deal with the inherent lack of robustness to modeling errors exhibited by most state estimators for unstable processes. We propose to addresses the question of how to reliably perform computations in a distributed manner to determine shared fixed points of families of functions which arise in distributed optimization and parameter estimation problems. Using recent ideas regarding the intersection of convex sets, we will explore possible approaches to defending a distributed fixed point calculator against malicious attack.

Document Details

Document Type
DoD Grant Award
Publication Date
Aug 12, 2021
Source ID
FA95502010037

Entities

People

  • A. Stephen Morse

Organizations

  • Air Force Office of Scientific Research
  • United States Air Force
  • Yale University

Tags

Fields of Study

  • Computer science

Readers

  • Applied Combinatorial Optimization and Logic Circuit Design.
  • Statistical inference.
  • Systems Analysis and Design